Sensor positioning for activity recognition using wearable accelerometers

Analyzing Optimal Wearable Motion Sensor Placement for Accurate Classification of Fall Directions

Falls represent a significant risk factor, necessitating accurate classification methods. This study aims to identify the optimal placement of wearable sensors-specifically accelerometers, gyroscopes, and magnetometers-for effective fall-direction classification. Although previous research identified optimal sensor locations for distinguishing falls from non-falls, limited attention has been given to the classification of fall direction across different body regions. This study assesses inertial measurement unit (IMU) sensors placed at 12 distinct body locations to determine the most effective positions for capturing fall-related data. The research was conducted in three phases: first, comparing classifiers across all sensor locations to identify the most effective; second, evaluating performance differences between sensors placed on the left and right sides of the body; and third, exploring the efficacy of combining sensors from the upper and lower body regions. Statistical analyses of the results for the most effective classifier model demonstrate that the support vector machine (SVM) is more effective than other classifiers across all sensor locations, with statistically significant differences in performance. At the same time, the comparison between the left and right sensor locations shows no significant performance differences within the same anatomical areas. Regarding optimal sensor placement, the findings indicate that sensors positioned on the pelvis and upper legs in the lower body, as well as on the shoulder and head in the upper body, were the most effective results for accurate fall-direction classification. The study concludes that the optimal sensor configuration for fall-direction classification involves strategically combining sensors placed on the pelvis, upper legs, and lower legs.

On-Device Semi-Supervised Activity Detection: A New Privacy-Aware Personalized Health Monitoring Approach

This paper presents an on-device semi-supervised human activity detection system that can learn and predict human activity patterns in real time. The clinical objective is to monitor and detect the unhealthy sedentary lifestyle of a user. The proposed semi-supervised learning (SSL) framework uses sparsely labelled user activity events acquired from Inertial Measurement Unit sensors installed as wearable devices. The proposed cluster-based learning model in this approach is trained with data from the same target user, thus preserving data privacy while providing personalized activity detection services. Two different cluster labelling strategies, namely, population-based and distance-based strategies, are employed to achieve the desired classification performance. The proposed system is shown to be highly accurate and computationally efficient for different algorithmic parameters, which is relevant in the context of limited computing resources on typical wearable devices. Extensive experimentation and simulation study have been conducted on multi-user human activity data from the public domain in order to analyze the trade-off between classification accuracy and computation complexity of the proposed learning paradigm with different algorithmic hyper-parameters. With 4.17 h of training time for 8000 activity episodes, the proposed SSL approach consumes at most 20 KB of CPU memory space, while providing a maximum accuracy of 90% and 100% classification rates.

A novel physical activity recognition approach using deep ensemble optimized transformers and reinforcement learning

In recent years, human physical activity recognition has increasingly attracted attention from different research fields such as healthcare, computer-human interaction, lifestyle monitoring, and athletics. Deep learning models have been extensively employed in developing physical activity recognition systems. To improve these models, their hyperparameters need to be initialized with optimal values. However, tuning these hyperparameters manually is time-consuming and may lead to inaccurate results. Moreover, the application of these models to different data resources and the integration of their results into the overall data processing pipeline are challenging issues in physical activity recognition systems. In this paper, we propose a novel ensemble method for physical activity recognition based on a deep transformer-based time-series classification model that uses heart rate, speed, and distance time-series data to recognize physical activities. In particular, we develop a modified arithmetic optimization algorithm to automatically adjust the optimal values of the classification models' hyperparameters. Moreover, a reinforcement learning-based ensemble approach is proposed to optimally integrate the results of the classification models obtained using heart rate, speed, and distance time-series data and, subsequently, recognize the physical activities. Experiments performed on a real-world dataset demonstrated that the proposed method achieves promising efficiency in comparison to other state-of-the-art models. More specifically, the proposed method increases the performance compared to the second-best performer by around 3.44 %, 9.45 %, 5.43 %, 2.54 %, and 7.53 % based on accuracy, precision, recall, specificity, and F1-score evaluation metrics, respectively.

Human Activity Recognition in a Free-Living Environment Using an Ear-Worn Motion Sensor

Human activity recognition (HAR) technology enables continuous behavior monitoring, which is particularly valuable in healthcare. This study investigates the viability of using an ear-worn motion sensor for classifying daily activities, including lying, sitting/standing, walking, ascending stairs, descending stairs, and running. Fifty healthy participants (between 20 and 47 years old) engaged in these activities while under monitoring. Various machine learning algorithms, ranging from interpretable shallow models to state-of-the-art deep learning approaches designed for HAR (i.e., DeepConvLSTM and ConvTransformer), were employed for classification. The results demonstrate the ear sensor's efficacy, with deep learning models achieving a 98% accuracy rate of classification. The obtained classification models are agnostic regarding which ear the sensor is worn and robust against moderate variations in sensor orientation (e.g., due to differences in auricle anatomy), meaning no initial calibration of the sensor orientation is required. The study underscores the ear's efficacy as a suitable site for monitoring human daily activity and suggests its potential for combining HAR with in-ear vital sign monitoring. This approach offers a practical method for comprehensive health monitoring by integrating sensors in a single anatomical location. This integration facilitates individualized health assessments, with potential applications in tele-monitoring, personalized health insights, and optimizing athletic training regimes.

A Novel Framework Based on Deep Learning Architecture for Continuous Human Activity Recognition with Inertial Sensors

Frameworks for human activity recognition (HAR) can be applied in the clinical environment for monitoring patients' motor and functional abilities either remotely or within a rehabilitation program. Deep Learning (DL) models can be exploited to perform HAR by means of raw data, thus avoiding time-demanding feature engineering operations. Most works targeting HAR with DL-based architectures have tested the workflow performance on data related to a separate execution of the tasks. Hence, a paucity in the literature has been found with regard to frameworks aimed at recognizing continuously executed motor actions. In this article, the authors present the design, development, and testing of a DL-based workflow targeting continuous human activity recognition (CHAR). The model was trained on the data recorded from ten healthy subjects and tested on eight different subjects. Despite the limited sample size, the authors claim the capability of the proposed framework to accurately classify motor actions within a feasible time, thus making it potentially useful in a clinical scenario.

AI in Rehabilitation Medicine: Opportunities and Challenges

Artificial intelligence (AI) tools are increasingly able to learn from larger and more complex data, thus allowing clinicians and scientists to gain new insights from the information they collect about their patients every day. In rehabilitation medicine, AI can be used to find patterns in huge amounts of healthcare data. These patterns can then be leveraged at the individual level, to design personalized care strategies and interventions to optimize each patient's outcomes. However, building effective AI tools requires many careful considerations about how we collect and handle data, how we train the models, and how we interpret results. In this perspective, we discuss some of the current opportunities and challenges for AI in rehabilitation. We first review recent trends in AI for the screening, diagnosis, treatment, and continuous monitoring of disease or injury, with a special focus on the different types of healthcare data used for these applications. We then examine potential barriers to designing and integrating AI into the clinical workflow, and we propose an end-to-end framework to address these barriers and guide the development of effective AI for rehabilitation. Finally, we present ideas for future work to pave the way for AI implementation in real-world rehabilitation practices.

Open-Source, Step-Counting Algorithm for Smartphone Data Collected in Clinical and Nonclinical Settings: Algorithm Development and Validation Study

BACKGROUND

Step counts are increasingly used in public health and clinical research to assess well-being, lifestyle, and health status. However, estimating step counts using commercial activity trackers has several limitations, including a lack of reproducibility, generalizability, and scalability. Smartphones are a potentially promising alternative, but their step-counting algorithms require robust validation that accounts for temporal sensor body location, individual gait characteristics, and heterogeneous health states.

OBJECTIVE

Our goal was to evaluate an open-source, step-counting method for smartphones under various measurement conditions against step counts estimated from data collected simultaneously from different body locations ("cross-body" validation), manually ascertained ground truth ("visually assessed" validation), and step counts from a commercial activity tracker (Fitbit Charge 2) in patients with advanced cancer ("commercial wearable" validation).

METHODS

We used 8 independent data sets collected in controlled, semicontrolled, and free-living environments with different devices (primarily Android smartphones and wearable accelerometers) carried at typical body locations. A total of 5 data sets (n=103) were used for cross-body validation, 2 data sets (n=107) for visually assessed validation, and 1 data set (n=45) was used for commercial wearable validation. In each scenario, step counts were estimated using a previously published step-counting method for smartphones that uses raw subsecond-level accelerometer data. We calculated the mean bias and limits of agreement (LoA) between step count estimates and validation criteria using Bland-Altman analysis.

RESULTS

In the cross-body validation data sets, participants performed 751.7 (SD 581.2) steps, and the mean bias was -7.2 (LoA -47.6, 33.3) steps, or -0.5%. In the visually assessed validation data sets, the ground truth step count was 367.4 (SD 359.4) steps, while the mean bias was -0.4 (LoA -75.2, 74.3) steps, or 0.1%. In the commercial wearable validation data set, Fitbit devices indicated mean step counts of 1931.2 (SD 2338.4), while the calculated bias was equal to -67.1 (LoA -603.8, 469.7) steps, or a difference of 3.4%.

CONCLUSIONS

This study demonstrates that our open-source, step-counting method for smartphone data provides reliable step counts across sensor locations, measurement scenarios, and populations, including healthy adults and patients with cancer.

Wearable Accelerometer and Gyroscope Sensors for Estimating the Severity of Essential Tremor

BACKGROUND

Several validated clinical scales measure the severity of essential tremor (ET). Their assessments are subjective and can depend on familiarity and training with scoring systems.

METHOD

We propose a multi-modal sensing using a wearable inertial measurement unit for estimating scores on the Fahn-Tolosa-Marin tremor rating scale (FTM) and determine the classification accuracy within the tremor type. 17 ET participants and 18 healthy controls were recruited for the study. Two movement disorder neurologists who were blinded to prior clinical information viewed video recordings and scored the FTM. Participants drew a guided Archimedes spiral while wearing an inertial measurement unit placed at the mid-point between the lateral epicondyle of the humerus and the anatomical snuff box. Acceleration and gyroscope recordings were analyzed. The ratio of the power spectral density between frequency bands 0.5-4 Hz and 4-12 Hz, and the sum of power spectrum density over the entire spectrum of 2-74 Hz, for both accelerometer and gyroscope data, were computed. FTM was estimated using regression model and classification using SVM was validated using the leave-one-out method.

RESULTS

Regression analysis showed a moderate to good correlation when individual features were used, while correlation was high ([Formula: see text] = 0.818) when suitable features of the gyro and accelerometer were combined. The accuracy for two-class classification of the combined features using SVM was 91.42% while for four-class it was 68.57%.

CONCLUSION

Potential applications of this novel wearable sensing method using a wearable Inertial Measurement Unit (IMU) include monitoring of ET and clinical trials of new treatments for the disorder.

Physical activity is low before and during hospitalisation: A secondary observational study in older Australian general medical patients

OBJECTIVES

To quantify physical activity in patients prior to and during an acute general medical hospital admission and explore relationships between mobility, pre- and in-hospital physical activity.

METHODS

This was a prospective, single-site secondary observational study conducted on general medical wards at a tertiary hospital. Prehospital physical activity was measured via the Physical Activity Scale for the Elderly (PASE; scored 0-400); in-hospital physical activity was measured via accelerometry (time at metabolic equivalents [METs] > 1.5), and mobility was measured via the de Morton Mobility Index (DEMMI). Associations were determined via Spearman's correlations.

RESULTS

Forty-six participants were included: median age 81 [76-85] years, 59% female, DEMMI on admission 39 [30-49]. Prehospital physical activity was low (PASE median 27.1 [1.6-61.9]). In-hospital physical activity was also low (0.5 [0.2-1.5] hours per day being physically active and 54 [16-194] steps per day taken). No statistically significant relationships existed between pre- and in-hospital physical activity (Spearman's rho (ρ) 0.24, 95% CI -0.08-0.53, p = 0.07). However, physical activity levels in the pre- and in-hospital settings were positively associated with patients' mobility in-hospital (Spearman's ρ 0.44, 95% CI 0.15-0.67, p = 0.002; Spearman's ρ 0.40, 95% CI 0.08-0.645, p = 0.011 respectively).

CONCLUSIONS

Physical activity is low both before and during a general medical admission. Assessment of usual physical activity patterns should be part of the clinical assessment of patients in general medicine; however, the low activity levels observed indicate a need for valid and reliable tools suitable for an older, frail cohort. Findings will inform the development of physical activity guidelines during hospitalisation.

Waist-What? Can a single sensor positioned at the waist detect parameters of gait at a speed and distance reflective of older adults' activity?

One of the problems facing an ageing population is functional decline associated with reduced levels of physical activity (PA). Traditionally researcher or clinician input is necessary to capture parameters of gait or PA. Enabling older adults to monitor their activity independently could raise their awareness of their activitiy levels, promote self-care and potentially mitigate the risks associated with ageing. The ankle is accepted as the optimum position for sensor placement to capture parameters of gait however, the waist is proposed as a more accessible body-location for older adults. This study aimed to compare step-count measurements obtained from a single inertial sensor positioned at the ankle and at the waist to that of a criterion measure of step-count, and to compare gait parameters obtained from the sensors positioned at the two different body-locations. Step-count from the waist-mounted inertial sensor was compared with that from the ankle-mounted sensor, and with a criterion measure of direct observation in healthy young and healthy older adults during a three-minute treadmill walk test. Parameters of gait obtained from the sensors at both body-locations were also compared. Results indicated there was a strong positive correlation between step-count measured by both the ankle and waist sensors and the criterion measure, and between ankle and waist sensor step-count, mean step time and mean stride time (r = .802-1.0). There was a moderate correlation between the step time variability measures at the waist and ankle (r = .405). This study demonstrates that a single sensor positioned at the waist is an appropriate method for the capture of important measures of gait and physical activity among older adults.

Relationship between physical activity and central sensitization in chronic low back pain: Insights from machine learning

BACKGROUND AND OBJECTIVES

Chronic low back pain (CLBP) is a leading cause of disability. The management guidelines for the management of CLBP often recommend optimizing physical activity (PA). Among a subsample of patients with CLBP, central sensitization (CS) is present. However, knowledge about the association between PA intensity patterns, CLBP, and CS is limited. The objective PA computed by conventional approaches (e.g. cut-points) may not be sensitive enough to explore this association. This study aimed to investigate PA intensity patterns in patients with CLBP and low or high CS (CLBP-, CLBP+, respectively) by using advanced unsupervised machine learning approach, Hidden semi-Markov model (HSMM).

METHODS

Forty-two patients were included (23 CLBP-, 19 CLBP+). CS-related symptoms (e.g. fatigue, sensitivity to light, psychological features) were assessed by a CS Inventory. Patients wore a standard 3D-accelerometer for one week and PA was recorded. The conventional cut-points approach was used to compute the time accumulation and distribution of PA intensity levels in a day. For the two groups, two HSMMs were developed to measure the temporal organization of and transition between hidden states (PA intensity levels), based on the accelerometer vector magnitude.

RESULTS

Based on the conventional cut-points approach, no significant differences were found between CLBP- and CLBP+ groups (p = 0.87). In contrast, HSMMs revealed significant differences between the two groups. For the 5 identified hidden states (rest, sedentary, light PA, light locomotion, and moderate-vigorous PA), the CLBP- group had a higher transition probability from rest, light PA, and moderate-vigorous PA states to the sedentary state (p < 0.001). In addition, the CBLP- group had a significantly shorter bout duration of the sedentary state (p < 0.001). The CLBP+ group exhibited longer durations of active (p < 0.001) and inactive states (p = 0.037) and had higher transition probabilities between active states (p < 0.001).

CONCLUSIONS

HSMM discloses the temporal organization and transitions of PA intensity levels based on accelerometer data, yielding valuable and detailed clinical information. The results imply that patients with CLBP- and CLBP+ have different PA intensity patterns. CLBP+ patients may adopt the distress-endurance response pattern with a prolonged bout duration of activity engagement.

Validation of an open-source smartphone step counting algorithm in clinical and non-clinical settings

Background

Step counts are increasingly used in public health and clinical research to assess wellbeing, lifestyle, and health status. However, estimating step counts using commercial activity trackers has several limitations, including a lack of reproducibility, generalizability, and scalability. Smartphones are a potentially promising alternative, but their step-counting algorithms require robust validation that accounts for temporal sensor body location, individual gait characteristics, and heterogeneous health states.

Objective

Our goal was to evaluate an open-source step-counting method for smartphones under various measurement conditions against step counts estimated from data collected simultaneously from different body locations ("internal" validation), manually ascertained ground truth ("manual" validation), and step counts from a commercial activity tracker (Fitbit Charge 2) in patients with advanced cancer ("wearable" validation).

Methods

We used eight independent datasets collected in controlled, semi-controlled, and free-living environments with different devices (primarily Android smartphones and wearable accelerometers) carried at typical body locations. Five datasets (N=103) were used for internal validation, two datasets (N=107) for manual validation, and one dataset (N=45) used for wearable validation. In each scenario, step counts were estimated using a previously published step-counting method for smartphones that uses raw sub-second level accelerometer data. We calculated mean bias and limits of agreement (LoA) between step count estimates and validation criteria using Bland-Altman analysis.

Results

In the internal validation datasets, participants performed 751.7±581.2 (mean±SD) steps, and the mean bias was -7.2 steps (LoA -47.6, 33.3) or -0.5%. In the manual validation datasets, the ground truth step count was 367.4±359.4 steps while the mean bias was -0.4 steps (LoA -75.2, 74.3) or 0.1 %. In the wearable validation dataset, Fitbit devices indicated mean step counts of 1931.2±2338.4, while the calculated bias was equal to -67.1 steps (LoA -603.8, 469.7) or a difference of 0.3 %.

Conclusions

This study demonstrates that our open-source step counting method for smartphone data provides reliable step counts across sensor locations, measurement scenarios, and populations, including healthy adults and patients with cancer.

Fusion-Based Body-Worn IoT Sensor Platform for Gesture Recognition of Autism Spectrum Disorder Children

The last decade's developments in sensor technologies and artificial intelligence applications have received extensive attention for daily life activity recognition. Autism spectrum disorder (ASD) in children is a neurological development disorder that causes significant impairments in social interaction, communication, and sensory action deficiency. Children with ASD have deficits in memory, emotion, cognition, and social skills. ASD affects children's communication skills and speaking abilities. ASD children have restricted interests and repetitive behavior. They can communicate in sign language but have difficulties communicating with others as not everyone knows sign language. This paper proposes a body-worn multi-sensor-based Internet of Things (IoT) platform using machine learning to recognize the complex sign language of speech-impaired children. Optimal sensor location is essential in extracting the features, as variations in placement result in an interpretation of recognition accuracy. We acquire the time-series data of sensors, extract various time-domain and frequency-domain features, and evaluate different classifiers for recognizing ASD children's gestures. We compare in terms of accuracy the decision tree (DT), random forest, artificial neural network (ANN), and k-nearest neighbour (KNN) classifiers to recognize ASD children's gestures, and the results showed more than 96% recognition accuracy.

Co-simulation of human digital twins and wearable inertial sensors to analyse gait event estimation

We propose a co-simulation framework comprising biomechanical human body models and wearable inertial sensor models to analyse gait events dynamically, depending on inertial sensor type, sensor positioning, and processing algorithms. A total of 960 inertial sensors were virtually attached to the lower extremities of a validated biomechanical model and shoe model. Walking of hemiparetic patients was simulated using motion capture data (kinematic simulation). Accelerations and angular velocities were synthesised according to the inertial sensor models. A comprehensive error analysis of detected gait events versus reference gait events of each simulated sensor position across all segments was performed. For gait event detection, we considered 1-, 2-, and 4-phase gait models. Results of hemiparetic patients showed superior gait event estimation performance for a sensor fusion of angular velocity and acceleration data with lower nMAEs (9%) across all sensor positions compared to error estimation with acceleration data only. Depending on algorithm choice and parameterisation, gait event detection performance increased up to 65%. Our results suggest that user personalisation of IMU placement should be pursued as a first priority for gait phase detection, while sensor position variation may be a secondary adaptation target. When comparing rotatory and translatory error components per body segment, larger interquartile ranges of rotatory errors were observed for all phase models i.e., repositioning the sensor around the body segment axis was more harmful than along the limb axis for gait phase detection. The proposed co-simulation framework is suitable for evaluating different sensor modalities, as well as gait event detection algorithms for different gait phase models. The results of our analysis open a new path for utilising biomechanical human digital twins in wearable system design and performance estimation before physical device prototypes are deployed.

A robust classification system for Southern Yellow cow behavior using 3-DoF accelerometers

Modern methods of monitoring help cow farmers save significantly monitoring time and improve cow health care efficiency. Behavioral changes when cows are sick may include increased or decreased daily activities such as increased lying or decreased walking time. Accelerometer advantages are low power consumption, small size, and lightweight. Thus, accelerometers have been widely used to monitor cow behavior. A cow monitoring system usually includes a central processor for receiving and processing information according to a behavioral classification algorithm through the cows’ movements. This paper introduces an effective classification system for Southern Yellow cow behavior using three degrees of freedom (3-DoF) accelerometers. The proposed classifier applied GBDT algorithm (16 seconds window) with five features, offers the good performance while investigating with four Southern Yellow cattle. The classification achievement was assessed and compared to existing ones regarding sensitivity, accuracy, and positive predictive value.

Guided regularized random forest feature selection for smartphone based human activity recognition

Human activity recognition (HAR) is an eminent area of research due to its extensive scope of applications in remote health monitoring, sports, smart home, and many more. Smartphone-based HAR systems use high-dimensional sensor data to infer human physical activities. Researchers continuously endeavor to select pertinent and non-redundant features without compromising the classification accuracy. In this work, our aim is to build an efficient HAR model that not only extracts the most relevant features from the 3-axial accelerometer and gyroscope signal data but also enhances the classification accuracy of the HAR system, without data loss using time-frequency domain features. We propose a feature selection method based on guided regularized random forest (GRRF) to determine the most pertinent and non-redundant features to reduce the time to recognize the human activities efficiently. After selecting the most relevant features, a support vector machine (SVM) is used to identify various human physical activities. The UCI public dataset and a self-collected dataset are used to assess the generalization capability and performance of the proposed feature selection method. Eventually, the accuracy reached 99.10% and 99.30% on the self-collected and UCI HAR datasets, respectively.

Real-time prediction of smoking activity using machine learning based multi-class classification model

Smoking cessation efforts can be greatly influenced by providing just-in-time intervention to individuals who are trying to quit smoking. Detecting smoking activity accurately among the confounding activities of daily living (ADLs) being monitored by the wearable device is a challenging and intriguing research problem. This study aims to develop a machine learning based modeling framework to identify the smoking activity among the confounding ADLs in real-time using the streaming data from the wrist-wearable IMU (6-axis inertial measurement unit) sensor. A low-cost wrist-wearable device has been designed and developed to collect raw sensor data from subjects for the activities. A sliding window mechanism has been used to process the streaming raw sensor data and extract several time-domain, frequency-domain, and descriptive features. Hyperparameter tuning and feature selection have been done to identify best hyperparameters and features respectively. Subsequently, multi-class classification models are developed and validated using in-sample and out-of-sample testing. The developed models obtained predictive accuracy (area under receiver operating curve) up to 98.7% for predicting the smoking activity. The findings of this study will lead to a novel application of wearable devices to accurately detect smoking activity in real-time. It will further help the healthcare professionals in monitoring their patients who are smokers by providing just-in-time intervention to help them quit smoking. The application of this framework can be extended to more preventive healthcare use-cases and detection of other activities of interest.

SUPPLEMENTARY INFORMATION

The online version contains supplementary material available at 10.1007/s11042-022-12349-6.

Communication Requirements in 5G-Enabled Healthcare Applications: Review and Considerations

Fifth generation (5G) mobile communication technology can enable novel healthcare applications and augment existing ones. However, 5G-enabled healthcare applications demand diverse technical requirements for radio communication. Knowledge of these requirements is important for developers, network providers, and regulatory authorities in the healthcare sector to facilitate safe and effective healthcare. In this paper, we review, identify, describe, and compare the requirements for communication key performance indicators in relevant healthcare use cases, including remote robotic-assisted surgery, connected ambulance, wearable and implantable devices, and service robotics for assisted living, with a focus on quantitative requirements. We also compare 5G-healthcare requirements with the current state of 5G capabilities. Finally, we identify gaps in the existing literature and highlight considerations for this space.

Upper extremity function in the free living environment of adults with traumatic brachial plexus injuries

Transition of data acquisition out of the laboratory, into the real world offers a previously inaccessible perspective of physical function. This proves to be beneficial when assessing surgical intervention, especially after a traumatic brachial plexus injury (BPI) causing loss of motor function in an upper extremity (UE). Moving towards the use of real world data in clinical practice as an outcome measure, this study developed a method to report bilateral UE activity in patients with BPI. Three groups of ten subjects each participated in this study-healthy controls, subjects with traumatic BPI prior to surgical treatment (pre-), and subjects who had surgical reconstruction to treat BPI (post-). Subjects wore four activity monitors on bilateral forearms and upper arms for four days. Tri-axial acceleration data were used to calculate asymmetry indices for forearm and upper arm usage. Analysis revealed a bimodal distribution in the post- group, prompting division of this group into two subgroups based on injury type: pan-plexus and upper trunk. While median asymmetry indices at the forearm and upper arm were decreased in the post- group when compared to the pre- group, these differences were not significant. Compared to controls, the pre-surgery group (p < 0.0001, p < 0.0001) and post-surgery group with pan-plexus injuries (p = 0.0074, p = 0.0242) both exhibited statistically significant differences in forearm and upper arm asymmetry, respectively. Further investigation to establish clinically significant differences in asymmetry index is warranted. Importantly, analyzing the activity of UEs following treatment of a BPI provides objective real world evidence of function.

Load Position Estimation Method for Wearable Devices Based on Difference in Pulse Wave Arrival Time

With the increasing use of wearable devices equipped with various sensors, information on human activities, biometrics, and surrounding environments can be obtained via sensor data at any time and place. When such devices are attached to arbitrary body parts and multiple devices are used to capture body-wide movements, it is important to estimate where the devices are attached. In this study, we propose a method that estimates the load positions of wearable devices without requiring the user to perform specific actions. The proposed method estimates the time difference between a heartbeat obtained by an ECG sensor and a pulse wave obtained by a pulse sensor, and it classifies the pulse sensor position from the estimated time difference. Data were collected at 12 body parts from four male subjects and one female subject, and the proposed method was evaluated in both user-dependent and user-independent environments. The average F-value was 1.0 when the number of target body parts was from two to five.

An Information Gain-Based Model and an Attention-Based RNN for Wearable Human Activity Recognition

Wearable sensor-based HAR (human activity recognition) is a popular human activity perception method. However, due to the lack of a unified human activity model, the number and positions of sensors in the existing wearable HAR systems are not the same, which affects the promotion and application. In this paper, an information gain-based human activity model is established, and an attention-based recurrent neural network (namely Attention-RNN) for human activity recognition is designed. Besides, the attention-RNN, which combines bidirectional long short-term memory (BiLSTM) with attention mechanism, was tested on the UCI opportunity challenge dataset. Experiments prove that the proposed human activity model provides guidance for the deployment location of sensors and provides a basis for the selection of the number of sensors, which can reduce the number of sensors used to achieve the same classification effect. In addition, experiments show that the proposed Attention-RNN achieves F1 scores of 0.898 and 0.911 in the ML (Modes of Locomotion) task and GR (Gesture Recognition) task, respectively.

Reliability and Validity of Running Cadence and Stance Time Derived from Instrumented Wireless Earbuds

Instrumented earbuds equipped with accelerometers were developed in response to limitations of currently used running wearables regarding sensor location and feedback delivery. The aim of this study was to assess test-retest reliability, face validity and concurrent validity for cadence and stance time in running. Participants wore an instrumented earbud (new method) while running on a treadmill with embedded force-plates (well-established method). They ran at a range of running speeds and performed several instructed head movements while running at a comfortable speed. Cadence and stance time were derived from raw earbud and force-plate data and compared within and between both methods using t-tests, ICC and Bland-Altman analysis. Test-retest reliability was good-to-excellent for both methods. Face validity was demonstrated for both methods, with cadence and stance time varying with speed in to-be-expected directions. Between-methods agreement for cadence was excellent for all speeds and instructed head movements. For stance time, agreement was good-to-excellent for all conditions, except while running at 13 km/h and shaking the head. Overall, the measurement of cadence and stance time using an accelerometer embedded in a wireless earbud showed good test-retest reliability, face validity and concurrent validity, indicating that instrumented earbuds may provide a promising alternative to currently used wearable systems.

HARTH: A Human Activity Recognition Dataset for Machine Learning

Existing accelerometer-based human activity recognition (HAR) benchmark datasets that were recorded during free living suffer from non-fixed sensor placement, the usage of only one sensor, and unreliable annotations. We make two contributions in this work. First, we present the publicly available Human Activity Recognition Trondheim dataset (HARTH). Twenty-two participants were recorded for 90 to 120 min during their regular working hours using two three-axial accelerometers, attached to the thigh and lower back, and a chest-mounted camera. Experts annotated the data independently using the camera's video signal and achieved high inter-rater agreement (Fleiss' Kappa =0.96). They labeled twelve activities. The second contribution of this paper is the training of seven different baseline machine learning models for HAR on our dataset. We used a support vector machine, k-nearest neighbor, random forest, extreme gradient boost, convolutional neural network, bidirectional long short-term memory, and convolutional neural network with multi-resolution blocks. The support vector machine achieved the best results with an F1-score of 0.81 (standard deviation: ±0.18), recall of 0.85±0.13, and precision of 0.79±0.22 in a leave-one-subject-out cross-validation. Our highly professional recordings and annotations provide a promising benchmark dataset for researchers to develop innovative machine learning approaches for precise HAR in free living.

A machine learning approach for semi-automatic assessment of IADL dependence in older adults with wearable sensors

BACKGROUND AND OBJECTIVE

The assessment of dependence in older adults currently requires a manual collection of data taken from questionnaires. This process is time consuming for the clinicians and intrudes the daily life of the elderly. This paper aims to semi-automate the acquisition and analysis of health data to assess and predict the dependence in older adults while executing one instrumental activity of daily living (IADL).

METHODS

In a mobile-health (m-health) scenario, we analyze whether the acquisition of data through wearables during the performance of IADLs, and with the help of machine learning techniques could replace the traditional questionnaires to evaluate dependence. To that end, we collected data from wearables, while older adults do the shopping activity. A trial supervisor (TS) labelled the different shopping stages (SS) in the collected data. We performed data pre-processing techniques over those SS and analyzed them with three machine learning algorithms: k-Nearest Neighbors (k-NN), Random Forest (RF) and Support Vector Machines (SVM).

RESULTS

Our results confirm that it is possible to replace the traditional questionnaires with wearable data. In particular, the best learning algorithm we tried reported an accuracy of 97% in the assessment of dependence. We tuned the hyperparameters of this algorithm and used embedded feature selection technique to get the best performance with a subset of only 10 features out of the initial 85. This model considers only features extracted from four sensors of a single wearable: accelerometer, heart rate, electrodermal activity and temperature. Although these features are not observational, our current proposal is semi-automatic, because it needs a TS labelling the SS (with a smartphone application). In the future, this labelling process could be automatic as well.

CONCLUSIONS

Our method can semi-automatically assess the dependence, without disturbing daily activities of elderly people. This method can save clinicians' time in the evaluation of dependence in older adults and reduce healthcare costs.

Activity Tracking Using Ear-Level Accelerometers

Introduction: By means of adding more sensor technology, modern hearing aids (HAs) strive to become better, more personalized, and self-adaptive devices that can handle environmental changes and cope with the day-to-day fitness of the users. The latest HA technology available in the market already combines sound analysis with motion activity classification based on accelerometers to adjust settings. While there is a lot of research in activity tracking using accelerometers in sports applications and consumer electronics, there is not yet much in hearing research.

Objective: This study investigates the feasibility of activity tracking with ear-level accelerometers and how it compares to waist-mounted accelerometers, which is a more common measurement location.

Method: The activity classification methods in this study are based on supervised learning. The experimental set up consisted of 21 subjects, equipped with two XSens MTw Awinda at ear-level and one at waist-level, performing nine different activities.

Results: The highest accuracy on our experimental data as obtained with the combination of Bagging and Classification tree techniques. The total accuracy over all activities and users was 84% (ear-level), 90% (waist-level), and 91% (ear-level + waist-level). Most prominently, the classes, namely, standing, jogging, laying (on one side), laying (face-down), and walking all have an accuracy of above 90%. Furthermore, estimated ear-level step-detection accuracy was 95% in walking and 90% in jogging.

Conclusion: It is demonstrated that several activities can be classified, using ear-level accelerometers, with an accuracy that is on par with waist-level. It is indicated that step-detection accuracy is comparable to a high-performance wrist device. These findings are encouraging for the development of activity applications in hearing healthcare.

Energy-Positive Activity Recognition - From Kinetic Energy Harvesting to Smart Self-Sustainable Wearable Devices

Wearable, intelligent, and unobtrusive sensor nodes that monitor the human body and the surrounding environment have the potential to create valuable data for preventive human-centric ubiquitous healthcare. To attain this vision of unobtrusiveness, the smart devices have to gather and analyze data over long periods of time without the need for battery recharging or replacement. This article presents a software-configurable kinetic energy harvesting and power management circuit that enables self-sustainable wearable devices. By exploiting the kinetic transducer as an energy source and an activity sensor simultaneously, the proposed circuit provides highly efficient context-aware control features. Its mixed-signal nano-power context awareness allows reaching energy neutrality even in energy-drought periods, thus significantly relaxing the energy storage requirements. Furthermore, the asynchronous sensing approach also doubles as a coarse-grained human activity recognition frontend. Experimental results, using commercial micro-kinetic generators, demonstrate the flexibility and potential of this approach: the circuit achieves a quiescent current of 57 nA and a maximum load current of 300 mA, delivered with a harvesting efficiency of 79%. Based on empirically collected motion data, the system achieves an energy surplus of over 232 mJ per day in a wrist-worn application while executing activity recognition at an accuracy of 89% and a latency of 60 s.

Machine-learning models for activity class prediction: A comparative study of feature selection and classification algorithms

PURPOSE

Machine-learning (ML) approaches have been repeatedly coupled with raw accelerometry to classify physical activity classes, but the features required to optimize their predictive performance are still unknown. Our aim was to identify appropriate combination of feature subsets and prediction algorithms for activity class prediction from hip-based raw acceleration data.

METHODS

The hip-based raw acceleration data collected from 27 participants was split into training (70 %) and validation (30 %) subsets. A total of 206 time- (TD) and frequencydomain (FD) features were extracted from 6-second non-overlapping windows of the signal. Feature selection was done using seven filter-based, two wrapper-based, and one embedded algorithm, and classification was performed with artificial neural network (ANN), support vector machine (SVM), and random forest (RF). For every combination between the feature selection method and the classifiers, the most appropriate feature subsets were found and used for model training within the training set. These models were then validated with the left-out validation set.

RESULTS

The appropriate number of features for the ANN, SVM, and RF ranged from 20 to 45. Overall, the accuracy of all the three classifiers was higher when trained with feature subsets generated using filter-based methods compared with when they were trained with wrapper-based methods (range: 78.1 %-88 % vs. 66 %-83.5 %). TD features that reflect how signals vary around the mean, how they differ with one another, and how much and how often they change were more frequently selected via the feature selection methods.

CONCLUSIONS

A subset of TD features from raw accelerometry could be sufficient for ML-based activity classification if properly selected from different axes.

Objective Assessment of Postoperative Morbidity After Breast Cancer Treatments with Wearable Activity Monitors: The "BRACELET" Study

Background

Current validated tools to measure upper limb dysfunction after breast cancer treatment, such as questionnaires, are prone to recall bias and do not enable comparisons between patients. This study aimed to test the feasibility of wearable activity monitors (WAMs) for achieving a continuous, objective assessment of functional recovery by measuring peri-operative physical activity (PA).

Methods

A prospective, single-center, non-randomized, observational study was conducted. Patients undergoing breast and axillary surgery were invited to wear WAMs on both wrists in the peri-operative period and then complete upper limb function (DASH) and quality-of-life (EQ-5D-5L) questionnaires. Statistical analyses were performed to determine the construct validity and concurrent validity of WAMs.

Results

The analysis included 39 patients with a mean age of 55 ± 13.2 years. Regain of function on the surgically treated side was observed to be an increase of arm activity as a percentage of preoperative levels, with the greatest increase observed between the postoperative days 1 and 2. The PA was significantly greater on the side not treated by surgery than on the surgically treated side after week 1 (mean PA, 75.8% vs. 62.3%; p < 0.0005) and week 2 (mean PA, 91.6% vs. 77.4%; p < 0.005). Subgroup analyses showed differences in recovery trends between different surgical procedures. Concurrent validity was demonstrated by a significant negative moderate correlation between the PA and DASH questionnaires (R = −0.506; p < 0.05).

Conclusion

This study demonstrated the feasibility and validity of WAMs to objectively measure postoperative recovery of upper limb function after breast surgery, providing a starting point for personalized rehabilitation through early detection of upper limb physical morbidity.

Supplementary Information

The online version contains supplementary material available at 10.1245/s10434-021-10458-4.

A Framework for Maternal Physical Activities and Health Monitoring Using Wearable Sensors

We propose a physical activity recognition and monitoring framework based on wearable sensors during maternity. A physical activity can either create or prevent health issues during a given stage of pregnancy depending on its intensity. Thus, it becomes very important to provide continuous feedback by recognizing a physical activity and its intensity. However, such continuous monitoring is very challenging during the whole period of maternity. In addition, maintaining a record of each physical activity, and the time for which it was performed, is also a non-trivial task. We aim at such problems by first recognizing a physical activity via the data of wearable sensors that are put on various parts of body. We avoid the use of smartphones for such task due to the inconvenience caused by wearing it for activities such as “eating”. In our proposed framework, a module worn on body consists of three sensors: a 3-axis accelerometer, 3-axis gyroscope, and temperature sensor. The time-series data from these sensors are sent to a Raspberry-PI via Bluetooth Low Energy (BLE). Various statistical measures (features) of this data are then calculated and represented in features vectors. These feature vectors are then used to train a supervised machine learning algorithm called classifier for the recognition of physical activity from the sensors data. Based on such recognition, the proposed framework sends a message to the care-taker in case of unfavorable situation. We evaluated a number of well-known classifiers on various features developed from overlapped and non-overlapped window size of time-series data. Our novel dataset consists of 10 physical activities performed by 61 subjects at various stages of maternity. On the current dataset, we achieve the highest recognition rate of 89% which is encouraging for a monitoring and feedback system.

Physical Activity Recognition Based on a Parallel Approach for an Ensemble of Machine Learning and Deep Learning Classifiers

Human activity recognition (HAR) by wearable sensor devices embedded in the Internet of things (IOT) can play a significant role in remote health monitoring and emergency notification to provide healthcare of higher standards. The purpose of this study is to investigate a human activity recognition method of accrued decision accuracy and speed of execution to be applicable in healthcare. This method classifies wearable sensor acceleration time series data of human movement using an efficient classifier combination of feature engineering-based and feature learning-based data representation. Leave-one-subject-out cross-validation of the method with data acquired from 44 subjects wearing a single waist-worn accelerometer on a smart textile, and engaged in a variety of 10 activities, yielded an average recognition rate of 90%, performing significantly better than individual classifiers. The method easily accommodates functional and computational parallelization to bring execution time significantly down.

A new approach for physical human activity recognition based on co-occurrence matrices

In recent years, it has been observed that many researchers have been working on different areas of detection, recognition and monitoring of human activities. The automatic determination of human physical activities is often referred to as human activity recognition (HAR). One of the most important technology that detects and tracks the activity of the human body is sensor-based HAR technology. In recent days, sensor-based HAR attracts attention in the field of computers due to its wide use in daily life and is a rapidly growing field of research. Activity recognition (AR) application is carried out by evaluating the signals obtained from various sensors placed in the human body. In this study, a new approach is proposed to extract features from sensor signals using HAR. The proposed approach is inspired by the Gray Level Co-Occurrence Matrix (GLCM) method, which is widely used in image processing, but it is applied to one-dimensional signals, unlike GLCM. Two datasets were used to test the proposed approach. The datasets were created from the signals obtained from the accelerometer, gyro and magnetometer sensors. Heralick features were obtained from co-occurrence matrix created after 1D-GLCM (One (1) Dimensional-Gray Level Co-Occurrence Matrix) was applied to the signals. HAR operation has been carried out for different scenarios using these features. Success rates of 96.66 and 93.88% were obtained for two datasets, respectively. It has been observed that the new approach proposed within the scope of the study provides high success rates for HAR applications. It is thought that the proposed approach can be used in the classification of different signals.

A Deep Learning Approach for Recognizing Activity of Daily Living (ADL) for Senior Care: Exploiting Interaction Dependency and Temporal Patterns

Ensuring the health and safety of senior citizens who live alone is a growing societal concern. The Activity of Daily Living (ADL) approach is a common means to monitor disease progression and the ability of these individuals to care for themselves. However, the prevailing sensor-based ADL monitoring systems primarily rely on wearable motion sensors, capture insufficient information for accurate ADL recognition, and do not provide a comprehensive understanding of ADLs at different granularities. Current healthcare IS and mobile analytics research focuses on studying the system, device, and provided services, and is in need of an end-to-end solution to comprehensively recognize ADLs based on mobile sensor data. This study adopts the design science paradigm and employs advanced deep learning algorithms to develop a novel hierarchical, multiphase ADL recognition framework to model ADLs at different granularities. We propose a novel 2D interaction kernel for convolutional neural networks to leverage interactions between human and object motion sensors. We rigorously evaluate each proposed module and the entire framework against state-of-the-art benchmarks (e.g., support vector machines, DeepConvLSTM, hidden Markov models, and topic-modeling-based ADLR) on two real-life motion sensor datasets that consist of ADLs at varying granularities: Opportunity and INTER. Results and a case study demonstrate that our framework can recognize ADLs at different levels more accurately. We discuss how stakeholders can further benefit from our proposed framework. Beyond demonstrating practical utility, we discuss contributions to the IS knowledge base for future design science-based cybersecurity, healthcare, and mobile analytics applications.

Classifying diverse manual material handling tasks using a single wearable sensor

The use of inertial measurement units (IMUs) for monitoring and classifying physical activities has received substantial attention in recent years, both in occupational and non-occupational contexts. However, a "user-friendly" approach is needed to promote this approach to quantify physical demands in actual workplaces. We explored the use of a single IMU for extracting information about different manual material handling (MMH) tasks (i.e., specific type of task performed, and associated duration and frequency), using a bidirectional long short-term memory network for classification. Classification performance using single IMUs placed on several body parts was compared with performance using multiple IMU configurations (2, 3, and 17 IMUs). Overall, the use of a single sensor led to satisfactory results (e.g., median accuracy >97%) in classifying MMH tasks and estimating task duration and frequency. Limited benefits were obtained using additional sensors, and several sensor locations yielded similar outcomes. Classification performance, though, was relatively inferior for push/pull vs. other tasks.

Development of the systematic observation of COVID-19 mitigation (SOCOM): Assessing face covering and distancing in schools

INTRODUCTION

During the COVID-19 pandemic, some K-12 schools resumed in-person classes with varying degrees of mitigation plans in the fall 2020. Physical distancing and face coverings can minimize SARS-CoV-2 spread, the virus that causes COVID-19. However, no research has focused on adherence to mitigation strategies during school days. Thus, we sought to develop a systematic observation protocol to capture COVID-19 mitigation strategy adherence in school environments: The Systematic Observation of COVID-19 Mitigation (SOCOM).

METHODS

We extended previously validated and internationally used tools to develop the SOCOM training and implementation protocols to assess physical-distancing and face-covering behaviors. SOCOM was tested in diverse indoor and outdoor settings (classrooms, lunchrooms, physical education [PE], and recess) among diverse schools (elementary, secondary, and special needs).

RESULTS

For the unique metrics of physical-distancing and face-covering behaviors, areas with less activity and a maximum of 10-15 students were more favorable for accurately capturing data. Overall proportion of agreement was high for physical distancing (90.9%), face covering (88.6%), activity type (89.2%), and physical activity level (87.9%). Agreement was lowest during active recess, PE, and observation areas with ≥20 students.

CONCLUSIONS

Millions of children throughout the USA are likely to return to school in the months ahead. SOCOM is a relatively inexpensive research tool that can be implemented by schools to determine mitigation strategy adherence and to assess protocols that allow students return to school safely and slow the spread of COVID-19.

Chest-Worn Inertial Sensors: A Survey of Applications and Methods

Inertial Measurement Units (IMUs) are frequently implemented in wearable devices. Thanks to advances in signal processing and machine learning, applications of IMUs are not limited to those explicitly addressing body movements such as Activity Recognition (AR). On the other hand, wearing IMUs on the chest offers a few advantages over other body positions. AR and posture analysis, cardiopulmonary parameters estimation, voice and swallowing activity detection and other measurements can be approached through chest-worn inertial sensors. This survey tries to introduce the applications that come with the chest-worn IMUs and summarizes the existing methods, current challenges and future directions associated with them. In this regard, this paper references a total number of 57 relevant studies from the last 10 years and categorizes them into seven application areas. We discuss the inertial sensors used as well as their placement on the body and their associated validation methods based on the application categories. Our investigations show meaningful correlations among the studies within the same application categories. Then, we investigate the data processing architectures of the studies from the hardware point of view, indicating a lack of effort on handling the main processing through on-body units. Finally, we propose combining the discussed applications in a single platform, finding robust ways for artifact cancellation, and planning optimized sensing/processing architectures for them, to be taken more seriously in future research.

Optimization and Validation of a Classification Algorithm for Assessment of Physical Activity in Hospitalized Patients

Low amounts of physical activity (PA) and prolonged periods of sedentary activity are common in hospitalized patients. Objective PA monitoring is needed to prevent the negative effects of inactivity, but a suitable algorithm is lacking. The aim of this study is to optimize and validate a classification algorithm that discriminates between sedentary, standing, and dynamic activities, and records postural transitions in hospitalized patients under free-living conditions. Optimization and validation in comparison to video analysis were performed in orthopedic and acutely hospitalized elderly patients with an accelerometer worn on the upper leg. Data segmentation window size (WS), amount of PA threshold (PA Th) and sensor orientation threshold (SO Th) were optimized in 25 patients, validation was performed in another 25. Sensitivity, specificity, accuracy, and (absolute) percentage error were used to assess the algorithm’s performance. Optimization resulted in the best performance with parameter settings: WS 4 s, PA Th 4.3 counts per second, SO Th 0.8 g. Validation showed that all activities were classified within acceptable limits (>80% sensitivity, specificity and accuracy, ±10% error), except for the classification of standing activity. As patients need to increase their PA and interrupt sedentary behavior, the algorithm is suitable for classifying PA in hospitalized patients.

Posture and Physical Activity Detection: Impact of Number of Sensors and Feature Type

Studies using wearable sensors to measure posture, physical activity (PA), and sedentary behavior typically use a single sensor worn on the ankle, thigh, wrist, or hip. Although the use of single sensors may be convenient, using multiple sensors is becoming more practical as sensors miniaturize.

PURPOSE

We evaluated the effect of single-site versus multisite motion sensing at seven body locations (both ankles, wrists, hips, and dominant thigh) on the detection of physical behavior recognition using a machine learning algorithm. We also explored the effect of using orientation versus orientation-invariant features on performance.

METHODS

Performance (F1 score) of PA and posture recognition was evaluated using leave-one-subject-out cross-validation on a 42-participant data set containing 22 physical activities with three postures (lying, sitting, and upright).

RESULTS

Posture and PA recognition models using two sensors had higher F1 scores (posture, 0.89 ± 0.06; PA, 0.53 ± 0.08) than did models using a single sensor (posture, 0.78 ± 0.11; PA, 0.43 ± 0.03). Models using two nonwrist sensors for posture recognition (F1 score, 0.93 ± 0.03) outperformed two-sensor models including one or two wrist sensors (F1 score, 0.85 ± 0.06). However, two-sensor models for PA recognition with at least one wrist sensor (F1 score, 0.60 ± 0.05) outperformed other two-sensor models (F1 score, 0.47 ± 0.02). Both posture and PA recognition F1 scores improved with more sensors (up to seven; 0.99 for posture and 0.70 for PA), but with diminishing performance returns. Models performed best when including orientation-based features.

CONCLUSIONS

Researchers measuring posture should consider multisite sensing using at least two nonwrist sensors, and researchers measuring PA should consider multisite sensing using at least one wrist sensor and one nonwrist sensor. Including orientation-based features improved both posture and PA recognition.

Machine Learning-Based Physical Activity Tracking with a view to Frailty Analysis

Frailty in old age is defined as the individual intrinsic susceptibility of having bad outcomes following a health problem. It relies on sarcopenia, mobility and activity. Recognizing and monitoring a range of physical activities is a necessary step which precedes the analysis of this syndrome. This paper investigates the optimal tools for this recognition in terms of type and placement of wearable sensors. Two machine learning procedures are proposed and compared on a public dataset. The first one is based on deep learning, where feature extraction is done manually, by constructing activity images from raw signals and applying convolutional neural networks to learn optimal features from these images. The second one is based on shallow learning, where hundreds of handcrafted features are extracted manually, followed by a novel feature selection approach to retain the most discriminant subset.Clinical relevance- This analysis is an indispensable prerequisite to develop efficacious way in order to identify people with frailty using sensors and moreover, to take on the challenge of frailty prevention, an actual world health organization priority.

Badminton Activity Recognition Using Accelerometer Data

A thorough analysis of sports is becoming increasingly important during the training process of badminton players at both the recreational and professional level. Nowadays, game situations are usually filmed and reviewed afterwards in order to analyze the game situation, but these video set-ups tend to be difficult to analyze, expensive, and intrusive to set up. In contrast, we classified badminton movements using off-the-shelf accelerometer and gyroscope data. To this end, we organized a data capturing campaign and designed a novel neural network using different frame sizes as input. This paper shows that with only accelerometer data, our novel convolutional neural network is able to distinguish nine activities with 86% precision when using a sampling frequency of 50 Hz. Adding the gyroscope data causes an increase of up to 99% precision, as compared to, respectively, 79% and 88% when using a traditional convolutional neural network. In addition, our paper analyses the impact of different sensor placement options and discusses the impact of different sampling frequenciess of the sensors. As such, our approach provides a low cost solution that is easy to use and can collect useful information for the analysis of a badminton game.

Estimating wearable motion sensor performance from personal biomechanical models and sensor data synthesis

We present a fundamentally new approach to design and assess wearable motion systems based on biomechanical simulation and sensor data synthesis. We devise a methodology of personal biomechanical models and virtually attach sensor models to body parts, including sensor positions frequently considered for wearable devices. The simulation enables us to synthesise motion sensor data, which is subsequently considered as input for gait marker estimation algorithms. We evaluated our methodology in two case studies, including running athletes and hemiparetic patients. Our analysis shows that running speed affects gait marker estimation performance. Estimation error of stride duration varies between athletes across 834 simulated sensor positions and can soar up to 54%, i.e. 404 ms. In walking patients after stroke, we show that gait marker performance differs between affected and less-affected body sides and optimal sensor positions change over a period of movement therapy intervention. For both case studies, we observe that optimal gait marker estimation performance benefits from personally selected sensor positions and robust algorithms. Our methodology enables wearable designers and algorithm developers to rapidly analyse the design options and create personalised systems where needed, e.g. for patients with movement disorders.

A novel gait parameter estimation method for healthy adults and postoperative patients with an ear-worn sensor

OBJECTIVE

Gait analysis helps to assess recovery during rehabilitation. Previous gait analysis studies are primarily applicable to healthy subjects or to postoperative patients. The purpose of this paper is to construct a new gait parameter estimation platform based on an ear-worn activity recognition (e-AR) sensor, which can be used for both normal and pathological gait signals.

APPROACH

Thirty healthy adults and eight postoperative patients participated in the experiment. A method based on singular spectrum analysis (SSA) and iterative mean filtering (IMF) is proposed to detect gait events and estimate three key gait parameters, i.e. stride time, swing time, and stance time.

MAIN RESULTS

Experimental results show that the estimated gait parameters provided by the proposed method are very close to the gait parameters provided by the gait assessment system. For normal gait signals, the average absolute errors of stride, swing, and stance time are 27.8 ms, 35.8 ms, and 37.5 ms, respectively. For pathological gait signals, the average absolute error of stride time is 32.1 ms.

SIGNIFICANCE

The proposed parameter estimation method can be applied to both general analysis for healthy subjects and rehabilitation evaluation for postoperative patients. The convenience and comfort of the ear-worn sensor increase its potential for practical applications.

Heart Rate Variability and Accelerometry as Classification Tools for Monitoring Perceived Stress Levels-A Pilot Study on Firefighters

Chronic stress is the main cause of health problems in high-risk jobs. Wearable sensors can become an ecologically valid method of stress level assessment in real-life applications. We sought to determine a non-invasive technique for objective stress monitoring. Data were collected from firefighters during 24-h shifts using sensor belts equipped with a dry-lead electrocardiograph (ECG) and a three-axial accelerometer. Levels of stress experienced during fire incidents were evaluated via a brief self-assessment questionnaire. Types of physical activity were distinguished basing on accelerometer readings, and heart rate variability (HRV) time series were segmented accordingly into corresponding fragments. Those segments were classified as stress/no-stress conditions. Receiver Operating Characteristic (ROC) analysis showed true positive classification as stress condition for 15% of incidents (while maintaining almost zero False Positive Rate), which parallels the amount of truly stressful incidents reported in the questionnaires. These results show a firm correspondence between the perceived stress level and physiological data. Psychophysiological measurements are reliable indicators of stress even in ecological settings and appear promising for chronic stress monitoring in high-risk jobs, such as firefighting.

STUDY OF GAIT PATTERN RECOGNITION BASED ON FUSION OF MECHANOMYOGRAPHY AND ATTITUDE ANGLE SIGNAL

The study of lower limb movements plays an important role in many fields, such as rehabilitation and treatment of disabled patients, detection, and monitoring of daily life, as well as the interaction between people and machine, like the application of intelligent prosthetics. In this paper, the wireless device was used to collect the mechanomyography (MMG) signals of four thigh muscles (rectus femoris, vastus lateralis, vastus medialis, and semitendinosus) and the attitude angle of rectus femoris. High precision was achieved in 11 gait movements, including 3 static activities, 4 dynamic transition activities, and 4 dynamic activities. It has been verified that the hidden Markov model (HMM) could not only be applied to the MMG-based gait recognition with high veracity but also support comparative analysis between support vector machine (SVM) and quadratic discriminant analysis (QDA). In addition, the experiment was conducted from the perspectives of feature selections, channel combinations, and muscle contribution rates. The results show that the average classification accuracy of dynamic motions based on MMG is 98.27%, while based on attitude angle, the average recognition rate of static motions and dynamic transition motions could achieve 98.33% and 100%, respectively. Generally, the average recognition rate of 11 gait motions is 98.91%.

The new design of cows' behavior classifier based on acceleration data and proposed feature set

Monitor and classify behavioral activities in cows is a helpful support solution for livestock based on the analysis of data from sensors attached to the animal. Accelerometers are particularly suited for monitoring cow behaviors due to small size, lightweight and high accuracy. Nevertheless, the interpretation of the data collected by such sensors when characterizing the type of behaviors still brings major challenges to developers, related to activity complexity (i.e., certain behaviors contain similar gestures). This paper presents a new design of cows' behavior classifier based on acceleration data and proposed feature set. Analysis of cow acceleration data is used to extract features for classification using machine learning algorithms. We found that with 5 features (mean, standard deviation, root mean square, median, range) and 16-second window of data (1 sample/second), classification of seven cow behaviors (including feeding, lying, standing, lying down, standing up, normal walking, active walking) achieved the overall highest performance. We validated the results with acceleration data from a public source. Performance of our proposed classifier was evaluated and compared to existing ones in terms of the sensitivity, the accuracy, the positive predictive value, and the negative predictive value.

Concurrent validity of Nokia Go activity tracker in walking and free-living conditions

RATIONALE, AIMS, AND OBJECTIVES

Consumer-based activity trackers aim at quantifying physical activity in a wide range of contexts. Nevertheless, they need to be validated before they are confidently used. This study assessed the concurrent validity of the Nokia Go against reference devices, according to different sensor locations, in two measurement conditions: during a walking task and during a 24-hour free-living condition.

METHODS

We examined the agreement between devices and between locations in the number of steps and total sleep time by using intraclass correlation coefficient and Bland-Altman method.

RESULTS

In the walking task, the agreement is good to excellent for steps between the Nokia Go and the reference device. In the free-living condition, there is a systematic underestimation of steps in comparison with the ActiGraph. Excellent agreement was found between locations. The device worn at the hip indicated the lowest number of steps, and the device located at the dominant wrist indicated the greatest number of steps.

CONCLUSIONS

There are high discrepancies in step count between devices because of the different types of activities in daily life. The Nokia Go may be confidently used for step counting during pure walking tasks, at different locations. However, the lack of concurrent validity with ActiGraph call for caution regarding their use in daily living conditions.

Sensor Data Required for Automatic Recognition of Athletic Tasks Using Deep Neural Networks

Movement screens are used to assess the overall movement quality of an athlete. However, these rely on visual observation of a series of movements and subjective scoring. Data-driven methods to provide objective scoring of these movements are being developed. These currently use optical motion capture and require manual pre-processing of data to identify the start and end points of each movement. Therefore, we aimed to use deep learning techniques to automatically identify movements typically found in movement screens and assess the feasibility of performing the classification based on wearable sensor data. Optical motion capture data were collected on 417 athletes performing 13 athletic movements. We trained an existing deep neural network architecture that combines convolutional and recurrent layers on a subset of 278 athletes. A validation subset of 69 athletes was used to tune the hyperparameters and the final network was tested on the remaining 70 athletes. Simulated inertial measurement data were generated based on the optical motion capture data and the network was trained on this data for different combinations of body segments. Classification accuracy was similar for networks trained using the optical and full-body simulated inertial measurement unit data at 90.1 and 90.2%, respectively. A good classification accuracy of 85.9% was obtained using as few as three simulated sensors placed on the torso and shanks. However, using three simulated sensors on the torso and upper arms or fewer than three sensors resulted in poor accuracy. These results for simulated sensor data indicate the feasibility of classifying athletic movements using a small number of wearable sensors. This could facilitate objective data-driven methods that automatically score overall movement quality using wearable sensors to be easily implemented in the field.