The impact of accelerometer use in exercise-associated hypoglycemia prevention in type 1 diabetes

Exercise-associated hypoglycemia is a common adverse event in people with type 1 diabetes. Previous in silico testing by our group demonstrated superior exercise-associated hypoglycemia mitigation when a predictive low glucose suspend (PLGS) algorithm was augmented to incorporate activity data. The current study investigates the effectiveness of an accelerometer-augmented PLGS algorithm in an outpatient exercise protocol. Subjects with type 1 diabetes on insulin pump therapy participated in two structured soccer sessions, one utilizing the algorithm and the other using the subject's regular basal insulin rate. Each subject wore their own insulin pump and a Dexcom G4™ Platinum continuous glucose monitor (CGM); subjects on-algorithm also wore a Zephyr BioHarness™ 3 accelerometer. The algorithm utilized a Kalman filter with a 30-minute prediction horizon. Activity and CGM readings were manually entered into a spreadsheet and at five-minute intervals, the algorithm indicated whether the basal insulin infusion should be on or suspended; any changes were then implemented by study staff. The rate of hypoglycemia during and after exercise (until the following morning) was compared between groups. Eighteen subjects (mean age 13.4 ± 3.7 years) participated in two separate sessions 7-22 days apart. The difference in meter blood glucose levels between groups at each rest period did not achieve statistical significance at any time point. Hypoglycemia during the session was recorded in three on-algorithm subjects, compared to six off-algorithm subjects. In the postexercise monitoring period, hypoglycemia occurred in two subjects who were on-algorithm during the session and four subjects who were off-algorithm. The accelerometer-augmented algorithm failed to prevent exercise-associated hypoglycemia compared to subjects on their usual basal rates. A larger sample size may have achieved statistical significance. Further research involving an automated system, a larger sample size, and an algorithm design that favors longer periods of pump suspension is necessary.

Gait assessment using the Kinect RGB-D sensor

Patients with concussions, strokes and neuromuscular disease such as Parkinson's disease, often have difficulties in keeping balance and suffer from abnormal gaits. Gait assessment conducted by a physician or therapist in clinics is standard clinical practice for assessing such injuries. However, this approach is subjective, leading to potential problems of unrepeatability, poor sensitivity and unreliability. To conduct the assessment in an objective way, a computer-based gait assessment system is designed and presented in this paper. The system performs assessments on dynamic balance and gaits by analyzing the skeleton frames of a subject captured by the Microsoft Kinect RGB-D sensor. Results show that the proposed system effectively scores subjects.

Use of smartphones and portable media devices for quantifying human movement characteristics of gait, tendon reflex response, and Parkinson's disease hand tremor

Smartphones and portable media devices are both equipped with sensor components, such as accelerometers. A software application enables these devices to function as a robust wireless accelerometer platform. The recorded accelerometer waveform can be transmitted wireless as an e-mail attachment through connectivity to the Internet. The implication of such devices as a wireless accelerometer platform is the experimental and post-processing locations can be placed anywhere in the world. Gait was quantified by mounting a smartphone or portable media device proximal to the lateral malleolus of the ankle joint. Attributes of the gait cycle were quantified with a considerable accuracy and reliability. The patellar tendon reflex response was quantified by using the device in tandem with a potential energy impact pendulum to evoke the patellar tendon reflex. The acceleration waveform maximum acceleration feature of the reflex response displayed considerable accuracy and reliability. By mounting the smartphone or portable media device to the dorsum of the hand through a glove, Parkinson's disease hand tremor was quantified and contrasted with significance to a non-Parkinson's disease steady hand control. With the methods advocated in this chapter, any aspect of human movement may be quantified through smartphones or portable media devices and post-processed anywhere in the world. These wearable devices are anticipated to substantially impact the biomedical and healthcare industry.

Quadriceps femoris spasticity in children with cerebral palsy: measurement with the pendulum test and relationship with gait abnormalities

BACKGROUND

Development of a reliable and objective test of spasticity is important for assessment and treatment of children with cerebral palsy. The pendulum test has been reported to yield reliable measurements of spasticity and to be sensitive to variations in spasticity in these children. However, the relationship between the pendulum test scores and other objective measures of spasticity has not been studied. The present study aimed to assess the effectiveness of an accelerometer-based pendulum test as a measurement of spasticity in CP, and to explore the correlation between the measurements of this test and the global index of deviation from normal gait in in children with cerebral palsy.

METHODS

We studied thirty-six children with cerebral palsy, including 18 with spastic hemiplegia and 18 with spastic diplegia, and a group of 18 typically-developing children. Knee extensor spasticity was assessed bilaterally using the accelerometer-based pendulum test and three-dimensional gait analysis. The Gillette Gait Index was calculated from the results of the gait analysis.

RESULTS

The data from the accelerometer-based pendulum test could be used to distinguish between able-bodied children and children with cerebral palsy. Additionally, two of the measurements, first swing excursion and relaxation index, could be used to differentiate the degree of knee extensor spasticity in the children with cerebral palsy. Only a few moderate correlations were found between the Gillette Gait Index and the pendulum test data.

CONCLUSIONS

This study demonstrates that the pendulum test can be used to discriminate between typically developing children and children with CP, as well as between various degrees of spasticity, such as spastic hemiplegia and spastic diplegia, in the knee extensor muscle of children with CP. Deviations from normal gait in children with CP were not correlated with the results of the pendulum test.

Reporting the reliability of accelerometer data with and without missing values

OBJECTIVES

Participants with complete accelerometer data often represent a low proportion of the total sample and, in some cases, may be distinguishable from participants with incomplete data. Because traditional reliability methods characterize the consistency of complete data, little is known about reliability properties for an entire sample. This study employed Generalizability theory to report an index of reliability characterizing complete (7 days) and observable (1 to 7 days) accelerometer data.

DESIGN

Cross-sectional.

METHODS

Accelerometer data from the Study of Early Child Care and Youth Development were analyzed in this study. Missing value analyses were conducted to describe the pattern and mechanism of missing data. Generalizability coefficients were derived from variance components to report reliability parameters for complete data and also for the entire observable sample. Analyses were conducted separately by age (9, 11, 12, and 15 yrs) and daily wear time criteria (6, 8, 10, and 12 hrs).

RESULTS

Participants with complete data were limited (<34%) and, most often, data were not considered to be missing completely at random. Across conditions, reliability coefficients for complete data were between 0.74 and 0.87. Relatively lower reliability properties were found across all observable data, ranging from 0.52 to 0.67. Sample variability increased with longer wear time criteria, but decreased with advanced age.

CONCLUSIONS

A reliability coefficient that includes all participants, not just those with complete data, provides a global perspective of reliability that could be used to further understand group level associations between activity and health outcomes.

Human fall detection on embedded platform using depth maps and wireless accelerometer

Since falls are a major public health problem in an aging society, there is considerable demand for low-cost fall detection systems. One of the main reasons for non-acceptance of the currently available solutions by seniors is that the fall detectors using only inertial sensors generate too much false alarms. This means that some daily activities are erroneously signaled as fall, which in turn leads to frustration of the users. In this paper we present how to design and implement a low-cost system for reliable fall detection with very low false alarm ratio. The detection of the fall is done on the basis of accelerometric data and depth maps. A tri-axial accelerometer is used to indicate the potential fall as well as to indicate whether the person is in motion. If the measured acceleration is higher than an assumed threshold value, the algorithm extracts the person, calculates the features and then executes the SVM-based classifier to authenticate the fall alarm. It is a 365/7/24 embedded system permitting unobtrusive fall detection as well as preserving privacy of the user.

Long-term activity recognition from wristwatch accelerometer data

With the development of wearable devices that have several embedded sensors, it is possible to collect data that can be analyzed in order to understand the user's needs and provide personalized services. Examples of these types of devices are smartphones, fitness-bracelets, smartwatches, just to mention a few. In the last years, several works have used these devices to recognize simple activities like running, walking, sleeping, and other physical activities. There has also been research on recognizing complex activities like cooking, sporting, and taking medication, but these generally require the installation of external sensors that may become obtrusive to the user. In this work we used acceleration data from a wristwatch in order to identify long-term activities. We compare the use of Hidden Markov Models and Conditional Random Fields for the segmentation task. We also added prior knowledge into the models regarding the duration of the activities by coding them as constraints and sequence patterns were added in the form of feature functions. We also performed subclassing in order to deal with the problem of intra-class fragmentation, which arises when the same label is applied to activities that are conceptually the same but very different from the acceleration point of view.

Quantitative assessment of upper limb motion in neurorehabilitation utilizing inertial sensors

Two inertial sensor systems were developed for 3-D tracking of upper limb movement. One utilizes four sensors and a kinematic model to track the positions of all four upper limb segments/joints and the other uses one sensor and a dead reckoning algorithm to track a single upper limb segment/joint. Initial evaluation indicates that the system using the kinematic model is able to track orientation to 1 degree and position to within 0.1 cm over a distance of 10 cm. The dead reckoning system combined with the "zero velocity update" correction can reduce errors introduced through double integration of errors in the estimate in offsets of the acceleration from several meters to 0.8% of the total movement distance. Preliminary evaluation of the systems has been carried out on ten healthy volunteers and the kinematic system has also been evaluated on one patient undergoing neurorehabilitation over a period of ten weeks. The initial evaluation of the two systems also shows that they can monitor dynamic information of joint rotation and position and assess rehabilitation process in an objective way, providing additional clinical insight into the rehabilitation process.

Quantitative measurement of physical activity in acute ischemic stroke and transient ischemic attack

BACKGROUND AND PURPOSE

The purpose of this study was to quantitatively measure and describe the amount and pattern of physical activity in patients within the first week after acute ischemic stroke and transient ischemic attack using accelerometers.

METHODS

A total of 100 patients with acute ischemic stroke or transient ischemic attack admitted to our acute stroke unit wore Actical accelerometers attached to both wrists and ankles and the hip for ≤7 days. Patients were included within 72 hours of symptom onset. Accelerometer output was measured in activity counts (AC). Patients were tested daily with Scandinavian Stroke Scale.

RESULTS

Physical activity peaked in the morning and declined during the rest of the day. In patients with stroke, total AC were 71% lower than in patients with transient ischemic attack. AC were 80% lower in the paretic compared with those in the nonparetic arm in patients with ischemic stroke. For the legs AC were 44% lower on the paretic side and an overall increase in AC with time was found. There was a significant increase in AC with increasing Scandinavian Stroke Scale and a decrease in AC with increasing age.

CONCLUSIONS

This study demonstrates the feasibility of using accelerometers to quantitatively and continuously measure physical activity simultaneously from all 4 extremities and the hip in patients with acute ischemic stroke and transient ischemic attack. Our study provides quantitative evidence of physical inactivity in patients with acute ischemic stroke. The method offers a low cost and noninvasive tool for future clinical interventional physiotherapeutic and early mobilization studies.

CLINICAL TRIAL REGISTRATION URL

http://www.clinicaltrials.gov. Unique identifier: NCT01560520.

Comparison of the Thorax Dynamic Responses of Small Female and Midsize Male Post Mortem Human Subjects in Side and Forward Oblique Impact Tests

Despite the increasing knowledge of the thorax mechanics in impact loadings, the effects of inter-individual differences on the mechanical response are difficult to take into account. For example, the biofidelity corridors for the small female or large male are extrapolated from the midsize male corridors. The present study reports on the results of new tests performed on small female Post Mortem Human Subjects (PMHS), and compares them with test results on midsize male PMHS. Three tests in pure side impact and three tests in forward oblique impact were performed on the thorax of small female specimens. The average weight and stature were 43 kg and 1.58 m for the small female specimens. The initial speed of the impactor was 4.3 m/s. The mass and the diameter of the impactor face were respectively 23.4 kg and 130 mm. The instrumentation and methodology was the same as for the tests published in 2008 by Trosseille et al. on midsize male specimens. The rib cages were instrumented with accelerometers on the T1, T4 and T12 vertebrae, upper and lower sternum, and the ribs were instrumented with up to 110 strain gauges. A force transducer and an accelerometer were mounted on the impactor in order to record the force applied onto the thorax. Targets fixed on vertebrae were tracked using high speed cameras in order to estimate the thoracic deflection. For the six midsize males, the test conditions were exactly the same as for the small female specimens, except for the diameter of the impactor face which was 152 mm. The average weight and stature were 70.3 kg and 1.70 m for the midsize male specimens. The force and thoracic deflection time-histories and the injury assessments are given for each specimen. The thorax force magnitude varied from 1.05 to 1.45 kN and from 1.63 to 2.34 kN, respectively for the small female and midsize male groups. The maximum deflection varied from 51 to 117 mm and from 59 to 81 mm, respectively for the small female and midsize male groups. The maximum forces appeared to be a function of the total body mass for each loading angle.

Detection of c, d, and e waves in the acceleration photoplethysmogram

Analyzing the acceleration photoplethysmogram (APG) is becoming increasingly important for diagnosis. However, processing an APG signal is challenging, especially if the goal is to detect its small components (c, d, and e waves). Accurate detection of c, d, and e waves is an important first step for any clinical analysis of APG signals. In this paper, a novel algorithm that can detect c, d, and e waves simultaneously in APG signals of healthy subjects that have low amplitude waves, contain fast rhythm heart beats, and suffer from non-stationary effects was developed. The performance of the proposed method was tested on 27 records collected during rest, resulting in 97.39% sensitivity and 99.82% positive predictivity.

Automatic annotation of seismocardiogram with high-frequency precordial accelerations

Seismocardiogram (SCG) is the low-frequency vibrations signal recorded from the chest using accelerometers. Peaks on dorsoventral and sternal SCG correspond to specific cardiac events. Prior research work has shown the potential of extracting such peaks for various types of monitoring and diagnosis applications. However, annotation of these peaks is not a trivial task and complicated in some subjects. In this paper, an automated method is proposed to annotate these peaks. The high-frequency accelerations obtained from the same accelerometer, used to record SCG with, were used to facilitate the annotation of the SCG. Algorithms were developed for detection of isovolumic moment (IM) and aortic valve closure (AC) points of SCG. Four different envelope calculation methods were used: cardiac sound characteristic waveform (CSCW), Shannon, absolute, and Hilbert. The algorithms were evaluated based on a dataset including 18 subjects undergoing lower body negative pressure and were further tested with another dataset, which included 67 subjects. These datasets had been previously manually annotated. The algorithm based on CSCW envelope calculation produced the highest detection accuracy for both IM and AC. The overall CSCW algorithm detection accuracy for the test dataset was 98.7% and 99.1% for the IM and AC points, respectively.

Ambulatory Estimation of Relative Foot Positions by Fusing Ultrasound and Inertial Sensor Data

Relative foot position estimation is important for rehabilitation, sports training and functional diagnostics. In this paper an extended Kalman filter fusing ultrasound range estimates and inertial sensors is described. With this filter several gait parameters can be estimated ambulatory. Step lengths and stride widths from 54 walking trials of three healthy subjects were estimated and compared to an optical reference. Mean ( ± standard deviation) of absolute difference was 1.7 cm ( ±1.8 cm) and 1.2 cm ( ± 1.2 cm) for step length and stride width respectively. Walking with a turn and walking around in a square area were also investigated and resulted in mean absolute differences of 1.7 cm ( ±2.0 cm) and 1.5 cm ( ± 1.5 cm) for step lengths and stride widths. In addition to these relative positions, velocities, orientations and stance and swing times can also be estimated. We conclude that the presented system is low-cost and provides a complete description of footstep kinematics and timing.

Gait characteristic analysis and identification based on the iPhone's accelerometer and gyrometer

Gait identification is a valuable approach to identify humans at a distance. In this paper, gait characteristics are analyzed based on an iPhone's accelerometer and gyrometer,and a new approach is proposed for gait identification. Specifically, gait datasets are collected by the triaxial accelerometer and gyrometer embedded in an iPhone. Then, the datasets are processed to extract gait characteristic parameters which include gait frequency, symmetry coefficient, dynamic range and similarity coefficient of characteristic curves. Finally, a weighted voting scheme dependent upon the gait characteristic parameters is proposed forgait identification. Four experiments are implemented to validate the proposed scheme. The attitude and acceleration solutions are verified by simulation. Then the gait characteristics are analyzed by comparing two sets of actual data, and the performance of the weighted voting identification scheme is verified by 40 datasets of 10 subjects.

Navigation within buildings: novel movement detection algorithms supporting people with visual impairments

This study aimed at finding simple algorithms to identify three different movements registered by accelerometer and to detect differences in the acceleration signals of people with and without visual impairments. The Tactile Acoustical Navigation and Information Assistant (TANIA) is construed to provide persons suffering from visual impairments support for an independent navigation indoors and outdoors. Attaining this goal, TANIA uses vertical acceleration signal extrema to assess its user's walking distance. This study investigated first the sit-to-stand movement, stumbling and walking up- and down stairs of 25 subjects with visual impairments using TANIA sensor system. The objective was to improve the user's movement detection using sensors to get valid and reliable data. In a second step of the study it was investigated if there is a difference between the above-mentioned movements in people with or without visual impairments (n=10). The acceleration signals of the subjects were compared. Three simple algorithms were found, which are able to separate the movement signals based on accelerometers of the respective daily movements. The second step analysis revealed a detectable difference in the second phase of stumbling (p=.034), where the subjects had to get back into walking forward. No differences in the other acceleration signals were found.

Sedentary behavior and blood pressure control among osteoarthritis initiative participants

OBJECTIVE

To examine the association between sedentary behavior and blood pressure (BP) among Osteoarthritis Initiative (OAI) participants.

DESIGN

We conducted a cross-sectional analysis of the OAI 48-month visit participants whose physical activity was measured using accelerometers. Participants were classified into four quartiles according to the percentage of wear time that was sedentary (<100 activity counts per min). Users of antihypertensive medications or non-steroidal anti-inflammatory drugs (NSAIDs) were excluded. Our main outcomes were systolic and diastolic blood pressures (SBP and DBP) and "elevated BP" defined as BP ≥ 130/85 mm Hg.

RESULTS

For this study cohort (N = 707), mean BP was 121.4 ± 15.6/74.7 ± 9.5 mm Hg and 33% had elevated BP. SBP had a graded association with increased sedentary time (P for trend = 0.02). The most sedentary quartile had 4.26 mm Hg higher SBP (95% confidence interval (CI), 0.69-7.82; P = 0.02) than the least sedentary quartile, adjusting for age, moderate-to-vigorous (MV) physical activity, and other demographic and health factors. The probability of having elevated BP significantly increased in higher sedentary quartiles (P for trend = 0.046). There were no significant findings for DBP.

CONCLUSION

A strong graded association was demonstrated between sedentary behavior and increased SBP and elevated BP, independent of time spent in MV physical activity. Reducing daily sedentary time may lead to improvement in BP and reduction in cardiovascular risk.

Objective study of sensor relevance for automatic cough detection

The development of a system for the automatic, objective, and reliable detection of cough events is a need underlined by the medical literature for years. The benefit of such a tool is clear as it would allow the assessment of pathology severity in chronic cough diseases. Even though some approaches have recently reported solutions achieving this task with a relative success, there is still no standardization about the method to adopt or the sensors to use. The goal of this paper is to study objectively the performance of several sensors for cough detection: ECG, thermistor, chest belt, accelerometer, contact, and audio microphones. Experiments are carried out on a database of 32 healthy subjects producing, in a confined room and in three situations, voluntary cough at various volumes as well as other event categories which can possibly lead to some detection errors: background noise, forced expiration, throat clearing, speech, and laugh. The relevance of each sensor is evaluated at three stages: mutual information conveyed by the features, ability to discriminate at the frame level cough from these latter other sources of ambiguity, and ability to detect cough events. In this latter experiment, with both an averaged sensitivity and specificity of about 94.5%, the proposed approach is shown to clearly outperform the commercial Karmelsonix system which achieved a specificity of 95.3% and a sensitivity of 64.9%.

Association between cardiorespiratory fitness and accelerometer-derived physical activity and sedentary time in the general population

OBJECTIVE

To determine the association between cardiorespiratory fitness and sedentary behavior, independent of exercise activity.

PATIENTS AND METHODS

We included 2223 participants (aged 12-49 years; 1053 females [47%]) without known heart disease who had both cardiovascular fitness testing and at least 1 day of accelerometer data from the National Health and Nutrition Examination Survey 2003-2004. From accelerometer data, we quantified bouts of exercise as mean minutes per day for each participant. Sedentary time was defined as less than 100 counts per minute in mean minutes per day. Cardiorespiratory fitness was derived from a submaximal exercise treadmill test. Multivariable-adjusted linear regression analyses were performed with fitness as the dependent variable. Models were stratified by sex, adjusted for age, body mass index, and wear time, and included sedentary and exercise time.

RESULTS

An additional hour of daily exercise activity time was associated with a 0.88 (0.37-1.39; P<.001) metabolic equivalent of task (MET) higher fitness for men and a 1.37 (0.43-2.31; P=.004) MET higher fitness for women. An additional hour of sedentary time was associated with a -0.12 (-0.02 to -0.22; P=.03) and a -0.24 (-0.10 to -0.38; P<.001) MET difference in fitness for men and women, respectively.

CONCLUSION

After adjustment for exercise activity, sedentary behavior appears to have an inverse association with fitness. These findings suggest that the risk related to sedentary behavior might be mediated, in part, through lower fitness levels.

An accelerometry-based methodology for assessment of real-world bilateral upper extremity activity

BACKGROUND

The use of both upper extremities (UE) is necessary for the completion of many everyday tasks. Few clinical assessments measure the abilities of the UEs to work together; rather, they assess unilateral function and compare it between affected and unaffected UEs. Furthermore, clinical assessments are unable to measure function that occurs in the real-world, outside the clinic. This study examines the validity of an innovative approach to assess real-world bilateral UE activity using accelerometry.

METHODS

Seventy-four neurologically intact adults completed ten tasks (donning/doffing shoes, grooming, stacking boxes, cutting playdough, folding towels, writing, unilateral sorting, bilateral sorting, unilateral typing, and bilateral typing) while wearing accelerometers on both wrists. Two variables, the Bilateral Magnitude and Magnitude Ratio, were derived from accelerometry data to distinguish between high- and low-intensity tasks, and between bilateral and unilateral tasks. Estimated energy expenditure and time spent in simultaneous UE activity for each task were also calculated.

RESULTS

The Bilateral Magnitude distinguished between high- and low-intensity tasks, and the Magnitude Ratio distinguished between unilateral and bilateral UE tasks. The Bilateral Magnitude was strongly correlated with estimated energy expenditure (ρ = 0.74, p<0.02), and the Magnitude Ratio was strongly correlated with time spent in simultaneous UE activity (ρ = 0.93, p<0.01) across tasks.

CONCLUSIONS

These results demonstrate face validity and construct validity of this methodology to quantify bilateral UE activity during the performance of everyday tasks performed in a laboratory setting, and can now be used to assess bilateral UE activity in real-world environments.

Evolution of accelerometer methods for physical activity research

The technology and application of current accelerometer-based devices in physical activity (PA) research allow the capture and storage or transmission of large volumes of raw acceleration signal data. These rich data not only provide opportunities to improve PA characterisation, but also bring logistical and analytic challenges. We discuss how researchers and developers from multiple disciplines are responding to the analytic challenges and how advances in data storage, transmission and big data computing will minimise logistical challenges. These new approaches also bring the need for several paradigm shifts for PA researchers, including a shift from count-based approaches and regression calibrations for PA energy expenditure (PAEE) estimation to activity characterisation and EE estimation based on features extracted from raw acceleration signals. Furthermore, a collaborative approach towards analytic methods is proposed to facilitate PA research, which requires a shift away from multiple independent calibration studies. Finally, we make the case for a distinction between PA represented by accelerometer-based devices and PA assessed by self-report.

Estimation of feature importance for food intake detection based on Random Forests classification

Selection of the most representative features is important for any pattern recognition system. This paper investigates the importance of time domain (TD) and frequency domain (FD) features used for automatic food intake detection in a wearable sensor system by using Random Forests classification. Features were extracted from signals collected using 3 different sensor modalities integrated into the Automatic Ingestion Monitor (AIM): a jaw motion sensor, a hand gesture sensor and an accelerometer. Data was collected from 12 subjects wearing AIM in free-living for a 24-hr period where they experienced unrestricted intake. Features from the sensor signals were used to train the Random Forests classifier that estimated the importance of each feature as part of the training process. Results indicated that FD features from the jaw motion signal and TD features from the accelerometer signal were the most relevant features for food intake detection.

Individual characteristics associated with mismatches between self-reported and accelerometer-measured physical activity

Background

Accurate assessment tools are required for the surveillance of physical activity (PA) levels and the assessment of the effect of interventions. In addition, increasing awareness of PA is often used as the first step in pragmatic behavioural interventions, as discrepancies between the amount of activity an individual perceives they do and the amount actually undertaken may act as a barrier to change. Previous research has demonstrated differences in the amount of activity individuals report doing, compared to their level of physical activity when measured with an accelerometer. Understanding the characteristics of those whose PA level is ranked differently when measured with either self-report or accelerometry is important as it may inform the choice of instrument for future research. The aim of this project was to determine which individual characteristics are associated with differences between self-reported and accelerometer measured physical activity.

Methods

Participant data from the 2009 wave of the Commuting and Health in Cambridge study were used. Quartiles of self-reported and accelerometer-measured PA were derived by ranking each measure from lowest to highest. These quartiles were compared to determine whether individuals’ physical activity was ranked higher by either method. Multinomial logistic regression models were used to investigate the individual characteristics associated with different categories of mismatch.

Results

Data from 486 participants (70% female) were included in the analysis. In adjusted analyses, the physical activity of overweight or obese individuals was significantly more likely to be ranked higher by self-report than by accelerometer than that of normal-weight individuals (OR = 2.07, 95%CI = 1.28–3.34), particularly among women (OR = 3.97, 95%CI = 2.11–7.47).

Conclusions

There was a greater likelihood of mismatch between self-reported and accelerometer measured physical activity levels in overweight or obese adults. Future studies in overweight or obese adults should consider employing both methods of measurement.

Fusion of smartphone motion sensors for physical activity recognition

For physical activity recognition, smartphone sensors, such as an accelerometer and a gyroscope, are being utilized in many research studies. So far, particularly, the accelerometer has been extensively studied. In a few recent studies, a combination of a gyroscope, a magnetometer (in a supporting role) and an accelerometer (in a lead role) has been used with the aim to improve the recognition performance. How and when are various motion sensors, which are available on a smartphone, best used for better recognition performance, either individually or in combination? This is yet to be explored. In order to investigate this question, in this paper, we explore how these various motion sensors behave in different situations in the activity recognition process. For this purpose, we designed a data collection experiment where ten participants performed seven different activities carrying smart phones at different positions. Based on the analysis of this data set, we show that these sensors, except the magnetometer, are each capable of taking the lead roles individually, depending on the type of activity being recognized, the body position, the used data features and the classification method employed (personalized or generalized). We also show that their combination only improves the overall recognition performance when their individual performances are not very high, so that there is room for performance improvement. We have made our data set and our data collection application publicly available, thereby making our experiments reproducible.