Wavelet-Based Cascaded Adaptive Filter for Removing Baseline Drift in Pulse Waveforms

Supervised Machine-Learning Methodology for Industrial Robot Positional Health Using Artificial Neural Networks, Discrete Wavelet Transform, and Nonlinear Indicators

Robotic systems are a fundamental part of modern industrial development. In this regard, they are required for long periods, in repetitive processes that must comply with strict tolerance ranges. Hence, the positional accuracy of the robots is critical, since degradation of this can represent a considerable loss of resources. In recent years, prognosis and health management (PHM) methodologies, based on machine and deep learning, have been applied to robots, in order to diagnose and detect faults and identify the degradation of robot positional accuracy, using external measurement systems, such as lasers and cameras; however, their implementation is complex in industrial environments. In this respect, this paper proposes a method based on discrete wavelet transform, nonlinear indices, principal component analysis, and artificial neural networks, in order to detect a positional deviation in robot joints, by analyzing the currents of the actuators. The results show that the proposed methodology allows classification of the robot positional degradation with an accuracy of 100%, using its current signals. The early detection of robot positional degradation, allows the implementation of PHM strategies on time, and prevents losses in manufacturing processes.

Towards Generating Realistic Wrist Pulse Signals Using Enhanced One Dimensional Wasserstein GAN

For the past several years, there has been an increasing focus on deep learning methods applied into computational pulse diagnosis. However, one factor restraining its development lies in the small wrist pulse dataset, due to privacy risks or lengthy experiments cost. In this study, for the first time, we address the challenging by presenting a novel one-dimension generative adversarial networks (GAN) for generating wrist pulse signals, which manages to learn a mapping strategy from a random noise space to the original wrist pulse data distribution automatically. Concretely, Wasserstein GAN with gradient penalty (WGAN-GP) is employed to alleviate the mode collapse problem of vanilla GANs, which could be able to further enhance the performance of the generated pulse data. We compared our proposed model performance with several typical GAN models, including vanilla GAN, deep convolutional GAN (DCGAN) and Wasserstein GAN (WGAN). To verify the feasibility of the proposed algorithm, we trained our model with a dataset of real recorded wrist pulse signals. In conducted experiments, qualitative visual inspection and several quantitative metrics, such as maximum mean deviation (MMD), sliced Wasserstein distance (SWD) and percent root mean square difference (PRD), are examined to measure performance comprehensively. Overall, WGAN-GP achieves the best performance and quantitative results show that the above three metrics can be as low as 0.2325, 0.0112 and 5.8748, respectively. The positive results support that generating wrist pulse data from a small ground truth is possible. Consequently, our proposed WGAN-GP model offers a potential innovative solution to address data scarcity challenge for researchers working with computational pulse diagnosis, which are expected to improve the performance of pulse diagnosis algorithms in the future.

Estimation of central pulse wave velocity from radial pulse wave analysis

BACKGROUND AND OBJECTIVE

Arterial stiffness, commonly assessed by carotid-femoral pulse wave velocity (cfPWV), is an independent biomarker for cardiovascular disease. The measurement of cfPWV, however, has been considered impractical for routine clinical application. Pulse wave analysis using a single pulse wave measurement in the radial artery is a convenient alternative. This study aims to identify pulse wave features for a more accurate estimation of cfPWV from a single radial pulse wave measurement.

METHODS

From a dataset of 140 subjects, cfPWV was measured and the radial pulse waveform was recorded for 30 s twice in succession. Features were extracted from the waveforms in the time and frequency domains, as well as by wave separation analysis. All-possible regressions with bootstrapping, McHenry's select algorithm, and support vector regression were applied to compute models for cfPWV estimation.

RESULTS

The correlation coefficients between the measured and estimated cfPWV were r = 0.81, r = 0.81, and r = 0.8 for all-possible regressions, McHenry's select algorithm, and support vector regression, respectively. The features selected by all-possible regressions are physiologically interpretable. In particular, the amplitude ratio of the diastolic peak to the notch of the radial pulse waveform (Rn,dr,P) is shown to be correlated with cfPWV. This correlation was further evaluated and found to be independent of wave reflections using a dataset (n = 3,325) of simulated pulse waves.

CONCLUSIONS

The proposed method may serve as a convenient surrogate for the measurement of cfPWV. Rn,dr,P is associated with aortic pulse wave velocity and this association may not be dependent on wave reflection.

Wrist pulse signal acquisition and analysis for disease diagnosis: A review

Pulse diagnosis (PD) plays an indispensable role in healthcare in China, India, Korea, and other Orient countries. It requires considerable training and experience to master. The results of pulse diagnosis rely heavily on the practitioner's subjective analysis, which means that the results from different physicians may be inconsistent. To overcome these drawbacks, computational pulse diagnosis (CPD) is used with advanced sensing techniques and analytical methods. Focusing on the main processes of CPD, this paper provides a systematic review of the latest advances in pulse signal acquisition, signal preprocessing, feature extraction, and signal recognition. The most relevant principles and applications are presented along with current progress. Extensive comparisons and analyses are conducted to evaluate the merits of different methods employed in CPD. While much progress has been made, a lack of datasets and benchmarks has limited the development of CPD. To address this gap and facilitate further research, we present a benchmark to evaluate different methods. We conclude with observations of the status and prospects of CPD.

Improving the accuracy and robustness of carotid-femoral pulse wave velocity measurement using a simplified tube-load model

Arterial stiffness, as measured by pulse wave velocity, for the early non-invasive screening of cardiovascular disease is becoming ever more widely used and is an independent prognostic indicator for a variety of pathologies including arteriosclerosis. Carotid-femoral pulse wave velocity (cfPWV) is regarded as the gold standard for aortic stiffness. Existing algorithms for cfPWV estimation have been shown to have good repeatability and accuracy, however, further assessment is needed, especially when signal quality is compromised. We propose a method for calculating cfPWV based on a simplified tube-load model, which allows for the propagation and reflection of the pulse wave. In-vivo cfPWV measurements from 57 subjects and numerical cfPWV data based on a one-dimensional model were used to assess the method and its performance was compared to three other existing approaches (waveform matching, intersecting tangent, and cross-correlation). The cfPWV calculated using the simplified tube-load model had better repeatability than the other methods (Intra-group Correlation Coefficient, ICC = 0.985). The model was also more accurate than other methods (deviation, 0.13 ms⁻¹) and was more robust when dealing with noisy signals. We conclude that the determination of cfPWV based on the proposed model can accurately and robustly evaluate arterial stiffness.

Quantitative Comparison of the Performance of Piezoresistive, Piezoelectric, Acceleration, and Optical Pulse Wave Sensors

The accurate measurement of the arterial pulse wave is beneficial to clinical health assessment and is important for the effective diagnosis of many types of cardiovascular disease. A variety of sensors have been developed for the non-invasive detection of these waves, but the type of sensor has an impact on the measurement results. Therefore, it is necessary to compare and analyze the signals obtained under a range of conditions using various pulse sensors to aid in making an informed choice of the appropriate type. From the available types we have selected four: a piezoresistive strain gauge sensor (PESG) and a piezoelectric Millar tonometer (the former with the ability to measure contact force), a circular film acceleration sensor, and an optical reflection sensor. Pulse wave signals were recorded from the left radial, carotid, femoral, and digital arteries of 60 subjects using these four sensors. Their performance was evaluated by analyzing their susceptibilities to external factors (contact force, measuring site, and ambient light intensity) and by comparing their stability and reproducibility. Under medium contact force, the peak-to-peak amplitude of the signals was higher than that at high and low force levels and the variability of signal waveform was small. The optical sensor was susceptible to ambient light. Analysis of the intra-class correlation coefficients (ICCs) of the pulse wave parameters showed that the tonometer and accelerometer had good stability (ICC > 0.80), and the PESG and optical sensor had moderate stability (0.46 < ICC < 0.86). Intra-observer analysis showed that the tonometer and accelerometer had good reproducibility (ICC > 0.75) and the PESG and optical sensor had moderate reproducibility (0.42 < ICC < 0.91). Inter-observer analysis demonstrated that the accelerometer had good reproducibility (ICC > 0.85) and the three other sensors had moderate reproducibility (0.52 < ICC < 0.96). We conclude that the type of sensor and measurement site affect pulse wave characteristics and the careful selection of appropriate sensor and measurement site are required according to the research and clinical need. Moreover, the influence of external factors such as contact pressure and ambient light should be fully taken into account.

An Asynchronous Hybrid Spelling Approach Based on EEG-EOG Signals for Chinese Character Input

In this paper, we presented a novel asynchronous speller for Chinese sinogram input by incorporating electroencephalography (EOG) into the conventional electroencephalography (EEG)-based spelling paradigm. An EOG-based brain switch was used to activate a classic row-column P300-based speller only when spelling was needed, enabling an asynchronous operation of the system. Then, the user could input sinograms by alternately performing P300 and double-blink tasks until he or she intended to stop spelling. With the incorporation of an EOG detector, the system achieved rapid sinogram input. In addition to asynchronous operation, the performance of the proposed speller was compared with that achieved by a P300-based method alone across 18 subjects. The proposed system showed a mean communication speed of approximately 2.39 sinograms per minute, an increase of 0.83 sinograms per minute compared with the P300-based method. The preliminary online performance indicated that the proposed paradigm is a very promising approach for practical Chinese sinogram input application. This system may also be expanded to users whose languages are written in logographic scripts to serve as an assistive communication tool.

A New Method of Adaptive Filtering and Wavelet Transform to Filter Baseline Shift

This article describes how a baseline shift is a slow change in the orientation of the baseline over time. It often exists in the process of signals sampling, e g. ECG, TLC and so on. In order to filter the baseline shift, a combination method of wavelet transform and an adaptive filter is proposed. First, the wavelet transform method is used to decompose the original ECG signal and the high-frequency components are used to as Reference input data. Then, a new adaptive filtering algorithm, P-LMS, based on the power function, is proposed to conduct adaptive noise filtering. Finally, compared with the traditional normalized least mean square algorithm (NLMS), the proposed algorithm has the characteristics of faster convergence and the effect is better. Experiments on the ECG signal in MIT-BIH database, using the method of combining P-LMS and a wavelet transform is verified to effectively filter the baseline shift and maintain the geometric characteristics of the ECG signal.

A Novel AMARS Technique for Baseline Wander Removal Applied to Photoplethysmogram

A new digital filter, AMARS (aligning minima of alternating random signal) has been derived using trigonometry to regulate signal pulsations inline. The pulses are randomly presented in continuous signals comprising frequency band lower than the signal's mean rate. Frequency selective filters are conventionally employed to reject frequencies undesired by specific applications. However, these conventional filters only reduce the effects of the rejected range producing a signal superimposed by some baseline wander (BW). In this work, filters of different ranges and techniques were independently configured to preprocess a photoplethysmogram, an optical biosignal of blood volume dynamics, producing wave shapes with several BWs. The AMARS application effectively removed the encountered BWs to assemble similarly aligned trends. The removal implementation was found repeatable in both ear and finger photoplethysmograms, emphasizing the importance of BW removal in biosignal processing in retaining its structural, functional and physiological properties. We also believe that AMARS may be relevant to other biological and continuous signals modulated by similar types of baseline volatility.

Comparison of Three Different Types of Wrist Pulse Signals by Their Physical Meanings and Diagnosis Performance

Increasing interest has been focused on computational pulse diagnosis where sensors are developed to acquire pulse signals, and machine learning techniques are exploited to analyze health conditions based on the acquired pulse signals. By far, a number of sensors have been employed for pulse signal acquisition, which can be grouped into three major categories, i.e., pressure, photoelectric, and ultrasonic sensors. To guide the sensor selection for computational pulse diagnosis, in this paper, we analyze the physical meanings and sensitivities of signals acquired by these three types of sensors. The dependence and complementarity of the different sensors are discussed from both the perspective of cardiovascular fluid dynamics and comparative experiments by evaluating disease classification performance. Experimental results indicate that each sensor is more appropriate for the diagnosis of some specific disease that the changes of physiological factors can be effectively reflected by the sensor, e.g., ultrasonic sensor for diabetes and pressure sensor for arteriosclerosis, and improved diagnosis performance can be obtained by combining three types of signals.

Pulse waveform classification using support vector machine with Gaussian time warp edit distance kernel

Advances in signal processing techniques have provided effective tools for quantitative research in traditional Chinese pulse diagnosis. However, because of the inevitable intraclass variations of pulse patterns, the automatic classification of pulse waveforms has remained a difficult problem. Utilizing the new elastic metric, that is, time wrap edit distance (TWED), this paper proposes to address the problem under the support vector machines (SVM) framework by using the Gaussian TWED kernel function. The proposed method, SVM with GTWED kernel (GTWED-SVM), is evaluated on a dataset including 2470 pulse waveforms of five distinct patterns. The experimental results show that the proposed method achieves a lower average error rate than current pulse waveform classification methods.

New assessment model of pulse depth based on sensor displacement in pulse diagnostic devices

An accurate assessment of the pulse depth in pulse diagnosis is vital to determine the floating and sunken pulse qualities (PQs), which are two of the four most basic PQs. In this work, we proposed a novel model of assessing the pulse depth based on sensor displacement (SD) normal to the skin surface and compared this model with two previous models which assessed the pulse depth using contact pressure (CP). In contrast to conventional stepwise CP variation tonometry, we applied a continuously evolving tonometric mechanism at a constant velocity and defined the pulse depth index as the optimal SD where the largest pulse amplitude was observed. By calculating the pulse depth index for 18 volunteers, we showed that the pulse was deepest at Cheok (significance level: P < 0.01), while no significant difference was found between Chon and Gwan. In contrast, the two CP-based models estimated that the pulse was shallowest at Gwan (P < 0.05). For the repeated measures, the new SD-based model showed a smaller coefficient of variation (CV ≈ 7.6%) than the two CP-based models (CV ≈ 13.5% and 12.3%, resp.). The SD-based pulse depth assessment is not sensitive to the complex geometry around the palpation locations and temperature variation of contact sensors, which allows cost-effective sensor technology.

Human motion classification based on a textile integrated and wearable sensor array

A system for classification of motion patterns is presented based on a non-contact magnetic induction monitoring device. This device is textile integrated, wearable, and able to measure pulse and respiratory activity. The proposed classifiers are a neural network, support vector machine, and a decision tree algorithm generated by bootstrap aggregating. Their performance is compared using a data set comprising five different types of motion patterns. In addition, the dependence of the misclassification error on the input sample length is investigated. The features used for classification were based on information derived by discrete wavelet transform and on lower and higher order statistical measures. With the presented magnetic induction device, all tested classifiers were able to classify the defined motion pattern with an accuracy of over 93%. The proposed bootstrap aggregating decision tree algorithm produces the best classification performance (accuracy of 96%). The support vector machine classifier shows the least dependence on the sample length.

How accurate is pulse rate variability as an estimate of heart rate variability? A review on studies comparing photoplethysmographic technology with an electrocardiogram

BACKGROUND

The usefulness of heart rate variability (HRV) as a clinical research and diagnostic tool has been verified in numerous studies. The gold standard technique comprises analyzing time series of RR intervals from an electrocardiographic signal. However, some authors have used pulse cycle intervals instead of RR intervals, as they can be determined from a pulse wave (e.g. a photoplethysmographic) signal. This option is often called pulse rate variability (PRV), and utilizing it could expand the serviceability of pulse oximeters or simplify ambulatory monitoring of HRV.

METHODS

We review studies investigating the accuracy of PRV as an estimate of HRV, regardless of the underlying technology (photoplethysmography, continuous blood pressure monitoring or Finapresi, impedance plethysmography).

RESULTS/CONCLUSIONS

Results speak in favor of sufficient accuracy when subjects are at rest, although many studies suggest that short-term variability is somewhat overestimated by PRV, which reflects coupling effects between respiration and the cardiovascular system. Physical activity and some mental stressors seem to impair the agreement of PRV and HRV, often to an inacceptable extent. Findings regarding the position of the sensor or the detection algorithm are not conclusive. Generally, quantitative conclusions are impeded by the fact that results of different studies are mostly incommensurable due to diverse experimental settings and/or methods of analysis.

Novel diagnostic model for the deficient and excess pulse qualities

The deficient and excess pulse qualities (DEPs) are the two representatives of the deficiency and excess syndromes, respectively. Despite its importance in the objectification of pulse diagnosis, a reliable classification model for the DEPs has not been reported to date. In this work, we propose a classification method for the DEPs based on a clinical study. First, through factor analysis and Fisher's discriminant analysis, we show that all the pulse amplitudes obtained at various applied pressures at Chon, Gwan, and Cheok contribute on equal orders of magnitude in the determination of the DEPs. Then, we discuss that the pulse pressure or the average pulse amplitude is appropriate for describing the collective behaviors of the pulse amplitudes and a simple and reliable classification can be constructed from either quantity. Finally, we propose an enhanced classification model that combines the two complementary variables sequentially.

Eye movement analysis for activity recognition using electrooculography

In this work, we investigate eye movement analysis as a new sensing modality for activity recognition. Eye movement data were recorded using an electrooculography (EOG) system. We first describe and evaluate algorithms for detecting three eye movement characteristics from EOG signals-saccades, fixations, and blinks-and propose a method for assessing repetitive patterns of eye movements. We then devise 90 different features based on these characteristics and select a subset of them using minimum redundancy maximum relevance (mRMR) feature selection. We validate the method using an eight participant study in an office environment using an example set of five activity classes: copying a text, reading a printed paper, taking handwritten notes, watching a video, and browsing the Web. We also include periods with no specific activity (the NULL class). Using a support vector machine (SVM) classifier and person-independent (leave-one-person-out) training, we obtain an average precision of 76.1 percent and recall of 70.5 percent over all classes and participants. The work demonstrates the promise of eye-based activity recognition (EAR) and opens up discussion on the wider applicability of EAR to other activities that are difficult, or even impossible, to detect using common sensing modalities.

Baseline wander correction in pulse waveforms using wavelet-based cascaded adaptive filter

Pulse diagnosis is a convenient, inexpensive, painless, and non-invasive diagnosis method. Quantifying pulse diagnosis is to acquire and record pulse waveforms by a set of sensor firstly, and then analyze these pulse waveforms. However, respiration and artifact motion during pulse waveform acquisition can introduce baseline wander. It is necessary, therefore, to remove the pulse waveform's baseline wander in order to perform accurate pulse waveform analysis. This paper presents a wavelet-based cascaded adaptive filter (CAF) to remove the baseline wander of pulse waveform. To evaluate the level of baseline wander, we introduce a criterion: energy ratio (ER) of pulse waveform to its baseline wander. If the ER is more than a given threshold, the baseline wander can be removed only by cubic spline estimation; otherwise it must be filtered by, in sequence, discrete Meyer wavelet filter and the cubic spline estimation. Compared with traditional methods such as cubic spline estimation, morphology filter and Linear-phase finite impulse response (FIR) least-squares-error digital filter, the experimental results on 50 simulated and 500 real pulse signals demonstrate the power of CAF filter both in removing baseline wander and in preserving the diagnostic information of pulse waveforms. This CAF filter also can be used to remove the baseline wander of other physiological signals, such as ECG and so on.

A wavelet packets approach to electrocardiograph baseline drift cancellation

Baseline wander elimination is considered a classical problem. In electrocardiography (ECG) signals, baseline drift can influence the accurate diagnosis of heart disease such as ischemia and arrhythmia. We present a wavelet-transform- (WT-) based search algorithm using the energy of the signal in different scales to isolate baseline wander from the ECG signal. The algorithm computes wavelet packet coefficients and then in each scale the energy of the signal is calculated. Comparison is made and the branch of the wavelet binary tree corresponding to higher energy wavelet spaces is chosen. This algorithm is tested using the data record from MIT/BIH database and excellent results are obtained.