Wilson Amplitude Research Articles

Recognition of Upper Limb Movements Based on Hybrid EEG and EMG Signals for Human-Robot Interaction

Upper limb amputation is a condition that severely limits the amputee’s movement. Patients who have lost the use of one or more of their upper extremities have difficulty performing activities of daily living. To help improve the control of upper limb prosthesis with pattern recognition, non-invasive approaches (EEG and EMG signals) is proposed in this paper and are integrated with machine learning techniques to recognize the upper-limb motions of subjects. EMG and EEG signals are combined, and five features are utilized to classify seven hand movements such as (wrist flexion (WF), outward part of the wrist (WE), hand open (HO), hand close (HC), pronation (PRO), supination (SUP), and rest (RST)). Experiments demonstrate that using mean absolute value (MAV), waveform length (WL), Wilson Amplitude (WAMP), Sine Slope Changes (SSC), and Cardinality features of the proposed algorithm achieves a classification accuracy of 89.6% when classifying seven distinct types of hand and wrist movement. Index Terms— Human Robot Interaction, Bio-signals Analysis, LDA classifier.

A 2-D Long Short-Term Memory Fusion Networks for Bearing Remaining Useful Life Prediction

Remaining useful life (RUL) prediction plays a significant role in prognostic and health management (PHM), and it can reduce the cost of unwanted failures and improve the reliability of industrial equipment and systems. In recent years, deep learning and sensor technology have boosted fault detection accuracy. This article proposes a two-stage prediction method based on 2-D long short-term memory (2D-LSTM) fusion networks with multisensor data for RUL prediction. This method first uses the Wilson Amplitude (WAMP) feature to automatically detect the fault occurrence time (FOT) and divide the bearing’s degradation process into two stages: health and degradation state. Then a 2D-LSTM fusion network is employed to predict the RUL of bearings, including multiple subnetworks. In each subnetwork, deep temporal features of a single sensor’s data are extracted by 2D-LSTM, which can capture both vertical and horizontal dependencies of data. Furthermore, an information fusion unit (IFU) is created to help the model incorporate features captured from each 2D-LSTM subnetwork. Experiments on two real-world bearing datasets show that our model’s effectiveness is comparable to that of other existing methods. In addition, ablation studies are performed to verify the requirement and efficacy of each component of our proposed model.

Optimization of Data Quality Related EMG Feature Extraction Parameters to Increase Hand Movement Classification Accuracy.

Many biomedical robotic interfaces (e.g., prostheses, exoskeletons) classify or estimate user movement intent based on features extracted from measured electromyograms (EMG). In most cases, the parameters of feature extraction are determined heuristically or assigned arbitrary values. We propose a more rigorous method, numerical optimization, to systematically identify parameters that maximize classification accuracy based on EMG signal characteristics. In this study, we used simulated annealing, a common global numerical optimization method, to find the optimal values of three feature extraction parameters based on the root mean square (rms) magnitude of the EMG signal. The EMG data, obtained from a public database, had been measured from 2 muscles (one hand flexor and one hand extensor) of 5 able-bodied participants performing 6 different movement tasks. Using optimization, we increased the offline movement classification accuracy by 3-5% for each participant and from 79.91% to 92.25% overall. The value of one optimized parameter (threshold of Wilson amplitude) was strongly correlated with the rms magnitude of the EMG signal (R2=0.81). Other parameters were suspected to be related to signal noise, since no strong correlation with rms magnitude was observed. Future studies will refine the optimization approach and test its practicality and effectiveness for improving online classification accuracy with robotic interfaces.

Feature Extraction and Evaluation for Classification Models of Injurious Falls Based on Surface Electromyography

Background: Only 2% of falls in older adults result in serious injuries (i.e., hip fracture). Therefore, it is important to differentiate injurious versus non-injurious falls, which is critical to develop effective interventions for injury prevention. Objects: The purpose of this study was to a. extract the best features of surface electromyography (sEMG) for classification of injurious falls, and b. find a best model provided by data mining techniques using the extracted features. Methods: Twenty young adults self-initiated falls and landed sideways. Falling trials were consisted of three initial fall directions (forward, sideways, or backward) and three knee positions at the time of hip impact (the impacting-side knee contacted the other knee (“knee together”) or the mat (“knee on mat”), or neither the other knee nor the mat was contacted by the impacting-side knee (“free knee”). Falls involved “backward initial fall direction” or “free knee” were defined as “injurious falls” as suggested from previous studies. Nine features were extracted from sEMG signals of four hip muscles during a fall, including integral of absolute value (IAV), Wilson amplitude (WAMP), zero crossing (ZC), number of turns (NT), mean of amplitude (MA), root mean square (RMS), average amplitude change (AAC), difference absolute standard deviation value (DASDV). The decision tree and support vector machine (SVM) were used to classify the injurious falls. Results: For the initial fall direction, accuracy of the best model (SVM with a DASDV) was 48%. For the knee position, accuracy of the best model (SVM with an AAC) was 49%. Furthermore, there was no model that has sensitivity and specificity of 80% or greater. Conclusion: Our results suggest that the classification model built upon the sEMG features of the four hip muscles are not effective to classify injurious falls. Future studies should consider other data mining techniques with different muscles.

Detection of Muscle Activities in the sEMG Signal by Using Frequency Features and Adaptive Decision Threshold

Reliable detection of muscle activities from the surface electromyography (sEMG) signal is an important factor that makes the sEMG controlled orthotic devices a practical tool for assisting disabled people. In spite of the advantages of employing the sEMG signal as a control signal, the changes in the amplitude characteristics of this signal due to many factors and consequent variations in the required decision threshold may impede this control paradigm from being a reliable control method for such devices. Therefore, the performance of the algorithms intended to detect muscle activities should be immune against the involuntary amplitude variations of the sEMG signal. Moreover, the decision threshold value must be adaptive to the changes in the sEMG signal characteristics to reduce the number of false alarms that may arise with the fixed threshold and lead to unintended movements to these devices. In this paper, an amplitude-independent algorithm had been developed with an adaptive decision threshold; the algorithm employed only frequency features of the sEMG signal to detect muscle activities. These features are the previously developed Adaptive Zero Crossing feature and the new proposed Adaptive Wilson Amplitude feature. The Mean Instantaneous Frequency value of the sEMG signal was used as an adaptive decision threshold value to improve the detection performance and to minimize the number of false alarms produced with the utilization of inappropriate fixed decision threshold value. A comparison with an amplitude-independent algorithm that employed fixed decision threshold had revealed an improved performance regarding the resistance against false alarms.

METHOD FOR DETERMINING A PERIODIC WAVE ITERACY OF PHOTOPLETISMOGRAM FOR BIOMETRIC IDENTIFICATION

Background. In recent years, the development of an automated identification process using biometric authentication has been observed, which has a high level of protection, since it allows you to evaluate the physical parameters and characteristics of a particular person. Such access control is more reliable since identifiers cannot be transferred to third parties or duplicated to bypass security systems. Over the past decades, a significant number of systems with biometric identification has been developed, but systems with identification according to the characteristics of the photoplethysmogram still receive little attention. The main task of biometric personality identification using photoplethysmogram is the search and implementation of machine learning methods to determine their belonging to a particular patient.Objective. The purpose of the paper is to develop an algorithm for distinguishing pulse wave iterations using the calculation of the temporal characteristics of photoplethysmograms, such as the maximum amplitude value, variance, mean absolute deviation, Wilson amplitude and the total sum of signal amplitude values.Methods. Based on the study of the temporal characteristics of the photoplethysmogram, an algorithm for distinguishing pulse wave iterations is created, which can be used for further biometric identification of a person using machine-learning methods.Results. The results can be used for further development of automated access control and management systems using biometric identification.Conclusions. Known methods of biometric identification are usually based on the static parameters of a person (the structure of the cornea of the eye, palm, fingerprints, geometry of the auricle, etc.), but have a low level of protection, since using special equipment you can create a copy of the biometric key. Therefore, today, the use of methods based on the parameters of dynamic biometric identification (plethysmogram, cardiogram and others) provides the highest degree of protection, but requires a more accurate software device to isolate and determine common symptoms. The proposed approach to calculating individual parameters of the photoplethysmogram with the aim of their subsequent classification by machine learning methods may be an acceptable solution for patient biometric identification systems.

Dynamic feature for an effective elbow-joint angle estimation based on electromyography signals

<a name="_Hlk561743"></a><span>Some physical parameters influence the electromyography signal (EMG). when the EMG signal is used to estimate the position of the elbow. An adaptable feature was important to reduce a variation on the parameters. The aim of this paper is to estimate the joint position of the elbow using EMG signal based on a dynamic function. The major contribution of this work is that the method proposed is capable of determining the elbow position using the non-pattern (NPR) recognition (PR) method. A Wilson amplitude (WAMP) which used a dynamic threshold was used to reduce the EMG signal. The dynamic threshold was generated from the root mean square (RMS) processor. With the dynamic threshold, the model could adapt to any variations on the independent variables. In order to confirm this opportunity, this work involved ten healthy male subjects to perform an experimental protocol. After a tuning and calibration process, the mean of RMS error and correlation coefficient are 9.83º±1.69º and 0.98±0.01 for a single cycle of motion, 10.39º±1.82º and 0.97±0.01 for a continuous cycle of motion and 15.19º±1.92º and 0.94±0.02 for the arbitrary gesture. For conclusion, the performance of the prediction did not significantly depend on the varying cycle of gesture (p-value>0.05). This study has confirmed that the success of the non-pattern recognition-based prediction of elbow position is adaptable to any different subjects, loads, and speed of motion</span><span>.</span>

SEMG-MMG State-Space Model for the Continuous Estimation of Multijoint Angle

Continuous joint angle estimation plays an important role in motion intention recognition and rehabilitation training. In this study, a surface electromyography- (sEMG-) mechanomyography (MMG) state-space model is proposed to estimate continuous multijoint movements from sEMG and MMG signals accurately. The model combines forward dynamics with a Hill-based muscle model that estimates joint torque only in a nonfeedback form, making the extended model capable of predicting the multijoint motion directly. The sEMG and MMG features, including the Wilson amplitude and permutation entropy, are then extracted to construct a measurement equation to reduce system error and external disturbances. Using the proposed model, a closed-loop prediction-correction approach, unscented particle filtering, is used to estimate the joint angle from sEMG and MMG signals. Comprehensive experiments are conducted on the human elbow and shoulder joint, and remarkable improvements are demonstrated compared with conventional methods.

Investigation of the Number of Features and Muscles for an Effective Hand Motion Classifier Using Electromyography Signal

The essential problem in the development of a prosthetic hand based on electromyography (EMG) signal is the choice of the features and number of muscles that will be used to recognize the hand motion. A minimal number of feature and channel could reduce the cost in the development of the prosthetic hand and processing time. Therefore, it is important to obtain the correct number of feature and muscles in the system. The objective of this study is to evaluate the effectiveness of using the number of feature and muscle to recognize the hand motion for amputee person using the EMG signal. In this study, the EMG dataset was obtained from five transradial amputee persons. Ten disposable electrodes (Ag/AgCl) were placed on the residual hand with equal space between the electrodes. In order to obtain the EMG features which are related to the hand motion, each of the EMG signal was extracted using six-time domain features which are root mean square (RMS), integrated EMG (IEMG), waveform length (WL), difference absolute standard deviation value (DASDV), Wilson amplitude, and myopulse rate (MYOP). Each feature was evaluated using Decision Tree, Linear Discriminant Analysis (LDA), Support Vector Machine (SVM), K-nearest neighborhood, and Ensemble machine learning (ML). The number of class to be classified in this study was eighteen motions. The effectiveness of using the number of features and muscles was evaluated by varying the number of features and muscles. Further, the accuracy, in the discrimination of the motion, was calculated and compared among the machine learnings. The results of this study show that six muscles can effectively classify the eighteen of hand motion. The ML with WL feature has the highest accuracy among the others (81.3% based on quadratic SVM). The study suggested the effective number of muscle and feature which can be used in the prosthetics hand development so that the effective prosthetic machine can be built with a low cost in the time computing and the hardware for data acquisition.

Adaptive threshold to compensate the effect of muscle fatigue on elbow-joint angle estimation based on electromyography

Muscle fatigue is a major problem in the study based on electromyography (EMG) signal. When the muscle in the fatigue condition, the power of the EMG signal increases significantly. This phenomenon can be a problem in the elbow-joint angle estimation. The purpose of this study is to develop a method in order to compensate the effect of the muscle fatigue on the elbow-joint angle estimation based on time domain features using EMG signal. The EMG signal was collected from biceps while the subjects performed a fatiguing motion of flexion and extension. The EMG was extracted using four time-domain features, namely zero crossing (ZC), sign slope change (SSC), Wilson amplitude (WAMP) and myopulse percentage rate (MYOP). The yielded features were filtered using second order Butterworth low pass filter. In the proposed method, to compensate the effect of the muscle fatigue, the RMS of the EMG signal was calculated for every cycle and used it as a threshold value of the features. The results show that the proposed method is able to compensate the effect of muscle fatigue with a consistent root mean square error (RMSE). The improvement of the performance ranges between 17.41% and 37.9% (for all adaptive features).

Muscle fatigue compensation of the electromyography signal for elbow joint angle estimation using adaptive feature

The purpose of this study is to develop and evaluate an adaptive feature to compensate the effect of muscle fatigue in the elbow joint angle estimation. In the experiment protocol, subjects are asked to move the elbow in flexion and extension for 15 min. The electromyography (EMG) signal collected from biceps was extracted using the Wilson amplitude (WAMP) and modified WAMP features. A modified WAMP feature consists of the standard WAMP feature and an adaptive threshold; to smooth the features, the Kalman filter was applied. The results prove that muscle fatigue affects the EMG signal. The RMSEs resulting from the standard WAMP feature in the non-fatigue and fatigue conditions are 12.41° ± 5.16° and 18.17° ± 5.53°, respectively. The RMSEs resulting from modified WAMP feature in the non-fatigue and fatigue conditions are 12.63° ± 5.22° and 13.86° ± 4.7°, respectively. This proposed method is able to compensate the effect of muscle fatigue on the elbow joint angle estimation.

Identification and classification of damaged corn kernels with impact acoustics multi-domain patterns

An impact acoustic signal device was tested with undamaged, insect-damaged, and mildew-damaged corn kernels, and the different signals were compared using ensemble empirical mode decomposition methods. These methods were adopted based on their known superiority in processing of non-stationary signals and in suppressing of mode mixing. Time domain, frequency domain, and Hilbert domain features were extracted from an ensemble empirical mode decomposition of the impact acoustic signals. Four features were extracted from the time domain: the average amplitude change, Wilson amplitude, average absolute value, and peak-to-peak value. Three features were extracted from the frequency domain: the mean square frequency, the root mean square of the power spectrum, and the frequency band variance. The energy of the high-frequency and low-frequency bands and the average values of the envelopes were extracted from the Hilbert domain. Subsequently, these features were used as inputs to a support vector machine which was optimized by particle swarm optimization. The use of hybrid features enabled higher classification accuracy than usage of features in each domain separately. In this study, achieving the classification accuracies were 99.2% for undamaged kernels, 99.6% for insect-damaged kernels and 99.3% for mildew-damaged kernels. These results, based on ensemble empirical mode decomposition and integration of multi-domain features, are encouraging for the potential of an automated inspection system.

Non-Uniform Sample Assignment in Training Set Improving Recognition of Hand Gestures Dominated with Similar Muscle Activities.

So far, little is known how the sample assignment of surface electromyogram (sEMG) features in training set influences the recognition efficiency of hand gesture, and the aim of this study is to explore the impact of different sample arrangements in training set on the classification of hand gestures dominated with similar muscle activation patterns. Seven right-handed healthy subjects (24.2 ± 1.2 years) were recruited to perform similar grasping tasks (fist, spherical, and cylindrical grasping) and similar pinch tasks (finger, key, and tape pinch). Each task was sustained for 4 s and followed by a 5-s rest interval to avoid fatigue, and the procedure was repeated 60 times for every task. sEMG were recorded from six forearm hand muscles during grasping or pinch tasks, and 4-s sEMG from each channel was segmented for empirical mode decomposition analysis trial by trial. The muscle activity was quantified with zero crossing (ZC) and Wilson amplitude (WAMP) of the first four resulting intrinsic mode function. Thereafter, a sEMG feature vector was constructed with the ZC and WAMP of each channel sEMG, and a classifier combined with support vector machine and genetic algorithm was used for hand gesture recognition. The sample number for each hand gesture was designed to be rearranged according to different sample proportion in training set, and corresponding recognition rate was calculated to evaluate the effect of sample assignment change on gesture classification. Either for similar grasping or pinch tasks, the sample assignment change in training set affected the overall recognition rate of candidate hand gesture. Compare to conventional results with uniformly assigned training samples, the recognition rate of similar pinch gestures was significantly improved when the sample of finger-, key-, and tape-pinch gesture were assigned as 60, 20, and 20%, respectively. Similarly, the recognition rate of similar grasping gestures also rose when the sample proportion of fist, spherical, and cylindrical grasping was 40, 30, and 30%, respectively. Our results suggested that the recognition rate of hand gestures can be regulated by change sample arrangement in training set, which can be potentially used to improve fine-gesture recognition for myoelectric robotic hand exoskeleton control.

Analysis of Statistical Time-Domain Features Effectiveness in Identification of Bearing Faults From Vibration Signal

The frequency of rolling element failures in induction motor is high and may lead to losses due to sudden downtime of machine. Researchers are fervent to identify an effective fault diagnosing scheme with less computational burden using optimum number of good discriminating features. We attempted time domain features, namely, waveform length (WL), slope sign changes (SSC), simple sign integral and Wilson amplitude for the first time in addition to established mean absolute value and zero crossing (ZC) for identification of mechanical faults of induction motor. Ten data sets are derived from publicly available vibration database of Case Western Reserve University to identify the capability of features in identification of faults under various conditions. The results are compared with six conventional features for tenfold cross validation using linear discriminant analysis, naive Bayes, and support vector machine classifiers. The results have shown that WL, WAMP, ZC, and SSC outperform other features. Furthermore, area under receiver operator characteristics curve analyses showed an average of 0.9987 with the proposed statistical features and 0.97618 with six conventional features. We also attempted to study the effect of data length and percentage of overlap in classification and found accuracy improves with increase in length but not significant beyond the window length of 3000 with 50% overlap. The proposed statistical features are validated using the brute force method and Laplaician method of feature selection and shown an average accuracy rate of 0.9936 and 0.9894, respectively.

Evaluation of Feature Extraction and Recognition for Activity Monitoring and Fall Detection Based on Wearable sEMG Sensors.

As an essential subfield of context awareness, activity awareness, especially daily activity monitoring and fall detection, plays a significant role for elderly or frail people who need assistance in their daily activities. This study investigates the feature extraction and pattern recognition of surface electromyography (sEMG), with the purpose of determining the best features and classifiers of sEMG for daily living activities monitoring and fall detection. This is done by a serial of experiments. In the experiments, four channels of sEMG signal from wireless, wearable sensors located on lower limbs are recorded from three subjects while they perform seven activities of daily living (ADL). A simulated trip fall scenario is also considered with a custom-made device attached to the ankle. With this experimental setting, 15 feature extraction methods of sEMG, including time, frequency, time/frequency domain and entropy, are analyzed based on class separability and calculation complexity, and five classification methods, each with 15 features, are estimated with respect to the accuracy rate of recognition and calculation complexity for activity monitoring and fall detection. It is shown that a high accuracy rate of recognition and a minimal calculation time for daily activity monitoring and fall detection can be achieved in the current experimental setting. Specifically, the Wilson Amplitude (WAMP) feature performs the best, and the classifier Gaussian Kernel Support Vector Machine (GK-SVM) with Permutation Entropy (PE) or WAMP results in the highest accuracy for activity monitoring with recognition rates of 97.35% and 96.43%. For fall detection, the classifier Fuzzy Min-Max Neural Network (FMMNN) has the best sensitivity and specificity at the cost of the longest calculation time, while the classifier Gaussian Kernel Fisher Linear Discriminant Analysis (GK-FDA) with the feature WAMP guarantees a high sensitivity (98.70%) and specificity (98.59%) with a short calculation time (65.586 ms), making it a possible choice for pre-impact fall detection. The thorough quantitative comparison of the features and classifiers in this study supports the feasibility of a wireless, wearable sEMG sensor system for automatic activity monitoring and fall detection.

Automatic detection of non-convulsive seizures: A reduced complexity approach

Detection of non-convulsive seizures (NCSz) is a challenging task because they lack convulsions, meaning no physical visible symptoms are there to detect the presence of a seizure activity. Hence their diagnosis is not easy, also continuous observation of full length EEG for the detection of non-convulsive seizures (NCSz) by an expert or a technician is a very exhaustive, time consuming job. A technique for the automatic detection of NCSz is proposed in this paper. The database used in this research was recorded at the All India Institute of Medical Sciences (AIIMS), New Delhi. 13 EEG recordings of 9 subjects consisting of a total 23 seizures of 29.42min duration were used for analysis. Normalized modified Wilson amplitude is used as a key feature to classify between normal and seizure activity. The main advantage of this study lies in the fact that no classifier is used here and hence algorithm is very simple and computationally fast. With the use of only one feature, all of the seizures under test were detected correctly, and hence the median sensitivity and specificity of 100% and 99.21% were achieved respectively.

Pattern Recognition of Eight Hand Motions Using Feature Extraction of Forearm EMG Signal

In this manuscript, eight hand motions were classified using ten different extracted features from sEMG signals. These signals were collected from four different muscles placed on the forearm. It was found out that the performance of a classifier was improved through the implementation of more than one feature. We tested two feature combinations; the classification accuracy rate of 94 % was achieved using linear discriminant analysis (LDA) based on wavelength (WAVE), Wilson amplitude (WAMP), and root mean square combination. The performance of four wavelet families was tested to select the proper wavelet family that leads to highest classification rate. Our experimental results demonstrate that the highest average classification accuracy was 95 % achieved by implementing general neural network (GRNN) classification method based on energy of wavelet coefficients (using Sym4 family). Moreover, this study investigated the performance of three SVM-kernel functions (support vector machine) and found that polynomial function is the optimal choice in most cases. The highest achieved classification accuracy was 93 % using extracted wavelet coefficients.

Hysteresis in the electromyography–force relationship: Toward an optimal model for the estimation of force

In this study we analyzed the presence of hysteresis in the relationship between features of electromyography (EMG) and force. Intramuscular EMG and surface EMG (sEMG) were recorded concurrently from the flexor digitorum profundus muscle from 0% to 100% maximum voluntary contraction (MVC) in 11 subjects. Two features, mean absolute value (MAV) and Wilson amplitude (WAMP), were computed using either the first-order (poly1) or third-order (poly3) polynomial. We detected hysteresis in the EMG-force relationship for both features in all subjects. In general, the hysteresis-based models performed better than the overall model (which does not take into account the hysteresis in the EMG-force relationship), with R(2) values about 0.98 (averaged across subjects) and root mean square error around 5% of the MVC force. These results imply the existence of a path-dependent model, which may improve the accuracy of force estimation. Muscle Nerve, 2012.