Pornchai Phukpattaranont

Feature reduction and selection for EMG signal classification

Feature extraction is a significant method to extract the useful information which is hidden in surface electromyography (EMG) signal and to remove the unwanted part and interferences. To be successful in classification of the EMG signal, selection of a feature vector ought to be carefully considered. However, numerous studies of the EMG signal classification have used a feature set that have contained a number of redundant features. In this study, most complete and up-to-date thirty-seven time domain and frequency domain features have been proposed to be studied their properties. The results, which were verified by scatter plot of features, statistical analysis and classifier, indicated that most time domain features are superfluity and redundancy. They can be grouped according to mathematical property and information into four main types: energy and complexity, frequency, prediction model, and time-dependence. On the other hand, all frequency domain features are calculated based on statistical parameters of EMG power spectral density. Its performance in class separability viewpoint is not suitable for EMG recognition system. Recommendation of features to avoid the usage of redundant features for classifier in EMG signal classification applications is also proposed in this study.

Application of Wavelet Analysis in EMG Feature Extraction for Pattern Classification

Application of Wavelet Analysis in EMG Feature Extraction for Pattern Classification Nowadays, analysis of electromyography (EMG) signal using wavelet transform is one of the most powerful signal processing tools. It is widely used in the EMG recognition system. In this study, we have investigated usefulness of extraction of the EMG features from multiple-level wavelet decomposition of the EMG signal. Different levels of various mother wavelets were used to obtain the useful resolution components from the EMG signal. Optimal EMG resolution component (sub-signal) was selected and then the reconstruction of the useful information signal was done. Noise and unwanted EMG parts were eliminated throughout this process. The estimated EMG signal that is an effective EMG part was extracted with the popular features, i.e. mean absolute value and root mean square, in order to improve quality of class separability. Two criteria used in the evaluation are the ratio of a Euclidean distance to a standard deviation and the scatter graph. The results show that only the EMG features extracted from reconstructed EMG signals of the first-level and the second-level detail coefficients yield the improvement of class separability in feature space. It will ensure that the result of pattern classification accuracy will be as high as possible. Optimal wavelet decomposition is obtained using the seventh order of Daubechies wavelet and the forth-level wavelet decomposition.

An optimal wavelet function based on wavelet denoising for multifunction myoelectric control

The aim of this study was to investigate and select the wavelet function that is optimum to denoise the surface electromyography (sEMG) signal for multifunction myoelectric control. Wavelet denoising algorithm has been used to find the optimal wavelet function for removing white Gaussian noise (WGN) at various signal-to-noise ratios (SNRs) from sEMG signals. A total of 53 wavelet functions were used in evaluation of the denoised performance. The wavelets are Daubechies, Symlets, Coiflet, BiorSplines, ReverseBior, and Discrete Meyer. Universal thresholding method has been used to estimate threshold value. Soft, hard, hyperbolic, and garrote thresholding are applied. Evaluations of the performance of these algorithms are mean squared error (MSE). The results show that the best wavelet functions for denoising are the first order of Daubechies, BioSplines, and ReverseBior wavelets (db1, bior1.1, rbio1.1). Various families can be used except the third order of decomposition of BiorSplines (bior3.1, bior3.3, bior3.5, bior3.7, bior3.9) and Discrete Meyer (dmey) are not recommended to use in wavelet denoising of sEMG signal. In addition, performance of soft thresholding is better than the others modified thresholding.

Post-beamforming second-order Volterra filter for pulse-echo ultrasonic imaging

We present a new algorithm for deriving a second-order Volterra filter (SVF) capable of separating linear and quadratic components from echo signals. Images based on the quadratic components are shown to provide contrast enhancement between tissue and ultrasound contrast agents (UCAs) without loss in spatial resolution. It is also shown that the quadratic images preserve the low scattering regions due to their high dynamic range when compared with standard B-mode or harmonic images. A robust algorithm for deriving the filter has been developed and tested on real-time imaging data from contrast and tissue-mimicking media. Illustrative examples from image targets containing contrast agent and tissue-mimicking media are presented and discussed. Quantitative assessment of the contrast enhancement is performed on both the RF data and the envelope-detected log-compressed image data. It is shown that the quadratic images offer levels of enhancement comparable or exceeding those from harmonic filters while maintaining the visibility of low scattering regions of the image.

The Usefulness of Mean and Median Frequencies in Electromyography Analysis

Rich useful information can be obtained from the muscles and researchers can use such information in a wide class of clinical and engineering applications by measuring surface electromyography (EMG) signals (Merletti & Parker, 2004). Normally, EMG signals are acquired by surface electrodes that are placed on the skin superimposed on the targeted muscle. In order to use the EMG signal as a diagnosis signal or a control signal, a feature is often extracted before performing analysis or classification stage (Phinyomark et al., 2012a) because a lot of information, both useful information and noise (Phinyomark et al., 2012b), is contained in the raw EMG data. An EMG feature is a distinct characteristic of the signal that can be described or observed quantitatively, such as being large or small, spiky or smooth, and fast or slow. Generally, EMG features can be computed in numerical form from a finite length time interval and can change as a function of time, i.e. a voltage or a frequency. They can be computed in several domains, such as time domain, frequency domain, timefrequency and time-scale representations (Boostani & Moradi, 2003). However, frequencydomain features show the better performance than other-domain features in case of the assessing muscle fatigue (Al-Mulla et al., 2012). Mean frequency (MNF) and median frequency (MDF) are the most useful and popular frequency-domain features (Phinyomark et al., 2009) and frequently used for the assessment of muscle fatigue in surface EMG signals (Cifrek et al., 2009).

A Comparative Study of Wavelet Denoising for Multifunction Myoelectric Control

The aim of this study was to investigate the application of wavelet denoising in noise reduction for multifunction myoelectric control system. Six upper limb motions including hand open, hand close, wrist extension, wrist flexion, pronation, and supination. For each motion, two channels of electrodes were applied. A comparative study of four classical denoising algorithms including universal thresholding, SURE thresholding, hybrid thresholding, and minimax thresholding have been used to remove white Gaussian noise at various signal-to-noise ratios (SNRs) from EMG signals. Applications of soft and hard thresholding as well as threshold rescaling methods were considered and the whole procedures of noise reduction were applied with different wavelet functions and different decomposition levels. Evaluations of the performance of noise reduction are determined using mean squared error (MSE). The results show that Daubechies wavelet with second orders (db2) provides marginally better performance than other possibilities. Suitable number of decomposition levels is four. Universal and soft thresholding is the best of wavelet denoising algorithms from eight possible denoising processes under investigation. In addition, the threshold using a level-dependent estimation of level noise showed better than others.

A Review of Control Methods for Electric Power Wheelchairs Based on Electromyography Signals with Special Emphasis on Pattern Recognition

Abstract Electric Power Wheelchairs (EPWs) are becoming increasingly important in assistive technology and rehabilitation devices. Normally EPWs are controlled by a joystick. However, this may not be suitable for disabled people who lack full control of their upper-limbs. Recent advances in the control of EPWs based on electromyography (EMG) signals are able to meet the needs of users with restricted limb movement and provide high performance control. Hence, EPWs controlled by EMG signals are highly appropriate for elderly and disabled users. The purpose of this article is to review the state-of-the-art of EMG controlled EPWs and to present the achievements so far in this technology. A study of a variety of methods for EMG-based control in literature was studied here. Two types of control methods for EPWs, pattern recognition and hybrid recognition systems are discussed. Four major criteria are applied to compare the quality of control resulting from the use of these control methods: Accuracy of control, response time or real-time operation, robustness, and intuitiveness of control. Based on these four criteria, the use of the support vector machine classifier using features based on the time domain such as mean absolute value, waveform length, and zero crossing are suggested for the pattern recognition method. Furthermore, a combination of the pattern recognition and non-pattern recognition methods is recommended in order to increase the control commands by use of a small number of muscle positions.

Fractal analysis features for weak and single-channel upper-limb EMG signals

Electromyography (EMG) signals are the electrical manifestations of muscle contractions. EMG signals may be weak or at a low level when there is only a small movement in the major corresponding muscle group or when there is a strong movement in the minor corresponding muscle group. Moreover, in a single-channel EMG classification identifying the signals may be difficult. However, weak and single-channel EMG control systems offer a very convenient way of controlling human-computer interfaces (HCIs). Identifying upper-limb movements using a single-channel surface EMG also has a number of rehabilitation and HCI applications. The fractal analysis method, known as detrended fluctuation analysis (DFA), has been suggested for the identification of low-level muscle activations. This study found that DFA performs better in the classification of EMG signals from bifunctional movements of low-level and equal power as compared to other successful and commonly used features based on magnitude and other fractal techniques.

QRS detection algorithm based on the quadratic filter

QRS detection algorithm with noise removal method based on the quadratic filter.Results show the significant improvement in QRS signal to noise ratio.Use only a single fixed threshold without additional post processing techniques.Evaluate the algorithm with 109,483 beats of QRS complexes from MIT-BIH database.Achieve overall sensitivity 99.82% and positive predictive rate 99.81%. QRS detection in the electrocardiogram signal is very crucial as a preliminary step for obtaining QRS complex, beat segmentation, and beat-to-beat intervals. Two main problems in QRS detection are a variety of noise types and various types of abnormal morphologies. We propose a QRS detection algorithm consisting of the quadratic filter for enhancing QRS signal to noise ratio. Results show that significant improvement in QRS signal to noise ratio can be obtained from challenging situations including low amplitude QRS complexes corrupted by baseline drift and abnormal morphologies such as an aberrated atrial premature beat, a premature ventricular contraction beat, a fusion of ventricular and normal beat, and a fusion of paced and normal beat. The enhancements in QRS signal to noise ratio allow us to use a single fixed threshold without any additional post-processing techniques in beat detection step. The performance of proposed algorithm was evaluated with the electrocardiogram data from MIT-BIH arrhythmia database. Results show that the quadratic filter is capable of enhancing QRS signal to noise very well leading to the average detection error rate of 0.38% from 48 records.

Application of linear discriminant analysis in dimensionality reduction for hand motion classification

Application of Linear Discriminant Analysis in Dimensionality Reduction for Hand Motion Classification The classification of upper-limb movements based on surface electromyography (EMG) signals is an important issue in the control of assistive devices and rehabilitation systems. Increasing the number of EMG channels and features in order to increase the number of control commands can yield a high dimensional feature vector. To cope with the accuracy and computation problems associated with high dimensionality, it is commonplace to apply a processing step that transforms the data to a space of significantly lower dimensions with only a limited loss of useful information. Linear discriminant analysis (LDA) has been successfully applied as an EMG feature projection method. Recently, a number of extended LDA-based algorithms have been proposed, which are more competitive in terms of both classification accuracy and computational costs/times with classical LDA. This paper presents the findings of a comparative study of classical LDA and five extended LDA methods. From a quantitative comparison based on seven multi-feature sets, three extended LDA-based algorithms, consisting of uncorrelated LDA, orthogonal LDA and orthogonal fuzzy neighborhood discriminant analysis, produce better class separability when compared with a baseline system (without feature projection), principle component analysis (PCA), and classical LDA. Based on a 7-dimension time domain and time-scale feature vectors, these methods achieved respectively 95.2% and 93.2% classification accuracy by using a linear discriminant classifier.