A flexible approach for segregating physiological signals

Highlights

• SEMG signals were recorded using designed sensor system.

• Labview simulated code for extracting features was explored.

• A mechanism for interpretation of electromyogram signal was discussed.

• Factorial analysis technique reports effectiveness of recorded signal.

Abstract

The interpretation of Surface Electromyogram (SEMG) signals at multiple muscle points for different operations of arm was investigated. Myoelectric signals were detected using designed acquisition setup consists of an instrumentation amplifier, filter circuit, an amplifier with gain adjustment. Further, Labview softscope code was used to record the SEMG signal for independent movements. The whole system consists of bipolar noninvasive electrodes, signal acquisition protocols and signal conditioning at different levels. Feature extraction was done for exercising Statistic Measured Index method to evaluate the distance between two independent groups by directly addressing the quality of signal. Thereafter factorial analysis of variance technique was investigated to analyze the effectiveness of recorded signal. Finally the SEMG feature evaluation index based reported work is a step forward to develop more powerful, flexible, efficient applications leading to prosthetic design.

Introduction

Surface Electromyography (SEMG) involves recording the action potentials that activate skeletal muscle fibers. The action potentials are initiated at the muscle fiber membranes and resemble electrical waves that pass along the fibers to stimulate muscle contraction. In most cases, the action potentials are detected with bipolar arrangement placed on the surface of the skin over the muscle. Thus, during muscle contraction, the action potentials travel through the tissue overlying the muscle and are picked up by the electrodes on the skin surface. Because these electrodes are not very selective, the surface EMG signal is generated by a summation of the action potentials from many muscle fibers. The SEMG is the recorded sequences of two principal bioelectric activities i.e. (a) propogation of motor nerve impulses and their transmission at the neuromuscular junctions of a motor unit, and (b) propogation of muscle impulses by the sarcolemma and the T-tubular systems resulting in excitation and contraction coupling [1].

In order to predict the effect of SEMG corresponding to voluntary muscle contraction, various models have been developed by researchers [2], [3], [4], [5], [6]. Although it has been shown that all the aforementioned properties of a model are important for interpreting real data, no model is available in the literature that includes all these characteristics. From the late 1990s, due to the availability of computer systems and software modules, the researchers focused on two main subjects: (a) the relationship between the myoelectric signals collected by surface electrodes and the working of muscles and the nervous system, and (b) new techniques to extract information about the central nervous systems strategy for controlling motor units [7], [8], [9], [10], [11], [12], [13].

The advancement of electromyography, combined with modern digital processing techniques, has strongly renewed researchers’ interest in detecting upper-limb motion intentions for prosthetic control [14], [15] but the difficulties in SEMG signal classification for prosthetic applications are the selection of electrode locations on the arm, signal processing and the extraction of a feature vectors which are able to classify several motions [16], [17], [18], [19], [20], [21]. Among different methods to be used for the investigation of SEMG, the spectral estimation [22] plays a vital role, and the most widely used method for this is the Fourier transform. The application of the FFT requires a stationary signal, but the nature of the signal changes with contraction level and time which indicated that its nature is non stationary and non linear. Even when there is no voluntary change of muscle state, characteristics of signal changes due to the variation in the blood flow. Therefore, in order to analyze the characteristics of non stationary SEMG signal, more advanced and powerful computer aided signal processing algorithms need to be used both in time and frequency domain.

The objective of this present work is to acquire SEMG; its pre-processing; extraction of features (comparison of recorded data against two participated muscle locations); implementation of statistical technique; and finally the cross validation of results using different techniques for ensuring class separability among four exercised movements.