Mining Significant Features Set For Predicting Isotonic Muscle Fatigue Using Surface Electromyography (sEMG) Signal

Ku, Man Yi (2015) Mining Significant Features Set For Predicting Isotonic Muscle Fatigue Using Surface Electromyography (sEMG) Signal. Project Report. UTeM, Melaka, Malaysia. (Submitted)

Text (24 Pages) Mining Significant Features Set For Predicting Isotonic Muscle Fatigue Using Surface Electromyography (sEMG) Signal 24 Pages.pdf - Submitted Version Download (389kB)

Abstract

Surface electromyography (sEMG) has been widely used in sport science as the muscular biofeedback in order to design the suitable training to increase their muscle endurance based on the signal feedback. However, the research of muscle fatigue caused by isotonic training receives less attention due to higher signal noise than isometric data. Good representative features are known to reduce signal noise. However, there exists no standard significant features set for isotonic training data. Therefore, this study aims to propose the significant features set for isotonic muscle fatigue. Correlation based feature subset selection (CFS) was used to select the significant features set. The commonly used features in sEMG analysis are Mean Absolute Value (MAV), Modified Mean Absolute Value 1 and 2 (MMAV1, MMAV2), Root Mean Square (RMS), Simple Square Integral (SSI), Variance of EMG (VAR), Waveform Length (WL), Mean Frequency (MF) and Median Frequency (MDF). Support Vector Machine was to measure the performance of features in prediction accuracy. Shapiro-Wilk normality test and ANOVA were performed to further verify the experimental results. Frequency domain features showed higher accuracy than time domain features, however features set with both domains showed better accuracy than a single domain. The MAV, RMS, MF and MDF features set is similar to full nine features set in performance. At the same time, CFS has selected two features which is MF and MDF. However, this deducts set is significantly worse than MAV, RMS, MF and MDF features set. Thus, features MAV, RMS, MDF and MF are proposed as the significant features set for isotonic muscle fatigue. For future improvement, research can be tested on different body muscle and more aggressive isotonic training.

Actions (login required)

View Item