Features extraction of surface electromyopgraphy in term of force

Muhamad Syazwan , Mohd Jasni (2014) Features extraction of surface electromyopgraphy in term of force. Project Report. UTeM. (Submitted)

Text FEATURES EXTRACTION OF SURFACE ELECTROMYOGRPAHY IN TERM OF FORCE 24pages.pdf Download (343kB)

Abstract

The research of feature extraction based on surface electromyography in term of force is done to help the people with upper limb amputation disabilities. Therefore, this research will focus on the relationship between the normalized electromyography (EMG) signal and force. This objective of this research is to do feature extraction of electromyography (EMG) in term of force in time domain. Then, the statistical analysis by using a simple linear regression technique on scatter plot was done to analyze the relationship between force and electromyography (EMG). 10 subjects are selected that mainly based on criteria of weight and the health condition of the person. The selected muscle is long head biceps brachii. The experiment is divided into three main tasks which consist of angles of 450,900 and 1200. In addition, three tasks which consist of loads of 2kg, 4kg, and 6kg are done. The feature extractions with mean absolute value, root mean square (RMS), variance and standard deviation are analyzed by using simple linear regression analysis. The calculation of force formula from electromyography (EMG) signal is used to predict force. The average value is used to develop the equation of force because it has high value of correlation coefficient as compared to the value for all the subjects. Two methods to determine the reliability of the equation are based on the percentage of calculating force error and percentage of average predicted force error. The result has stated that mean shows the best feature extraction based on simple linear regression analysis characteristics. The result has shown that the performance for calculating force and average predicted force are inaccurate because the value of percentage of error is high. Hill-Based model and neural network are ways to improve the inaccuracy of simple linear regression technique to predict force.

Actions (login required)

View Item