Characterisation of Muscular Activities for Lower Extremities

Foo, Wei Jun (2020) Characterisation of Muscular Activities for Lower Extremities. Final Year Project (Bachelor), Tunku Abdul Rahman University College.

Text FooWeiJun_Full Text.pdf Restricted to Registered users only Download (4MB)

Abstract

This paper aims to research on the characterisation of muscular activity for lower extremities. For years, people are trying to determine the relationship between human intention and associated muscular voluntary contraction to allow movement. Electromyography (EMG) had been discovered and have gradually become significant with the development of technologies. The potential has been discovered, which could aid in different field, from medical, sport to military. However, several issues accompanied have as well been determined such as noises and interference which may causes significant change to the performance of study. In this project, various methods suggested by different researchers are studied and compared. Control methods as well been re-implemented and designed to enhance the conduction of data acquisition, processing and classification. Such methods, including implementation of detection pant at acquisition to ease out the repetitive installation of biomedical pads. Besides, Butterworth low pass filtering is performed on the raw data at 3rd order and 68Hz cut-off frequency and Savitzky-Golay smoothing filtering at default order from MATLAB toolbox at processing stage. In addition, EMG feature extraction is performed by utilising amplitude visualisation and gradient plot at characterisation stage. Moreover, to understand the human behaviour in carrying out daily routine gaits, several experiments are designed and carried out including sit-to-stand/stand-to-sit, walk-to-stop/stop-to-walk, full walking gait movement. Results and analysis are studied after the computing of data, by filtering, characterising and verifying for the consistency and accuracy of result. After signal acquisition being performed, for both Butterworth low pass filter and Savitzky-Golay Smoothing Filter, raw and filtered signal are being compared. Butterworth filter shows clearer end results for later characterising works while Savitzky-Golay Smoothing Filter is used for characterisation guidance due to its information preserving nature. The filtered signals are then characterised in simultaneous manners for 4 muscles groups and corresponding gradient trends are extracted. By calculating the mean of the positive and negative gradient value, the threshold magnitude for input pulse is generated for rising and falling phase change respectively. In general, 20% of the peak gradient plot is being set as the indicator of muscle activation. The pulses are re-input to other set of filtered signals to verify the accuracy and consistency. In overall, the verification is successful as most of the spikes generated are in synchronisation with the input pulse. In conclusion, the objectives of the project are fulfilled, where the number of EMG sensor pad to allow lower extremities signal acquisition; raw signals are being processed to obtain a clean and clear output for characterisation work; the pattern of EMG by varying the electrode pads position are determined. Lastly, the muscle movements and its corresponding EMG pattern are successfully characterised and analysed.