Signal analysis to predict driver’s vigilance state using emg signal /
Noor Aisyah Ab Rahman
- xvi, 179 pages : illustrations some color ; 30 cm. + 1 CD ROM
Faculty of Electrical & Electronics Engineering
Thesis (Master of Science) -- Universiti Malaysia Pahang – 2021
Includes bibliographical references
Lately, road accidents that have resulted in the loss of life, serious injuries and property loss have become one of the hot topics on the mass media and in the electronic media almost daily shows that road accidents are serious matter and need to be taken care of by certain parties. Some of the causes of road accidents are due to the poor driver vigilance when most drivers show signs of fatigue and loss of vigilance during long and monotonous driving. Muscle fatigue is a complex phenomenon that covers variety of causes, mechanisms and manifestations that develops due to changes in the efficiency of the nervous system that can be seen in the declining the performance. This study was presented brief explanation of the different methods for processing and classifying of electromyogram (EMG) signals for muscle fatigue. This study then compares numerous methods of analyzing signals, in terms of their performance. In this study, a physiological measure was used to estimate a driver's muscle fatigue while driving was used. The raw data were obtained by pairing the electrode to the bicep brachii of every 16 subjects while playing for Need for Speed (NFS) game for two hours. Before that, the subjects were answered a set of questionnaires and the score were calculated. From the questionnaire, the condition of the driver can be determined whether the driver is in non-fatigue or fatigue condition. The data were filtered using the Butterworth bandpass filter. Next, filtered data were analyzed using time domain, frequency domain and time-frequency domain analysis for feature extraction method. Then, the extracted features were undergone feature selection method. Finally, to evaluate the performance measure of the feature selection method and classify the driver’s condition, Artificial Neural Network (ANN) and Support Vector Machine (SVM) was used. From the results obtained, it can be concluded that ANN performance using features that undergo feature selection method produce a better classification accuracy compared to the ANN performance using features that did not undergo feature selection method with 93.8% against 87.5% accuracy achieved. While for SVM performance, 81.3% and 75% were the accuracy of features that undergo feature selection and features did not undergo feature selection method respectively. In conclusion, the performance of ANN and SVM increased by 6.3% and 18.7% respectively. Apart from that, EMG Signal Segmentation (ESS) methods were applied. EMG filtered signals were partitioned into few segments in segmentation process. The two hours signals were split into 60 minutes long segments for the first methodology. For the second methodology, the signals were divided into 30 minutes of segments. From each segment, the features were extracted and analyzed to contemplate the progressions of the features. Based on signal segmentation result, the change of features shows the clear condition of muscles and every subject yield different result because the different person has different muscle condition. As a conclusion, comprehensive muscle fatigue measurement methodology for driving in a simulated environment was described. Valuable information can be derived from muscles and the information can be used in evaluating the surface EMG signals to detect muscle fatigue in the driving task. Relevant parameters that useful for the development of human safety by assessing muscle fatigue during the driving task were found.
THE0009273(Local)
Faculty of Electrical & Electronics Engineering--Dissertations