<![CDATA[Paralysis is a condition in which the movement of a body part is impaired to the point where it is unable to move partially or completely. People with paralysis often need assistive devices to help with their mobility, such as wheelchairs. Smart wheelchairs that use a voice control system can help people who are unable to use their hands to control a wheelchair. This system uses the MFCC feature extraction method, which is the method that most closely approximates human hearing, and the CNN classification method, which has been proven to work well when trained with features extracted using MFCC. The system is run on a Jetson TX2 device and controls the wheelchair using an Arduino Uno by adjusting the pulse width modulation value according to the classification result of the system's command. The dataset used to train the CNN model is the speech_command v2 dataset created by Tensorflow, which contains over 500,000 data for 36 classes. In this research, however, 15,000 data were used for 4 command classes: "Go," "Right," "Left," and "Stop." The results of the system testing show an accuracy prediction value of 85% with a relatively fast average computation time of only 0.37 seconds to make a prediction.]]>
Copyrights © 2022
