Support Vector Machine and Random Forest Evaluation for Motion Intention Detection Using Surface EMG

Although the world's population is growing, the birth rate is currently declining and is expected to decline, resulting in an ageing population. Electric wheelchairs enhance social activities by expanding the mobility range of older people. However, the proportion of accidents involving electric mobility devices is high among people aged 70 years and above. This is because of decreased sensory and cognitive speeds and motor function. To improve these factors, we focused on surface electromyograms (sEMG) that are output before the start of movement. Using this signal, we used machine learning to estimate the distance and direction of movement of a hand before concluding the control operation of the wheelchair. Multiple features, including frequency and magnitude sum, were determined from the measured sEMG, and two types of machine learning, support vector machine (SVM) and random forest (RF), were used to identify the distance and direction of movement. Three types of movement distances and four types of movement directions were identified. The feature importance was calculated, and those contributing to the distance and direction discrimination were determined and compared for each machine learning method. The discrimination rate using an RF was higher than that using SVM. The results are discussed, focusing on the importance of features for each participant.