sEMG 신호를 이용한 Random Forest 기반 보행 상태 분류 모델 연구

Abstract

: This study evaluated the effectiveness of the Random Forest (RF) algorithm in classifying walking status using electromyography (sEMG) data. Data were collected from physically healthy men and women in their 20s and were preprocessed by extracting the root mean square (RMS) features in the 6-channel EMG data. The RF model was used to classify the three activities of walking, climbing stairs, and descending stairs, and was trained based on the optimal hyperparameters. As a result of the verification, the average accuracy was 96.48%, and the accuracy in the test data was 96.64%, indicating excellent performance. In particular, the F1 score for each class was analyzed as walking (0.9930), climbing stairs (0.9560), and descending stairs (0.9496). This study proved that the RF algorithm is effective in classifying walking status, and in future studies, we plan to further improve the performance by apply- ing deep learning algorithms such as LSTM and data augmentation techniques. These developments are expected to increase the possibility of application in practical applications such as walking monitoring and rehabilitation training systems for the elderly.