EMG Signal Processing with Clustering Algorithms for motor gesture Tasks

Recent research shows the possibility of using electromyography (EMG) electrical signals to control devices or prosthesis. The EMG signals are measured in muscles, such as the forearm. These signals can lead to determine the intentionality of the patient when performing any motor tasks, however the signals are susceptible to noise due to the voltage sensed, which is in the microvolts scale. In this work, the preprocessing of the EMG signals includes the design and test of a filter. Our designed filter allows eliminating any signal components from the electrical network or any other sources that are not EMG signals. To validate the preprocessing efficiency, we analyze the frequency components and the distribution of the filtered EMG signals. Later, the filtered data was processed with K-means, DBSCAN and Hierarchical Clustering algorithms to determine a subject's intention when performing a task. The results show that the K-means clustering algorithm was able to group the nine gestures made by the subjects, as compared to the DBSCAN and Hierarchical algorithms, which were not able to perform the clustering as expected. However, they match the performance of clustering two groups of combining gestures.