ABSTRACT
The variety of the EMG application areas and developments of telemedicine emphasize the importance of the storage and transmission of the EMG data. Novel methods focusing on compression have become a top priority in this regard. This chapter is focused on the cost and performance issues during compression of EMG signals that are utilized, specifically, for classification applications. The classification of facial EMG responses to fearful sound stimuli was chosen in this study and then, it can be categorized as a binary classification task. The compression of the facial EMG data was conducted via PCA and DCT. Compression of EMG data with PCA yielded favourable results with appropriate parameter choices in such a way that the classification accuracy varied between 86-91%. On the other hand, compression with DCT proves to be beneficial in real-time applications due to the fact that it allows for signal compression on the fly in real-time, with an efficient accuracy rate of 82-86% when all features are used. In addition, the accuracy rates of classification of EMG signals after compression with DCT are no less than those of original EMG signals. Therefore, the benefits of compression outweigh the small operational costs and related to computation time.
