Shift invariant feature extraction for sEMG-based speech recognition with electrode grid

For Japanese vowel recognition based on surface electromyography (sEMG), an electrode grid has been shown to be effective in our previous studies. In this study, we aim to leverage potential of the electrode grid further by using with a spatial shift invariant feature extraction method that can compensate deviation of the attached site of the electrode grid. We verified efficiency of the shift invariant feature extraction method in improving the recognition accuracy. 2-D dual tree complex wavelet transform was employed as such a shift invariant feature extraction method. Our result shows that shift invariant feature can provide additional information that cannot be provided when the channel signals are utilized independently.