Surface EMG Signal Classification Using a Selective Mix of Higher Order Statistics

We describe a novel application of Higher Order Statistics (HOS) for classifying Surface Electromyogram (sEMG) signals. We have followed seven approaches to identify discriminating signals representative of four primitive motions, i.e., elbow flexion/extension and forearm supination/pronation. The Sequential Forward Selection (SFS) method is utilized to reduce the number of HOS features to a sufficient minimum while retaining their discriminatory information. The SFS selected the kurtosis of sEMG as well as its second order statistics as discriminating features. Our method is robust, and does not require additional computations as compared to existing efficient methods for providing higher rates of correct classification of sEMG, which make it useful in practical sEMG' controlled prostheses.