A novel statistical decimal pattern-based surface electromyogram signal classification method using tunable q-factor wavelet transform

Surface electromyogram sensors have been widely used to acquire hand gestures signals. Many machine learning and artificial intelligence methods have been presented for automated surface electromyogram signals classification. In this method, a novel surface electromyogram signals recognition method is presented using a novel 1D local descriptor. The proposed descriptor is called as statistical decimal pattern and it is utilized as feature extractor in this study and tunable q-factor wavelet transform is used as pooling in this method. By using tunable q-factor wavelet transform and the proposed statistical decimal pattern, a multileveled learning method is constructed. Ten levels are created by using tunable q-factor wavelet transform. Statistical decimal pattern extracts features from tunable q-factor wavelet transform sub-bands of the raw surface electromyogram signal. Then, the generated features are concatenated, and to select distinctive features, ReliefF and neighborhood component analysis are used together. In the classification phase, k-nearest neighbor classifier with city block distance is chosen. To test performance of the proposed tunable q-factor wavelet transform and the proposed statistical decimal pattern-based surface electromyogram classification method, a freely and publicly published dataset was used. In this dataset, 10 hand gestures were defined. Experimental results clearly shown that the proposed tunable q wavelet transform and statistical decimal pattern-based method achieved 98.0%, 99.79% accuracy rates on two datasets and it outcomes other state-of-the-art methods according to these results.