Multi-view learning for classification of EMG template

Goal: The paper addresses a multi-view learning model (MVL) making use of subspace model for classification of Electromyography (EMG) templates. Method: For comprehensive representation of class information multi-view patterns are generated for predefined classes by dint of formulation pursuit, followed by correlation measures to estimate low dimensional embeddings (LDEs). The LDEs with statistical significant p-values are fused and fed to ensemble learning models. Results: The models have been demonstrated with benchmark EMG data sets. Experimental aftermaths and subsequent comparison with stateof-the-arts in terms of accuracy, sensitivity and specificity manifest the efficacy of proposed MVL. Conclusion: The algorithm effectively classifies both pathological and normal templates with learned patterns. Thus, it promises to alleviate the onus of clinician of large volume data and also expedite the large scale diagnosis research. Further, it ensures the suitable adaption to home-care health monitoring portable devices. Significance: The MVL automatically identifies EMG templates with significant reduction of dimensionality and complexity.