sEMG Signal Features Extraction and Machine Learning- Based Gesture Recognition for Prosthetic Hand

Abstract

Amputees around the world barely have any access to top-notch, smarter prosthetics due to the fact that they are either inaccurate or cost-inefficient. One of the more challenging tasks is the accurate detection of gestures and this paper illustrates a comparative analysis of the different machine learning-based algorithms for the gesture-identification. The first step in the process is the data extraction from the sEMG device, followed by the features extraction. Then, the six machine learning algorithms are applied to the testing and training data to compare the prediction accuracy and the region of convergence curve. The medium Gaussian SVM performs well under all conditions as compared to the K nearest neighbor, Fine Tree, Naïve Bayes, Linear Discriminator, and Subspace Discriminant. Different parameters are used for the comparison which include F1 score, Confusion Matrix, Accuracy, Precision, and Kappa. The conventional order strategies for hand gesture recognition based on sEMG have been thoroughly investigated and have yielded promising results. In any case, data miscalculation during feature extraction reduces recognition precision. The profound learning technique was presented to achieve greater precision. Therefore, the proposed design takes into account all aspects of the sEMG signal. The system secures a highest classification accuracy of 92.2% for the case of SVM algorithm.