A Novel Interval Type-2 Fuzzy Classifier Based on Explainable Neural Network for Surface Electromyogram Gesture Recognition

The existing hand gesture classification research based on surface electromyogram (sEMG) faces the challenges of low classification accuracy, weak real-time ability, weak robustness, few categories, and lack of explainability. In this article, we investigate how to classify sEMG signals for grasp recognition and human-robot interaction to consider these issues. A novel interval type-2 (IT2) fuzzy classifier based on explainable neural network is proposed for sEMG gesture recognition. Based on fully connected neural network, the adaptive moment estimation is applied to tune the antecedent parameters. The Ninapro data is adopted to test the performance of the proposed model, which realizes recognition of 52 gestures and achieves 95.04% categorization accuracy. Moreover, grasping experiments are conducted on computer, communication, and consumer electronics (3C) experiment platform to test the ability of the classifier in real scenarios. The experiment recognizes six gestures. The results of the 3C grasping experiment show that the proposed method achieves 99.4% offline training accuracy as well as 96.07% online test accuracy. Meanwhile, 89.4% of the classification results can be obtained within 0.5 s. The overall results demonstrate great potential for real-world applications, such as human intent detection and manipulator control.