In the past few years, Hand Gesture Recognition (HGR) utilizing EMG data has gained significant attention for improving human-machine interaction. However, effectively modeling multi-scale temporal variations and discriminative channel-wise features in EMG signals remains challenging. Although recent deep learning-based HGR methods have shown promising performance, many struggle to jointly capture multi-scale temporal dependencies and channel-wise feature relevance inherent in EMG signals. Deep learning (DL) methods offer several advantages over Machine Learning (ML) methods, including promising classification performance. In this research, a DL-based Multi-Scale Deep Residual Network (DRN) with an SE model that utilizes EMG signals to recognize hand gestures is proposed. The proposed architecture integrates multi-scale residual learning with squeeze-and-excitation–based channel recalibration, enabling the network to learn both temporal patterns at different scales and gesture-specific feature importance. Initially, the EMG-EPN-612 dataset is utilized to train and evaluate the research model. The research model includes data collection, preprocessing, feature extraction process and classification tasks. Standard EMG preprocessing, including moving average filtering, min–max normalization, and sliding-window segmentation, is applied, followed by gesture classification using a Multi-Scale DRN integrated with an SE module to capture temporal patterns and channel-wise dependencies. The dataset is divided into two parts, one of which contained 75% of the data for training and the other 25% for testing. The improved performance is attributed to the proposed model’s ability to effectively capture discriminative multi-scale temporal features and adaptively emphasize informative EMG channels. The proposed model not only achieved 99.24% accuracy, 99.15% precision, 99.17% f1-score, 99.10% specificity, and 99.20% recall, but also overcomes compared models.
Ramkumar et al. (Sat,) studied this question.