Attention-Driven Hybrid LSTM-GRU Model for Enhanced EMG Signal Hand Gesture Recognition
Ranjeesh R Chandran, D. Devaraj, Sreedeep Krishnan · International Journal of Electrical and Electronics Engineering · 2024
Hand Gesture Recognition (HGR) is crucial for Human Computer Interaction (HCI) with applications including assistive technologies for disabled persons to advanced human computer interfaces. HGR from Electromyography (EMG) signals possessed issues regarding noise and variability due to the complex muscle movements. So, for effectively recognizing hand gestures from EMG signals, this paper proposes a hybrid Deep Learning (DL) model combining Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) networks, incorporating an attention mechanism to highlight pertinent features and enhance the model’s focus on critical data segments. The study utilized a dataset with static hand movements captured using an MYO Thalmic bracelet with eight equally distributed sensors, and the raw EMG data was preprocessed. Feature extraction is analyzed in both the time and frequency domains, leading to a more robust and comprehensive analysis of the EMG signal. The proposed model achieved 98.875% accuracy, 97.82% precision, 98.07% recall, and 97.74% F1 score, outperforming existing models. Thus, the proposed model demonstrates significant advancements in hand gesture recognition with high accuracy, making it reliable for several real-time applications.