Transformer-based hand gesture recognition from instantaneous to fused neural decomposition of high-density EMG signals

Designing efficient and labor-saving prosthetic hands requires powerful hand gesture recognition algorithms that can achieve high accuracy with limited complexity and latency. In this context, the paper proposes a Compact Transformer-based Hand Gesture Recognition framework referred to as CT-HGR , which employs a vision transformer network to conduct hand gesture recognition using high-density surface EMG (HD-sEMG) signals. Taking advantage of the attention mechanism, which is incorporated into the transformer architectures, our proposed CT-HGR framework overcomes major constraints associated with most of the existing deep learning models such as model complexity; requiring feature engineering; inability to consider both temporal and spatial information of HD-sEMG signals, and requiring a large number of training samples. The attention mechanism in the proposed model identifies similarities among different data segments with a greater capacity for parallel computations and addresses the memory limitation problems while dealing with inputs of large sequence lengths. CT-HGR can be trained from scratch without any need for transfer learning and can simultaneously extract both temporal and spatial features of HD-sEMG data. Additionally, the CT-HGR framework can perform instantaneous recognition using sEMG image spatially composed from HD-sEMG signals. A variant of the CT-HGR is also designed to incorporate microscopic neural drive information in the form of Motor Unit Spike Trains (MUSTs) extracted from HD-sEMG signals using Blind Source Separation (BSS). This variant is combined with its baseline version via a hybrid architecture to evaluate potentials of fusing macroscopic and microscopic neural drive information. The utilized HD-sEMG dataset involves 128 electrodes that collect the signals related to 65 isometric hand gestures of 20 subjects. The proposed CT-HGR framework is applied to 31.25, 62.5, 125, 250 ms window sizes of the above-mentioned dataset utilizing 32, 64, 128 electrode channels. Our results are obtained via 5-fold cross-validation by first applying the proposed framework on the dataset of each subject separately and then, averaging the accuracies among all the subjects. The average accuracy over all the participants using 32 electrodes and a window size of 31.25 ms is 86.23