Abstract:Gesture recognition, being a natural and convenient means of human-computer interaction, has emerged as a prominent area of research. The goal of accurate and swift gesture recognition has driven numerous studies, as it plays a crucial role in meeting practical application demands. Real-time monitoring of surface electromyography (sEMG) signals collected from sEMG sensor during gesture execution enables the extraction of rich gesture action features, prompting researchers to delve into sEMG-based gesture recognition algorithms. However, existing algorithms often struggle to strike a favorable balance between prediction accuracy and computational efficiency. In response to these challenges, this paper proposes a new gesture recognition model leveraging multi-channel sEMG signals captured by a sEMG sensor. The approach involves employing a lightweight convolutional neural network (CNN) model as the backbone network, supplemented by a lightweight feature extraction block (LFE) to effectively aggregate multi-scale features. Additionally, the model benefits from pre-training weights based on transfer learning, enhancing feature extraction capabilities and generalization ability for faster convergence of the prediction model. Moreover, the introduction of a plug-and-play block called the broad attention learning block (BAL) significantly improves the overall recognition performance of the proposed model. To better capture contextual information from multi-channel sEMG signals, a pyramid input with varying resolutions is constructed, elevating the model's multi-scale feature extraction capability. The proposed model is evaluated and validated on a dynamic gesture dataset collected using wearable sensor (Myo). A series of ablation experiments, comparative analyses, and visualization outcomes unequivocally showcase the remarkable efficacy and superiority of our proposed method. Comparisons with other advanced models reveal that the proposed model successfully achieves a balanced compromise between prediction accuracy and processing speed.

Dynamic Hand Gesture Recognition for Wearable Devices with Low Complexity Recurrent Neural Networks

Dynamic hand gesture recognition using hidden Markov models

Research on Dynamic Gesture Recognition with Low-Pixel ToF-Sensors

Deep Learning Based Hand Gesture Recognition in Complex Scenes

Dynamic gesture recognition based on 2D convolutional neural network and feature fusion

Hand Gesture Recognition on a Resource-Limited Interactive Wristband

Next-Gen Dynamic Hand Gesture Recognition: MediaPipe, Inception-v3 and LSTM-Based Enhanced Deep Learning Model

A wearable sensor-based dynamic gesture recognition model via broad attention learning

High-speed Low-consumption sEMG-based Transient-state micro-Gesture Recognition

EfficientNetV2-based dynamic gesture recognition using transformed scalogram from triaxial acceleration signal

Online Human Gesture Recognition using Recurrent Neural Networks and Wearable Sensors

Dynamic Hand Gesture Recognition Based on a Leap Motion Controller and Two-Layer Bidirectional Recurrent Neural Network

CNN+RNN Depth and Skeleton based Dynamic Hand Gesture Recognition

Dynamic Gesture Recognition Using Inertial Sensors-based Data Gloves.

Motion Feature Augmented Recurrent Neural Network for Skeleton-based Dynamic Hand Gesture Recognition

Dynamic Gesture Recognition Based on RF Sensor and AE-LSTM Neural Network

A Rapid Adaptation Approach for Dynamic Air-Writing Recognition Using Wearable Wristbands with Self-Supervised Contrastive Learning

Dynamic gesture tracking and recognition algorithm based on deep learning

A Novel Hybrid Deep Learning Architecture for Dynamic Hand Gesture Recognition

Deep Learning Based Real-Time Recognition of Dynamic Finger Gestures Using a Data Glove

Machine-learned wearable sensors for real-time hand-motion recognition: toward practical applications