Data glove-based gesture recognition using CNN-BiLSTM model with attention mechanism.

Area of Science:

• Human-Computer Interaction
• Machine Learning
• Biomedical Engineering

Background:

• Hand gesture recognition (HGR) is a key human-machine interaction (HMI) technology.
• Visual-based HGR systems face challenges like occlusion and depth perception.
• Data gloves offer a robust alternative for HGR, particularly in complex environments like virtual reality and medical simulation.

Purpose of the Study:

• To propose and evaluate a novel data glove-based dynamic gesture recognition model.
• To enhance the accuracy and robustness of dynamic HGR using data gloves.
• To explore the potential of data gloves in advanced HMI applications.

Main Methods:

• Developed the Attention-based CNN-BiLSTM Network (A-CBLN) model.
• Utilized Convolutional Neural Networks (CNN) for local feature extraction.
• Employed Bidirectional Long Short-Term Memory (BiLSTM) networks for temporal feature analysis.
• Integrated attention mechanisms to weigh gesture features for improved understanding.

Main Results:

• The A-CBLN model demonstrated superior performance in dynamic gesture recognition.
• Achieved a high accuracy of 95.05% and precision of 95.43% on the test dataset.
• Effectively addressed challenges inherent in dynamic gesture recognition tasks.

Conclusions:

• The proposed A-CBLN model offers an effective solution for data glove-based dynamic HGR.
• Attention mechanisms significantly enhance the model's ability to interpret gesture nuances.
• This approach shows promise for applications in virtual reality, medical simulation, and other interactive fields.