Machine Learning-Based Gesture Recognition Glove: Design and Implementation.

Area of Science:

• Human-Computer Interaction (HCI)
• Robotics and Wearable Technology
• Machine Learning for Gesture Recognition

Background:

• Gesture recognition is crucial in HCI, with smart gloves being key devices.
• Existing research on data gloves primarily focuses on static gestures, neglecting dynamic gesture recognition.
• Dynamic gesture recognition is vital for interactive applications like gaming and virtual environments.

Purpose of the Study:

• To develop a low-cost smart glove prototype for capturing and classifying dynamic hand gestures.
• To implement a neural network-based classifier for accurate dynamic gesture recognition.
• To evaluate the prototype's effectiveness for game control applications.

Main Methods:

• A prototype data glove was designed with five flex sensors, five force sensors, and one Inertial Measurement Unit (IMU).
• A Convolutional Neural Network (CNN) classifier with three 2D convolutional layers and ReLU activation was developed for dynamic gesture classification.
• Performance was evaluated using metrics such as accuracy, precision, and recall, supported by confusion matrix analysis.

Main Results:

• The developed smart glove prototype successfully captured and classified dynamic hand gestures.
• The CNN-based classifier achieved a high accuracy of 90% for gesture recognition.
• The system demonstrated high classification accuracy, precision, and recall, indicating effective performance.

Conclusions:

• The low-cost smart glove provides a cost-effective and accurate solution for dynamic gesture recognition.
• The prototype shows significant potential for applications in gaming and Virtual/Augmented Reality (VR/AR) environments.
• Further research could expand the gesture vocabulary and participant pool for broader applicability.