HAGR-D: A Novel Approach for Gesture Recognition with Depth Maps.

Area of Science:

• Computer Vision
• Human-Computer Interaction
• Robotics

Background:

• Hand gestures are crucial for communication and interaction, with applications in medical fields, gaming, and sign language.
• While depth sensors have advanced gesture recognition, complex hand articulations present ongoing challenges.
• Existing computer vision methods struggle with the intricacies of dynamic hand gesture recognition.

Purpose of the Study:

• To propose a novel and effective approach for dynamic hand gesture recognition using depth maps.
• To address the complexity of hand articulations in gesture recognition systems.
• To develop a robust model that surpasses current state-of-the-art algorithms.

Main Methods:

• Utilized depth maps generated by the Microsoft Kinect Sensor.
• Implemented a variation of the convex invariant position based on RANSAC (CIPBR) algorithm.
• Employed a hybrid classifier combining dynamic time warping (DTW) and Hidden Markov models (HMM), termed HAGR-D.

Main Results:

• The proposed HAGR-D model demonstrated superior performance in hand gesture recognition tasks.
• Achieved a high classification rate of 97.49% on the MSRGesture3D dataset.
• Attained an impressive 98.43% classification rate on the RPPDI dynamic gesture dataset.

Conclusions:

• The HAGR-D approach offers a significant advancement in dynamic hand gesture recognition using depth data.
• The hybrid classifier effectively handles the complexity of hand movements for accurate recognition.
• This method provides a promising solution for real-world applications requiring reliable hand gesture interpretation.