Hand Gesture Recognition in Automotive Human⁻Machine Interaction Using Depth Cameras

Area of Science:

• Computer Science
• Robotics
• Human-Computer Interaction

Background:

• Hand gesture recognition is crucial for intuitive human-computer interaction.
• Depth data from time-of-flight (ToF) sensors offers rich information for gesture analysis.
• Evaluating machine learning models requires comprehensive and diverse datasets.

Purpose of the Study:

• To review current machine learning approaches for hand gesture recognition using ToF sensor data.
• To present research findings from the Computational Neuroscience laboratory at Ruhr West University of Applied Sciences.
• To introduce the REHAP dataset as a novel benchmark for 3D hand posture analysis.

Main Methods:

• Review of machine learning techniques, focusing on deep learning models like Convolutional Neural Networks (CNNs) and Long Short-Term Memory (LSTM).
• Investigation of sensor data fusion techniques within a deep learning framework.
• Conducting user studies to evaluate the practical performance of the developed gesture recognition system.

Main Results:

• Convolutional Neural Networks and Long Short-Term Memory models demonstrate the most reliable results for hand gesture recognition.
• Sensor data fusion techniques integrated into deep learning frameworks show promise for improved accuracy.
• User studies confirm the practical viability and effectiveness of the evaluated system.

Conclusions:

• Deep learning models, particularly CNNs and LSTMs, are highly effective for hand gesture recognition with ToF depth data.
• The REHAP dataset provides a valuable resource with over a million 3D hand posture samples for advancing research in this field.
• Future research should leverage advanced deep learning architectures and comprehensive datasets for robust gesture recognition systems.