Hand Gesture Recognition on Edge Devices: Sensor Technologies, Algorithms, and Processing Hardware

Area of Science:

• Computer Science
• Human-Computer Interaction
• Artificial Intelligence

Background:

• Hand gesture recognition (HGR) offers natural human-computer interaction, with existing research focusing on wearable devices.
• Device-free HGR, which does not require users to wear or hold any equipment, presents an alternative approach.

Purpose of the Study:

• To provide a comprehensive overview of device-free HGR systems.
• To analyze the technology, hardware, and algorithms involved in device-free HGR.
• To identify challenges and suggest future research directions for improving HGR.

Main Methods:

• Evaluation of sensor modalities including WIFI, vision, radar, mobile networks, and ultrasound.
• Exploration of pre-processing technologies such as stereo vision, MIMO, and various mapping techniques.
• Study of classification approaches, both with and without machine learning (ML), including tree structures and transformers.

Main Results:

• Demonstration of the interplay between timing, power, hardware, and algorithms in determining HGR granularity, accuracy, and gesture count.
• Assessment of system integration levels, including edge compatibility, real-time capability, continuous learning, robustness, ML application, and accuracy.
• A thorough understanding of the current state-of-the-art in device-free HGR, particularly for edge devices.

Conclusions:

• Device-free HGR systems offer a promising avenue for natural human-computer interaction.
• Further research is needed to address current challenges and enhance the efficiency and accuracy of these systems.
• The study provides a foundation for developing advanced, integrated, and robust device-free HGR solutions.