An Acoustic Sensing Gesture Recognition System Design Based on a Hidden Markov Model

Area of Science:

• Human-computer interaction
• Machine learning
• Signal processing

Background:

• Human activities heavily rely on tactile interactions with surrounding objects and surfaces.
• Recognizing touch-based gestures is crucial for developing interactive surface technologies.
• Acoustic-based input classification is an active research area due to its low computational cost.

Purpose of the Study:

• To compare the effectiveness of Artificial Neural Network (ANN) and Hidden Markov Models (HMM) for acoustic-based gesture recognition.
• To evaluate a low-cost hardware system for passive gesture recognition.
• To assess the performance of ANN and HMM using a simple alphabet of geometric shapes.

Main Methods:

• Developed a low-cost hardware system with a microphone, stethoscope, conditioning circuit, and microcontroller.
• Integrated the hardware with a surface to create a passive gesture recognition input system.
• Acquired acoustic signals, processed them via a microcontroller, and evaluated ANN and HMM models using MATLAB toolboxes.

Main Results:

• The Hidden Markov Model (HMM) technique demonstrated robustness in gesture recognition.
• HMM achieved a 90% success rate in classifying gestures (circle, square, triangle).
• HMM exhibited a shorter training time compared to the Artificial Neural Network (ANN).

Conclusions:

• Hidden Markov Models are a robust and efficient technique for acoustic-based gesture recognition.
• The developed low-cost system is effective for evaluating gesture recognition algorithms.
• Further research can explore the advantages and limitations of ANN and HMM in more complex gesture recognition tasks.