Gait Classification Based on Micro-Doppler Effect

Area of Science:

• Radar Signal Processing
• Biomedical Engineering
• Human Motion Analysis

Background:

• Gait feature extraction is crucial for human motion analysis and identification.
• Traditional state-space methods (SSM) for gait analysis require model order estimation.
• Micro-Doppler signatures from radar echoes contain rich information about human movement.

Purpose of the Study:

• To propose an improved state-space method (SSM) for gait feature extraction.
• To enhance the accuracy of micro-Doppler trajectory identification and micro-motion feature extraction.
• To validate the effectiveness of the proposed method for pedestrian classification.

Main Methods:

• Introduced zero-phase component analysis Whitening (ZCA Whitening) and estimated echo search algorithm for preprocessing pedestrian echoes.
• Utilized a two-channel echo input for the improved SSM to identify micro-Doppler trajectories.
• Extracted five key gait features: torso amplitude, stride length, walking cycle, torso maximum speed, and feet maximum speed.
• Validated the method using Boulic model simulations and real-world data from a 77 GHz FMCW radar.

Main Results:

• The improved SSM eliminates the need for model order estimation.
• Achieved more accurate torso micro-Doppler trajectories and effective micro-motion features of the feet compared to traditional SSM.
• Demonstrated successful pedestrian classification using a support vector machine (SVM) based on the extracted gait features.

Conclusions:

• The proposed improved state-space method offers a more accurate and efficient approach to gait feature extraction.
• The extracted gait features are effective for pedestrian identification using radar data.
• This method holds potential for applications in surveillance, healthcare, and human-computer interaction.