Acoustic⁻Seismic Mixed Feature Extraction Based on Wavelet Transform for Vehicle Classification in Wireless Sensor...

Area of Science:

• Signal Processing
• Machine Learning
• Wireless Sensor Networks

Background:

• Accurate vehicle target classification in wireless sensor networks is crucial for surveillance and security.
• Existing methods often struggle with noisy or complex signal environments.
• Integrating acoustic and seismic data offers a promising approach to enhance classification robustness.

Purpose of the Study:

• To propose and validate a novel acoustic-seismic mixed feature extraction method for vehicle target classification.
• To improve target classification accuracy by combining acoustic and seismic signal features.
• To assess the efficiency of feature simplification techniques.

Main Methods:

• Signal decomposition using the à trous algorithm to obtain wavelet coefficients.
• Wavelet Coefficient Energy Ratio (WCER) calculation for feature extraction.
• Hierarchical clustering for feature simplification.
• Support Vector Machine (SVM) for target classification.
• Validation using real-world experimental data.

Main Results:

• The WCER method effectively extracts target features from acoustic and seismic signals.
• Feature simplification reduced processing time without impacting classification accuracy.
• The proposed acoustic-seismic mixed features improved target classification accuracy by approximately 12% compared to using single-modality features.

Conclusions:

• The WCER-based acoustic-seismic mixed feature extraction method is effective for vehicle target classification.
• Feature simplification enhances the practical applicability of the method.
• Combining acoustic and seismic data significantly boosts classification performance in wireless sensor networks.