Adaptive and learning algorithms for seismic detection of personnel
1Department of Electrical Engineering, Southeastern Massachusetts University, North Dartmouth, MA 02747.
IEEE Transactions on Pattern Analysis and Machine Intelligence
|August 27, 2011
Summary
Adaptive Kalman filtering effectively detects intruders using seismic data by learning nonstationary noise patterns. This method outperforms adaptive filtering for long-term seismic record processing.
Area of Science:
- Geophysics
- Signal Processing
- Seismic Data Analysis
Background:
- Intruder detection using seismic sensors relies on extracting faint impulse signals from noisy environments.
- Correlated background noise in seismic data presents a significant challenge for accurate signal feature extraction.
Purpose of the Study:
- To develop and compare adaptive digital processing techniques for intruder detection using seismic sensor data.
- To evaluate the performance of adaptive digital filtering and adaptive Kalman filtering in extracting impulse-like signal features.
Main Methods:
- Development of adaptive digital filtering algorithms.
- Development of adaptive Kalman filtering algorithms tailored for seismic data.
- Comparative analysis of filtering methods on short and long seismic data segments.
Main Results:
- Both adaptive digital filtering and adaptive Kalman filtering showed similar performance on short data segments.
- Adaptive Kalman filtering demonstrated superior capability in learning nonstationary data characteristics for long seismic records.
- Adaptive Kalman filtering proved more effective in adaptively removing background noise over extended periods.
Conclusions:
- Adaptive Kalman filtering is a more robust technique for long-term seismic intruder detection due to its adaptability to changing noise conditions.
- The study highlights the importance of adaptive algorithms in processing nonstationary seismic data for enhanced detection accuracy.
