Robust Discrimination of Human vs. Animal Footsteps Using Seismic Signals

We propose to develop innovative statistical signal processing algorithms that discriminate between human and animal footsteps when processing seismic signatures acquired by Unattended Ground Sensors (UGS) and UGS constellations. The innovation is to extract and statistically analyze features (e.g., energy, pattern, and spectral properties) of the seismic signatures. Furthermore, to increase the probability of detection (Pd) and reduce false alarms (Pfa) the algorithms track persistent features and their locations when the seismic source is within the UGS detection range. In Phase I, PSI will develop and test the algorithms using available recorded signals. Performance of the algorithms in terms of Pd vs. Pfa will be investigated by adding noise and anomalous vibrations to the signal. We expect Pd exceeding 0.9 with Pfa below 0.01 for high signal-to-noise ratios (SNR), and Pd exceeding 0.75 with Pfa below 0.05 for low SNR. The capability to discriminate human vs. animal seismic sources in rural areas and borders is expected to benefit the military (e.g., surveillance, reconnaissance), DHS (e.g., border control), and civil communities (e.g., detection of construction near pipes or cables).