Diamondback: Sensor Fusion and Feature-Based Human/Animal Classification for UGS

Unattended ground sensors (UGS) are essential in order to determine if personnel are illegally walking across a border for homeland security. However, a problem with current UGS units and their implementation is that they are poor at discriminating between humans and animals. The main reason for this is that the signal processing algorithms for classification are currently limited to using cadence differences between humans and animals. Such an approach can be defeated by the adversary if they mimic such movement. False alarms set off by the UGS units increases the workload of US Customs and Border Patrol (CBP) agents unduly. New signal processing algorithms that generate fewer false alarms are possible if they utilize other feature vectors in the seismic data beyond cadence. Typical UGS units include seismic, acoustic and infra red sensors. A multimodal approach that uses acoustic and infra red sensors along with the seismic sensor is likely to further reduce the false alarm rate. Scientific Systems Company Inc. (SSCI) is teaming with Crane Wireless Monitoring Solutions (Crane) for this project. SSCI has extensive experience with signal processing and pattern recognition techniques and Crane has significant experience in developing UGS systems and intrusion detection algorithms.