Improving video surveillance systems using acoustical source localization

The complete surveillance of areas and area perimeters is of upcoming interest for high-level security standard estates. A major drawback of video systems is that complete surveillance of wide areas is very cost expensive considering the huge amount of video cameras needed to monitor areas without gaps. Using thermal cameras areas can be monitored even at night but making the system even more cost expensive. An acoustical source localization system based on generalized cross correlation using phase transform [1] weighting algorithm is proposed which is intended to fill the gaps between the inspection windows of video cameras and offering additional information about the ongoing in the area which could be further evaluated in a multi-modal system design. The system is based on sophisticated far-field broad-band microphone arrays which interact together to retrieve the position of active sound sources. Detection is performed using adaptive threshold technique based on statistical filters. The system is robust to wind noise and uncorrelated noise and has a 360° working radius of approximately 40-50 meters. Sound sources which could be localized well were e.g. person speaking, driving car, breaking, drilling and sawing materials and cutting wires.