ABSTRACT
Remote object signature detection is becoming increasingly important in noncooperative and hostile environments for many applications (Dedeoglu et al., 2008; Li et al., 2008). These include (1) remote and large area surveillance in frontier defense, 66maritime affairs, law enforcement, and so on; (2) perimeter protection for important locations and facilities such as forests, oil fields, railways, and high-voltage towers; and (3) search and rescue in natural and man-made disasters such as earthquakes, flooding, hurricanes, and terrorism attacks. In these situations, target signature detection, particularly signatures of humans, vehicles, and other targets or events, at a large distance is critical in order to watch out for the trespassers or events before taking appropriate actions, or make quick decisions to rescue the victims, with minimum risks. Although imaging and video technologies (including visible and infrared (IR)) have had great advancement in object signature detection at a large distance, there are still many limitations in noncooperative and hostile environments because of intentional camouflage and natural occlusions. Audio information, another important data source for target detection, still cannot match the range and signal qualities provided by video technologies for long-range sensing, particularly under a variety of large background noises. For obtaining better performance of human tracking in a near to mediate range, Beal et al. (2003) and Zou and Bhanu (2005) have reported the integrations of visual and acoustic sensors. By integration, each modality may compensate for the weaknesses of the other one. But in these systems, the acoustic sensors (microphones) need to be placed near the subjects in monitoring, and therefore cannot be used for long-range surveillance. A parabolic microphone, which can capture voice signals at a fairly large distance in the direction pointed by the microphone, could be used for remote hearing and surveillance. But it is very sensitive to noise caused by the surroundings (i.e., wind) or the sensor motion, and all the signals on the way are captured. Therefore, there is a great necessity to find a new type of acoustic sensor for long-range voice detection.
