Acoustic Pattern Recognition for Security Breaching Noise Detection

A proposed acoustic pattern recognition system will be based around an array of acoustic sensors to allow localization of sound events and cancellation of background noise. The number of sensors could be adjusted to vary the coverage area depending on the site. The measured signals will provide the input to a sophisticated signal processing algorithm that will separate the acoustic signals into constituent sound sources in order to detect, characterize, localize, and identify security breaching sounds, all amongst a high background noise environment. Following source separation, an adaptive classifier will be utilized to discriminate between nuisance and background sounds and security breaching sounds. A novel attribute of this classifier is its feature extraction approach, which leverages the images produced by auditory models and thus tightly couples the classifier to the mechanisms potentially used by the biological systems we propose to mimic.BENEFITS: lotsDevelopment of this sensor capability will provide a major step in developing a sensor recognition system that can detect the sounds of break-on and gunfire. It will support a number of military applications besides the facility security mission, such as monitoring trail use and protecting troops from infiltrators and ambush. Commercially, the increased need for a border, plant, and perimeter security sensor that can reduce the need for extensive manpower would be welcomed.