Intelligent Acoustic Source Detection/Recognition for Noisy Environments

To address the U.S. Navy needs for an intelligent, acoustic recognition system for extremely noisy environments at military installations, on ships, and in vehicles, Physical Optics Corporation (POC) proposes to develop a new intelligent acoustic threat detection/recognition system for extremely-low SNR environments; this system will be based on a neural network source separation structure capable of both robust multiple noise separation and reliability quantification. It will detect threats, separate targets, and quantify confidence by means of a multiple-neural net architecture. In addition to target separation, it will determine high-resolution acoustic source bearing, and information on the location, speed of movement, relevance (direction), and number of sources of each distinct target sound. The proposed technology will be flexibly deployable for data fusion with a hybrid array of hydrophones, vibration sensors and other acoustic transducers. In Phase I POC will demonstrate the feasibility of recognizing discrete sounds at 20 dB more than 95% of the time. In Phase II, POC will optimize a prototype system with a sensor array, and test it in context similar to a military installation or naval base.