A joint Feature Extraction and Data Compression Method For Low Bit Rate Transmission In Distributed Acoustic Sensor Environments

Unattended passive acoustic sensors are widely used for remote battlefield surveillance, situation awareness and monitoring applications. They are rugged and reliable and can be left in the field for a relatively long period of time after deployment. However, the current strategy of deploying expensive and large acoustic arrays of several microphones is being shifted toward deployment of a modest quantity of low cost, low power, small size wireless sensors that collectively form a sparse distributed array. This new paradigm is expected to provide many benefits including much better coverage of the surveillance area, improved spatial resolution for separating multiple closely spaced targets, potential for dynamic array configuration, and widespread use in urban warfare, homeland security, etc. Nevertheless, among the barrier issues is the development of a very low bit rate data compression and encoding system that can be integrated with the existing sensor packages and implemented on generic processing boards. Moreover, effective strategies for large deployment are needed to yield sensor networks with great robustness to various operating and environmental factors. Based upon our successful Phase I research, Information System Technologies, Inc. (ISTI) will further develop and fine-tune the new joint subband detection, feature extraction, and data compression system for transitioning to hardware implementation and prototyping. All the operations will be performed on a generic off-the-shelf processing board that contains local A/D and I/O interfaces, a micro-computer, and a wireless transmitter. This research will also lead to new methodologies for effective and economical sparse distributed sensor deployment that offer robustness to sensor location error, sensor failure, coherence loss, wavefront perturbations, and much better spatial resolution. The effectiveness of the systems developed in this Phase II research will be demonstrated in two field testing that involve multi-target scenarios and various sparse distributed sensor configurations. Testing will be done based upon the received signals at the mater station for data fusion, direction of arrival estimation and target classification.