Decorrelation of noise by means of Error Templates

OBJECTIVE: To explore the utility of adaptive techniques based on explicit characterization of interfering sounds in the design and developmetn of robust acoustic detection and classification. DESCRIPTION: Previous work (Patent 4,802,231) has shown that with proper adaptive learning schemes, interfering acoustic signatures (which tend to obscure or be misrecognized as one of the desired signatures) can be automatically identified, characterized, compared to and deleted from an incoming acoustic data stream. This process enhances the effective signal to clutter ratio and may expose a desired signature buried in a very obscure environment. In a typical use of this approach, sounds that tend to obscure or be trainig and on the fly, into templates defining error classes. These templates are then used to identify and suppress signals from template matching against reference templates. RESEARCH: Develop and code algorithms and exercise in a suitable simualtion testbed. Effort focused on demonstrating ability to perform detection and upon classificaiton accuracy, rather than on real time performance.