SBIR-STTR Award

Tanks of different models have unique acoustic signatures. These signatures can be used for detection and for identification. One approach is to analyze the signature in the spectral domain, using the Fourier transform, and comparing new sound samples with known templates. Problems arise when signatures vary or are corrupted by noise. The problemof signature recognition has much in common with speech recognition, an active field for several decades. We propose to apply traditional and novel speech-recognition techniques to the identification of acoustic signatures. Preliminary tests in this proposal, applied to Army-supplied signature data, show error rates of less than 1%. In Phase I wewill first develop "traditional" speech recognizers, which use non-Fourier features that have proven more robust, such as cepstral and linear-predictive-coding (LPC) coefficients. As back ends, we will investigate both hidden Markov models (HMMs) and neural algorithms. The algorithms will be modified to fit the problem domain in terms of frequencyranges and time scales. We will also apply more recent algorithms: the auditory image model (AIM) mimics the human auditory system; learning vector quantization (LVQ) is a novel neurally inspired, Bayesian optimal algorithm. The systems will be compared to the more traditional Fourier-based approach and evaluated for noise robustness. In Phase II, we will develop a noise-robust, real-time system to be evaluated in the field.

Benefits:
Commercial applications for robust acoustic-signature recognition are numerous. In particular, the field of speech-recognition is on the verge of leaving the laboratories and entering theconsumer market. Non-speech applications include fault detection in machinery and medical diagnostics.

Commercially available speech recognition ISR) systems perform adequately in quiet surroundings for medium-sized vocabularies, but performance plummets in noisy environments. Achieving noise robustness is a must before SR is widely deployed by the Army or becomes accepted in the consumer electronics market. Similarly, acoustic-signature recognizers (ASR) for army vehicles can be greatly improved over existing spectral front ends. During Phase I, we proved the feasibility of several novel, noise-robust preprocessors and training algorithms, in particular those based on cochlear modeling and neural-like training. We also showed how SR algorithms can successfully be applied to ASR. We will build and deliver a two-component system to be interfaced with the Army Research Laboratory's (ARL's) Real-world Interface robot. First, a continuous-speech, noise-robust SR system will be used with ARL's natural-language-processing (NLP) system to allow voice control of the robot. This system will be enhanced by an ultrasonic pulse microphone for lip reading. Second, an ASR system, also placed on the robot, will identify vehicles by their acoustic signatures and automatically analyze the signature to identify the most salient features that give away the vehicle. This system will interface with the visual-acoustic recognizer currently under development for the ARL.

Benefits:
Noise-robust speech-recognition is useful to the military (cockpits, ground crew, command centers, field operators, battlefield visualizers) and commercial sector (telephony, cell phones, answering machines, internal automobile environments, and home electronics) Acoustic-signature recognizers are useful for law enforcement and machinery fault detection.