Preprocessors for Noisy Speech 
Institution: Signition Inc.; P.O. Box 1020, Los Alamos, NM 87544-1020 
Principal Investigator: Dr. George Zweig; (505) 662-6855, zweig@earwig.lanl.gov 
Objectives: The objective of this project is to develop a preprocessor for speech 
recognition systems operating in noisy environments. The preprocessor, consisting 
of a nonlinear inhomogeneous transmission line, will be realized in software, 
although realization in hardware in FYgl should be possible. More specifically we 
will: 
1) Develop a nonlinear transmission line preprocessor that accurately simulates the 
mechanics of the mammalian inner ear at all sound pressure levels. 
2) Preprocess speech with the nonlinear transmission line and show that there is a 
substantial improvement in the signal to noise ratio. 
3) Assess the desirability and feasibility of implimenting either a digital or analog 
transmission line on a chip and using it as a preprocessor in the CMU, BBN, or 
MIT DARPA funded speech recognition systems. 
Recent accomplishments." A speech preprocessor based on a nonlinear transmission 
line model of the cochlea 1 was proposed. 2 Sounds of different intensities, such as 
voiced and unvoiced speech, are analyzed in radically different ways. The Q's of 
the preprocessor's nonlinear filters vary withinput amplitude, higher Q's (longer 
integration times) corresponding to quieter sounds. In addition, for quiet sounds, 
the preprocessor acts like a laser that traps and amplifies low level signals, thereby 
aiding in their detection and analysis. 
Work planed/or FYgO: Analytical and numerical methods for solving the non- 
linear transmission line equations will be developed. 
References 
1. G. Zweig, 
. 
"Cochlear mechanics," accepted for publication in J. Acoust. Soc. 
Am., 1990. 
G. Zweig, "Auditory speech preprocessors," in Proceedings o/ the DARPA 
Speech and Natural Language Workshop, Morgan Kauffman, 1989. 
456 
