IBN 97-28383 GOLDSTEIN Human hearing continuously monitors the frequency composition of sounds throughout the audio range (50-16,000 Hz). The inner ear allows the brain to monitor about 4,000 different overlapping frequency bands within this range. A model for how this is accomplished by the brain, the" simple spectrogram model," has been used widely throughout this century by scientists and engineers. The model represents auditory frequency analysis with 25 contiguous linear filters, each spanning 1/3 of an octave, with the brain monitoring the energy within short segments (10-250 milliseconds) of the filter responses. By assuming further that information received by the brain is limited by internal noise, this model has served as a basic framework for predicting human abilities to recognize speech and to discriminate changes in sounds. Limitations of the simple model are known, including failures to predict audible distortion tones generated by inner- ear nonlinearity, and sensitivity to signal waveform. In recent collaborative research with Bell Laboratory scientists, Dr. Goldstein discovered that detection of signal peaks rather than energy improves model predictions of discrimination phenomena. This revived interest in an earlier, "peak-detection," model of Dr. Goldstein's (1967) that was used to explain discrimination of waveform changes produced by modulation of sinusoidal carriers. As a result, evidence was found for two processes at each auditory filter in the peak-detection model. Because of other recent evidence that slow modulations within speech sounds contribute to their recognition, these newly recognized characteristics of hearing may be particularly relevant for audio science and engineering. In the present project, use of random signals minimizes confounding effects of inner-ear nonlinearity on interpretations of experiments. The project is exploiting this to develop further the model of auditory peak detection. In addition to upgradi ng the classical spectrogram model, this research is contributing knowledge of functions performed by neural mechanisms beyond the inner ear.
