| OCR Text |
Show treatf4,in'numerous references rl,2,31. In addition, the details of the physiological system or of its analogs are complex, resisting concise mathematical modelling, and hence have not, to date, been useful in solving the usual signal processing problems -- noise removal , parameter extraction and transmission, etc.. A mathematical transformation, on the other hand, can relate more easily to these problems. One such transform, the short-time Fourier transform, the properties of which are well-understood, provides both a forward and a reverse mapping to a domain which resembles the analysis domain of the ear. However, even the most superficial examination of the ear's physiology reveals weaknesses in the short-time Fourier transform as a model of auditory analysis. The transform conforms to the ear-property hypothesized by Helmholtz [4I, and later corroborated by Pekesy [5] and others, wherein spatially selective time-limited frequency analysis is performed. It fails, however, to emulate other aspects of the behavior observed by Bekesy. In particular, Bekesy observed that the basilar membrane behaves as a non-uniform or dispersive transmission line such that tones travel a distance inversely proportional to their frequency where they are sensed and then are rapidly attenuated. He further observed that the envelope of a tone traveling the length of the membrane maintains its shape as it moves the 35mm to the apex of the membrane. In other words, the mechanical analysis performed by the inner ear was reported |
