EUROSPEECH 2003 - INTERSPEECH 2003
A new and potentially important method for predicting, analysing and comparing responses of hearing aid algorithms is studied and presented here. This method is based on a time-frequency representation (TFR) generated by a computational auditory model. Hearing impairment is simulated by a change of parameters of the auditory model. To simulate the basilar membrane (BM) filtering part of the auditory model we propose a single parameter control version of the gammachirp filterbank and for simulating the neural processing in the auditory pathway we propose a signal processing model motivated by the physiological properties of the auditory nerve. This model then interprets the information processing in the auditory pathway through the use of a TFR called the auditory TFR (A-TFR) which matches the standard spectrogram in terms of both time and frequency resolution. Two popularly used hearing aid algorithms, a two channel wide band system and a nine channel compression system, are simulated and are used to compensate the impaired auditory model. The responses of the compensated system, in terms of the acoustic-phonetics cues that characterise speech intelligibility, are analysed and compared with one another and with that of a normal auditory system. It is shown that although the nine channel compression algorithm performs better than the two channel system both the hearing aid algorithms distort severely the acoustic-phonetic cues.
Bibliographic reference. Kiran, G.V. / Sreenivas, T.V. (2003): "A novel method of analysing and comparing responses of hearing aid algorithms using auditory time-frequency representation", In EUROSPEECH-2003, 61-64.