Speech recognition and frequency selectivity for hearing impaired listeners

The correlation between frequency selectivity and speech-syllable recognition in noise for hearing-impaired subjects was examined in four groups of subjects. Auditory filters were used as a measure of the subjects´ frequency selectivities. Naturally spoken consonant-vowel pairs were used as tokens for the speech-syllable recognition task. Speech recognition scores in the various levels of noise were found to be weakly correlated with the sharpness of the auditory filters. Simulation results from an auditory periphery model, utilizing subject auditory filters as the first stage, are examined. Spectral information from the model is analyzed for possible speech cues. In general, a subject´s speech recognition ability cannot be accounted for in the simulated output of the subject´s auditory periphery