Speech processing for the hearing-impaired: successes, failures, and implications for speech mechanisms

Abstract

People with sensorineural hearing impairment typically have more difficulty than normally hearing people in understanding speech in the presence of background sounds. This paper starts by quantifying the magnitude of the problem in various listening situations and with various types of background sound. It then considers some of the factors that contribute to this difficulty, including: reduced audibility; reduced frequency selectivity; loudness recruitment; and regions in the cochlea which have no surviving inner hair cells and/or neurones (dead regions). Methods of compensating for the effects of some of these factors are described and evaluated. Signal-processing methods to compensate for the effects of reduced frequency selectivity using the output of a single microphone have had only limited success, although methods using multiple microphones have worked well. Amplitude compression can compensate for some of the effects of loudness recruitment, allowing speech to be understood over a wide range of sound levels. The exact form of the compression (fast-acting versus slow-acting, single-channel versus multiple channel) does not seem to be critical, suggesting that the relative loudness of different components of speech, and dynamic aspects of loudness perception do not need to be restored to “normal”.