A particularly common complaint of listeners with mild to severe levels of sensorineural hearing loss is difficulty understanding speech in a background of noise. For a moderate hearing impairment, an increase of between 2.5 and 12 dB Signal-to-Noise-Ratio (SNR) is required to achieve similar speech discrimination scores to those of normal hearing.
Several modern hearing aids attempt to address this problem using simple adaptive response filters. Normally, this involves a filter having slope and/or passband sensitive to the frequency and amplitude of the background noise, which results in an attenuation of both signal and noise. A number of hearing aids are currently available with such systems. Size and power constraints limit the implementation to analog circuits, possibly with digital control. The current Level-dependent Automatic Signal Processing (ASP) strategies have been divided into three categories: 1) Bass Increases at Low Levels (BILL; i.e., lows are reduced at high input levels); 2) Treble Increases at Low Levels (TILL; i.e., highs are reduced at high input levels); and 3) Programmable increases at Low Levels (PILL; i.e., either low or highs are reduced at high input levels).
In the BILL approach, the bass response is attenuated relative to higher-frequency sounds, as the input sound pressure level increases. This may be useful when the added noise is predominantly low-frequency, as the masking effect of the noise is reduced. The BILL approach has been realized in several "noise-reduction" circuits.
In the TILL approach, the hearing aid response is flattened at high input levels. As the input sound pressure level is reduced, the treble response is increased. At lower levels, low-frequency (bass) amplification is reduced (i.e., treble increases) to reduce noise masking effects. The TILL approach has been realized in the "K-AMP" circuit (note that the K-Amp circuit contains other features as well, including mild compression.)
Programmable (PILL) hearing aids use frequency-dependent compression or multi-band compression to achieve their noise reduction capabilities. A wide range of capabilities is included in this category. Typically, the user may select from several frequency responses as desired for different listening conditions.
These analog signal processing methods reflect the state of the art in modern hearing aids. Unfortunately, these approaches have been implemented prior to careful behavioral testing to establish whether or not they offer any real advantage for hearing-impaired listeners (HILs).
The evaluation of an ASP scheme as part of an actual hearing aid is necessarily complicated by a variety of factors, including the other electroacoustic characteristics of the hearing aid, and the appropriateness of the hearing aid's gain function for each HIL.
As an alternative, we have studied various ASP systems using precise simulations of the ASP functions. In a series of pre-production experiments, we have investigated DSP-based simulations of various types of adaptive systems in various types of background noises. The present paper reports the results of one such simulation and evaluation.
Cite as: Jamieson, D.G., Brennan, R.L. (1992) Evaluation of speech enhancement strategies for normal and hearing-impaired listeners. Proc. ETRW on Speech Processing in Adverse Conditions, 155-158
@inproceedings{jamieson92_spac,
author={Donald G. Jamieson and Robert L. Brennan},
title={{Evaluation of speech enhancement strategies for normal and hearing-impaired listeners}},
year=1992,
booktitle={Proc. ETRW on Speech Processing in Adverse Conditions},
pages={155--158}
}