Compression-dependent differences in hearing aid gain between speech and nonspeech input signals

Rebecca Warner Henning; Ruth Bentler

doi:10.1097/00003446-200508000-00004

A primary purpose of fitting hearing aids is to improve the audibility of speech; however, hearing aid gain is typically measured by using standardized nonspeech signals, e.g., swept pure tones, speech-weighted broadband noise, or modulated noise. When compression hearing aids are tested with these nonspeech input signals, the measured gain can be substantially different than if a real speech input signal were used. The purpose of this study was to systematically evaluate the effects of release time, compression ratio, and number of compression channels, as well as interactions of these parameters, on the gain difference between several common nonspeech hearing aid test signals and speech. It was hypothesized that the difference in hearing aid gain between static nonspeech signals and speech would increase as release time, compression ratio, and number of channels increased. Speech and several common nonspeech hearing aid test signals, matched at overall root-mean-square levels corresponding to average (65 dB SPL) and loud (80 dB SPL) conversational speech, were input into a master hearing aid circuit, and the gain of the circuit was measured in one-third octave bands. The hearing aid was programmed as a moderate-gain (23 dB) wide dynamic range compression instrument with a compression threshold of 50 dB SPL. The release time, compression ratio, and number of compression channels of the circuit were systematically adjusted by programming software. The one-third octave band gain differences between the nonspeech signals and speech were measured for all combinations of the compression settings. Multiple regression analysis was used to evaluate the effects of each compression parameter, and interactions of the parameters, on the gain difference between each nonspeech signal and speech. One-third octave band gain differences between nonspeech and speech signals (calculated as nonspeech signal minus speech signal) ranged from -3.1 to 10.4 dB, depending on frequency, nonspeech test signal, and input signal level. In most cases, the compression parameters accounted for more than 70% of the variance in gain differences between the speech and nonspeech signals. At an input level of 65 dB, increases in the release time and compression ratio led to an increase in the gain difference between most nonspeech signals and speech at most frequencies. Increases in the number of channels caused an increase in the gain difference when the spectra of the nonspeech signals differed from the speech spectrum. The effects of release time and number of channels increased as the compression ratio increased. At an 80 dB input level, increasing the compression ratio led to a decrease in the gain difference between the nonspeech signals and speech. Release time and number of channels had little to no effect at the higher input level. The compression parameters of release time, compression ratio, and number of compression channels explain most of the variance in differences in hearing aid gain between nonspeech and speech signals. It may be cumbersome, however, to quantitatively define this relationship for all hearing aid circuits. It is therefore recommended that hearing aid "use" gain or output be measured with a real speech signal. If a nonspeech signal must be used, then it should have spectral and temporal properties that are similar to speech.

Compression-dependent differences in hearing aid gain between speech and nonspeech input signals

View Online

Abstract

Details

Metrics