Skip to main content
Text Size: sML

Second Biennial Hearing Aid Research and Development Conference

Hearing Aid Conference Logo

September 22-24, 1997
National Institutes of Health
Bethesda, Maryland

Application of the Aided Audibility Index to Understanding of Compressed Speech

Pamela E. Souza, Department of Speech and Hearing Sciences, University of Washington, Christopher W. Turner, Department of Speech Pathology and Audiology, University of Iowa

Audibility of the speech signal is believed to be the primary factor in recognition of compressed speech (Souza & Turner, in press). The purpose of this study was to evaluate use of the Aided Audibility Index (AAI; Stelmachowicz et. al., 1994) in quantifying recognition of compressed and linear speech. Recognition of VCV nonsense syllables was tested for sixteen subjects with mild to severe sensorineural hearing loss. The entire set of digitally recorded syllables was presented in six conditions: linear speech at input levels of 55, 70 and 85 dB SPL, and compressed speech at input levels of 55, 70 and 85 dB SPL. To create the compressed conditions, a computer algorithm performed two-channel, wide dynamic range compression. Compression parameters were typical of those in commercially available wide dynamic range compression hearing aids. For each subject and condition, an Aided Audibility Index was calculated. This measure, which can be used with either compressed or linear speech, is similar to the Articulation Index and expresses an importance-weighted index of the amount of the speech signal which is audible to the listener. Each listener's AAI was compared with their recognition score to determine the contribution of audibility in each condition. Results indicated that audibility, as expressed by the AAI, accounts for approximately 80% of the variance in recognition for linear speech and 65% of the variance in recognition for compressed speech. Observed recognition scores were also compared with those predicted using an AAI transfer function for nonsense syllables. For both linear and compressed conditions, observed performance was poorer than that predicted by the AAI solely on the basis of audibility. These differences were greater in the compressed condition, suggesting that factors other than audibility may play a relatively greater role in recognition of compressed speech.

Return to Conference Abstracts List