Evaluation of a wide range of amplitude-frequency responses for the hearing impaired

The long-term average frequency spectrum of speech was modified to 25 target frequency spectra in order to determine the effect of each of these spectra on speech intelligibility in noise and on sound quality. Speech intelligibility was evaluated using the test as developed by Plomp and Mimpen (1979), whereas sound quality was examined through judgments of loudness, sharpness, clearness, and pleasantness of speech fragments. Subjects had different degrees of sensorineural hearing loss and sloping audiograms, but not all of them were hearing aid users. The 25 frequency spectra were defined such that the entire dynamic range of each listener, from 5 dB above threshold to 5 dB below UCL, was covered. Frequency shaping of the speech was carried out on-line by means of Finite Impulse Response (FIR) filters. The tests on speech reception in noise indicated that the Speech-Reception Thresholds (SRTs) did not differ significantly for the majority of spectra. Spectra with high levels, especially at low frequencies (probably causing significant upward spread of masking), and also those with steep negative slopes resulted in significantly higher SRTs. Sound quality judgments led to conclusions virtually identical to those from the SRT data: frequency spectra with an unacceptably low sound quality were in most of the cases significantly worse on the SRT test as well. Because the SRT did not vary significantly among the majority of frequency spectra, it was concluded that a wide range of spectra between the threshold and UCL levels of listeners with hearing losses is suitable for the presentation of speech energy. This is very useful in everyday listening, where the frequency spectrum of speech may vary considerably.