Speech was mixed with different noise signals and then processed according to the well-known noise reduction method of 'spectral subtraction'. Three different algorithms were examined. The speech signals were subjected to a four alternative forced choice (4AFC) test. Both the processed and unprocessed signals were evaluated psycho-acoustically and objectively. Speech intelligibility was measured with the 4AFC test by presenting the signals via headphones to a group of normal-hearing and to a group of hearing-impaired listeners. The intelligibility scores were compared with the intelligibility scores predicted from a modified version of the Speech Transmission Index (STI). It appeared that although the noise reduction algorithms reduced the noise level, they did not improve the measured speech intelligibility, either for normal-hearing or for hearing-impaired listeners. This, however, was inconsistent with the scores estimated from STI, which erroneously predicted a significant improvement in intelligibility due to the noise reduction processing.
|Number of pages||6|
|Journal||Scandinavian Audiology, Supplement|
|Publication status||Published - 1993|
- Adaptive filtering
- Hearing loss
- Speech recognition