Multichannel compression processing for profound deafness.

Abstract

Three-channel amplitude compression followed by frequency shaping was used to process test sentences for five profoundly deaf subjects, and the recognition scores were compared to scores achieved with frequency shaping only. At preferred levels, the scores of three of the five subjects showed a statistically significant but not dramatic advantage for compression; the averages of the scores for these three subjects were 29.2 percent for the uncompressed speech and 39.5 percent for the compressed speech. The preferred-level scores of the other two subjects did not show a statistically significant advantage for compression; averages were 33.7 percent for uncompressed speech and 36.1 percent for compressed speech. Tests at input levels 10 dB and 15 dB below preferred levels were also given to four of the subjects (the fifth had to leave the experiment early). In the reduced-level tests all four subjects showed a statistically significant advantage for compression. The averages of reduced-level uncompressed scores for the four subjects were 10.7 percent and 15.2 percent at -15 dB and -10 dB levels, compared to compressed scores at these levels of 31.7 percent and 32.5 percent. When visual cues were added to the auditory presentation in an exploratory experiment with one subject, the benefit of compression carried over into the higher scores.