Development of a large-item environmental sound test and the effects of short-term training with spectrally-degraded stimuli

Abstract

Objectives: Accurate identification of environmental sounds plays an important role in maintaining listeners' awareness of their environment, and is a major concern for cochlear implant patients. Although research indicates that decreased spectral resolution has a negative effect on environmental sound identification, little is known about the processes underlying perceptual adaptation to spectrally-degraded input. The goals of this study were (1) to develop a test of environmental sound perception containing a large variety of easily identifiable and familiar sound sources, represented by multiple exemplars, and (2) to examine whether auditory training improves listeners' identification of spectrally-degraded environmental sounds.

Design: In experiment 1, familiarity ratings and identification accuracy were obtained for 21 normal-hearing subjects for 48 environmental sound sources; there were 4 exemplars of each sound source, for a total of 192 stimuli. A second test was developed using a subset of 40 sound sources (4 exemplars each, for a total of 160 stimuli). In experiment 2, seven normal-hearing subjects (who did not participate in experiment 1) were asked to identify spectrally-degraded environmental sounds processed by a four-channel noise-band vocoder. The second stimulus set developed in experiment 1 (40 sound sources, 4 exemplars each) was used in experiment 2. The subjects were tested in a pretest-posttest design with five training sessions between the pretest and the posttest. The training sounds were selected individually for each subject, and comprised one half of the sound sources that were misidentified in the pretest. Each sound source used in training was represented by two exemplars. During training, subjects received trial and block feedback. For each incorrect response, subjects were allowed to replay the stimulus up to five times after being shown the correct response.

Results: In experiment 1, listeners' average identification accuracy was 95% correct, with 178 of all sounds identified with an accuracy of 80% or more. The average identification accuracy of the 160 sounds selected for experiment 2 was 98% correct, and their average familiarity rating was 6.39 (on a 7-point scale). In experiment 2, the average identification accuracy of spectrally-degraded sounds was 33% correct on the pretest. However, after training, average identification accuracy across all sounds improved to 63% correct on the posttest. The largest improvement (86 percentage points) was obtained for the sound exemplars used during training. The identification accuracy for alternative exemplars of the training sounds (that referenced the same sources) improved by 36 percentage points. Finally, the identification of sound sources not included in the training, but perceived with equal difficulty on the pretest, improved by 18 percentage points.

Conclusions: These results demonstrate positive effects of training on the identification of spectrally-degraded environmental sounds and suggest that training effects can generalize to other sound exemplars and sources, although with a reduced magnitude of improvement. The findings also indicate a timeline for initial perceptual adaptation to spectrally-degraded environmental sounds, and provide a preliminary basis for incorporating environmental sounds into auditory rehabilitation programs for cochlear implant patients.