Objective evaluation of hearing loss simulators

Abstract

SINTEF s department in Trondheim, Norway, has developed a hearing loss simulator. Thepurpose of the simulator is to reconstruct hearing impaired s perception of sound. Thusfar, the simulator has only been judged subjectively, in which solely candidates with confidentialityhave participated in the assessment. In this thesis, 20 normal hearing subjectshave attended. None of the individuals were acquainted with the simulator prior to thetesting. Hence, their sound experience was objective. In this manner, it has been carriedout an objective evaluation of the hearing loss simulator.Audiograms from three hearing impaired have been applied in the experiment. A set ofaudio files was selected to be sent through the simulator. The resulting output audio filesrepresented the respective hearing impairments. The equivalent procedure was performedon a sound file containing white noise.In addition to the hearing loss simulator, SINTEF has implemented a hearing-in-noisetest. The hearing impaired have performed the test with the original audio files, whereasthe 20 normal hearing subjects have listened to the audio files that were sent through thesimulator. The hearing impaired s results were compared to the results from the simulations.Deviation between the results could indicate the extent of the simulator s ability toreconstruct a representative sound perception of a hearing impaired.The distribution of the data samples, and the statistical analysis of the results have beeninvestigated. The analysis showed that the deviation between the true hearing impairmentand the simulated hearing impairment substantially depended on the audiogram.The results from the first and the third simulation were scattered over relatively largeranges, compared to that of the second simulation. A hypothesis test of 95% confidenceinterval was applied to the results, in order to test the true hearing impairments against thesimulated hearing impairments. Overall, the probabilities were not very convicting. Thelargest probability was 63.9%, whereas the least was 0.0%.If the evaluation was solely based on the probabilities, the simulator would be consideredunsatisfactory. Including subjective and outer factors in the evaluation, has influencedthe analysis of the quantitative experiment. Thus, it has been concluded thatthe simulator provides the user with representative speech perceptions for different hearingimpairments. However, observations have indicated that simulation of white noise israther inadequate.