Evaluation of speech intelligibility in short-reverberant sound fields

Objective

The purpose of this study was to explore the differences in speech intelligibility in short-reverberant sound fields using deteriorated monosyllables. Generated using digital signal processing, deteriorated monosyllables can lack the redundancy of words, and thus may emphasize differences in sound fields in terms of speech clarity.

Methods

Ten participants without any hearing disorders identified 100 monosyllables convolved with eight impulse responses measured in different short-reverberant sound fields (speech transmission index >0.6 and reverberation time <1 s), and we compared speech recognition scores between normal and deteriorated monosyllables. Deterioration was produced using low-pass filtering (cut off frequency = 1600 Hz).

Results

Speech recognition scores associated with the deteriorated monosyllables were lower than those for the normal monosyllables. In addition, scores were more varied among the different sound fields, although this result was not significant according to an analysis of variance. In contrast, the variation among sound fields was significant for the normal monosyllables. When comparing the intelligibility scores to the acoustic parameters calculated from eight impulse responses, the speech recognition scores were the highest when the reverberant/direct sound energy ratio (R/D) was balanced.

Conclusions

Although our deterioration procedure obscured differences in intelligibility score among the different sound fields, we have established that the R/D is a useful parameter for evaluating speech intelligibility in short-reverberant sound fields.