p-ISSN 2083-389x   e-ISSN 2084-3127


Speech perception in noise: a comparison between sentence and prosody recognition

Marianne Van Zyl, Johan J Hanekom

JHS 2011; 1(2): EA54-56

ID: 882061


The perception of speech in the presence of interfering noise remains an important issue in the field of audiology. Successful perception of speech under adverse listening conditions is facilitated to a large extent by the redundancy of the speech signal. An important cue that contributes to the redundancy of the speech signal is prosody, or suprasegmental speech features. The present study investigated the acoustic cues of a particular prosodic pattern, validated its recognition in quiet, and assessed its recognition in noise by normal-hearing listeners. The prosody under investigation was conditional permission, approval or agreement. A collection of sentences were recorded from two speakers (one male, one female). Two versions of each sentence were recorded, one giving unconditional permission or approval and the other adding a condition which was subsequently removed from the digital recording to eliminate differences in content between the two versions while retaining prosodic differences. Recorded materials were validated in a group of normal-hearing listeners (n = 12) in a quiet listening condition. The recognition of the prosodic contrast was evaluated in a second group of listeners (n = 9) in speech-weighted noise, at three different signal-to-noise ratio’s (SNRs) and compared to recognition of words and sentences at the same SNRs. Findings indicated that the recognition of sentences and of words in sentences deteriorated significantly as the SNR deteriorated, while recognition of prosody did not, remaining significantly above chance, even at an SNR of -8 dB. These findings indicate the resilience of the prosodic pattern under investigation to the effects of noise.

Keywords: speech recognition, prosody, signal-to-noise ratio