On the effects of masking of perceptual cues in hearing-impaired ears

Speech perception;Psycho-acoustic;Acoustic cues;Hearing impaired;Signal processing;Amplification strategy

Cole, Cliston Luther

关键词: Speech perception;    Psycho-acoustic;    Acoustic cues;    Hearing impaired;    Signal processing;    Amplification strategy;   

Others  :  https://www.ideals.illinois.edu/bitstream/handle/2142/98279/COLE-DISSERTATION-2017.pdf?sequence=1&isAllowed=y

美国|英语

来源: The Illinois Digital Environment for Access to Learning and Scholarship

【 摘 要 】

One of the goals of the Human Speech Recognition (HSR) group is to understand the strategy of the hearing-impaired (HI) ear in detecting consonants. It has been uniformly assumed that audibility is the main factor in speech perception, for both normal-hearing (NH) and HI listeners (Zurek and Delhorne, 1987). Based on an entropy measure, Trevino and Allen (2013) have shown that at most comfortable level (MCL) audibility is not the main issue for the HI ear. This observation is counter-intuitive. In this research group, we hope to find answers to the following questions: What is the strategy of each HI ear in detecting consonants? How can we determine the subject’s strategy? From the 3DDS findings of perceptual cues (Li and Allen 2011; Li et al. 2012), results from two perceptual masking experiments (Li and Allen 2011; Kapoor and Allen 2012), and analysis of work by Han (2011) and Trevino and Allen (2013), we generalize the errors made by an HI ear with up to four strategies. S1: The frequency of the consonant’s primary cue is varied by changing the vowels, which slightly moves the cue frequency. S2: The conflicting cues are varied. Different tokens of the same consonant have different confusions, due to conflicting cues. S3: The masking of the primary cue is varied. The primary cue for many tokens of the same consonant-vowel is highly correlated with the NH SNR 90 . S4: The number of conflicting cues is varied, as measured by the error entropy. The entropy of a token tells us something about the number of conflicting cues and/or about the ambiguity of the primary cue. In this research, we focus on one strategy, the masking of the primary cue on HI ears, and hope it will lead us in a positive direction of generalization. An extension of three consonant identification experiments is proposed, derived from Miller and Nicely (1955), Li and Allen (2011), and Kapoor and Allen (2012). Both Li and Kapoor showed that masking of primary cue and/or removing the conflicting cues can improve speech perception for NH ears. To determine the strategy of the HI ear in detecting consonants, we study consonant group error patterns. If we can establish error generalizability in the HI ears, we will gain insight into that ear’s decoding strategy.