The effect of compression on speech perception as reflected by attention and intelligibility measures

The purpose of this study was to investigate the effect of amplitude compression on speech perception as reflected by attention and intelligibility. Previous investigations of the effects of compression on intelligibility have shown inconsistent results. Additional measures were investigated in the present work because percent-correct measures of intelligibility are not a complete indicator of compression effectiveness. Listening effort was considered as an additional parameter. Two dual-task paradigms were constructed to measure listening effort associated with amplitude compression. In the first experiment, an auditory word recognition task was employed as the primary task and a visual motor tracking task as the secondary task. Monosyllabic words were mixed with speech-shaped noise at a fixed signal-to-noise ratio and compressed using fast-acting wide dynamic range compression. Participants with normal hearing performed a word recognition and a pursuit rotor task in single and dual-task conditions. Results showed that compressed speech decreased visual motor tracking performance but not word recognition as compared to linearly processed speech. In the second experiment, an auditory word recognition task again served as the primary task, and a visual lexical decision task was employed as the secondary task. In the secondary task subjects were asked to decide whether an item on a computer screen was a word or not. The visual lexical decision task was expected to interfere more with the auditory word recognition task as compared to the visual motor tracking task of the first experiment because the lexical decision task shares more similarities in processing modes with the auditory task. However, the results showed that the lexical task did not interfere with the auditory task and did not reduce the dual-task performance. Several explanations for this finding are proposed. The use of dual-tasks to measure listening effort can be affected by many factors. Additional research will reveal the particular dual-task methods that are best for evaluating compression