Intelligibility of information in temporally desynchronized bands of speech
Christian E. Stilp, Keith R. Kluender, Joshua M. Alexander, Michael Kiefte · The Journal of the Acoustical Society of America · 2007
[Stilp et al., [J. Acoust. Soc. Am. 121, 3195 (2007)] demonstrated that acoustic measures of potential information, absent explicit linguistic information, are closely related to intelligibility of locally time-reversed sentences. Here, this finding is replicated in two experiments in which synchronization of speech energy within four spectral bands was perturbed in different ways. An ensemble of seven-syllable sentences was synthesized at three different speaking rates (2.5, 5.0, 10 syllables/s) and filtered by one-third-octave Butterworth filters centered at 335, 850, 2135, and 5400 Hz. In Experiment 1, onset of the 335- and 5,400-Hz bands was delayed by multiple durations (0, 25, 50, 100, 200, 400, 600 ms). Across speaking rates, intelligibility was a nonmonotonic function of delay with local minima corresponding to roughly one syllable. Smaller and larger delays yielded better performance. In Experiment 2, bands were nonlinearly scaled in time (jittered) to preserve overall duration with local rate adjusted to preserve the same overall degree of desynchronization as Experiment 1. Performance monotonically decreased with increasing jitter. Overall, results closely mimic previous findings for locally time-reserved sentences, and provide support that physical acoustic potential information closely relates to intelligibility of temporally distorted speech. [Work supported by NIDCD.]