The advantage of spatial and vocal characteristics in the recognition of competing speech
Martin D. Vestergaard, D. Timothy Ives, Roy D. Patterson · Proceedings of the International Symposium on Auditory and Audiological Research · 2009
In multi-speaker environments, listeners take advantage of a variety of cues that characterize the target and distracter speakers to improve speech recognition. Spatial cues like interaural time and intensity differences provide binaural unmasking and a better-ear advantage. Vocal characteristics such as pitch and resonance scale help to disambiguate concurrent speech. Temporal misalignment of competing speech signals can improve recognition by virtue of ‘listening in the dips’. In this paper, we review a series of experiments on the advantage of spatial and vocal characteristics in the recognition of concurrent speech. Syllable pairs were synthesized to simulate different speakers, and the recognition of syllables that varied in spatial and vocal characteristics was measured. The effect of temporal glimpsing was measured by aligning the temporal envelopes of the competing signals in a controlled way. The results show that spatial and vocal cues compete to provide selectivity of concurrent speech sounds. When they are clearly separated in space, vocal characteristics can only further improve performance marginally. However, when they are temporally and spatially aligned, a substantial advantage can be derived from the vocal characteristics. The paper discusses the interaction of spatial and vocal cues, and the patterns of syllable confusions that listeners make.