Authors: Pierre H Guilleminot1*; Tobias Reichenbach2
1Imperial College London
Background Speech is made of sentences that can be broken down words, syllables, all the way down to phonemes. To compute such a signal, the brain must first segment it into smaller subunits. This parsing process relies on neural oscillations in the auditory cortex, especially in the delta and theta frequency bands (1-4Hz and 4-8Hz). The latter is typically associated with the tracking of incoming syllables and can be modulated through and somatosensory system, therefore potentially modulating speech processing.
Methods Here, we used vibrotactile pulses derived from the syllable timing to improve speech-in-noise comprehension in healthy subjects. The pulses were aligned to the perceptual centre of syllables. In addition, we evaluated the effect of different lags between the auditory and tactile streams. Speech comprehension was measured as the proportion of keywords understood in semantically unpredictable sentences. Neural activity was measured using electroencephalographic recordings (EEG) in response to continuous speech.
Results Vibrotactile feedback was found to modulate speech-in-noise comprehension and even improve it when auditory and tactile streams are synchronized, with no subjective discomfort for the subject. Moreover, speech comprehension and neural encoding of both auditory and tactile stimuli were all found to be modulated by the lag between sensory streams. Finally, the subjective comfort of subjects was found to be correlated with biomarkers of audio-tactile speech integration.
Conclusions Our results therefore present evidence of the role of neural oscillations in speech processing as well as the possibility to use somatosensory information in order to improve speech in noise comprehension, thus opening avenues for novel multisensory hearing aids.