AVSP 2003 - International Conference on Audio-Visual Speech Processing
September 4-7, 2003
In the experiments described in this paper, [aba] and [aga] speech sequences are combined in such a way that an [ada] stimuli is obtained, referring to the well-known McGurk effect. But contrary to the standard experiments, where audio [aba] and visual [aga] stimuli are combined, only audio signals are considered here, resulting in what is called a pure audio McGurk effect. The processing consists in modelling the audio signals by the classical linear prediction model and then linearly combining the [aba] and [aga] sequence LP filters that model the contribution of the vocal tract before resynthesis. The relation of the experiments with the problem of data representation and the nature of the audio-visual integration space is discussed.
Bibliographic reference. Girin, Laurent (2003): "Pure audio McGurk effect", In AVSP 2003, 139-144.