5th International Conference on Spoken Language Processing
Some research questions regarding the speech perception can only be answered with natural speech stimuli especially in noisy environment. In this paper we are going to answer a couple of questions on visual support of audio signal at speech recognition. How much support can give the video signal for the audio one? The impact of nature of the noise. How can help the visual information to identify the place of articulation? Does the voices of different class of excitation get the same visual support? In order to answer these questions we have performed intelligibility study on consonants between the same vowel supported or not by the speaker's image with different signal to noise ratio. The noise is either white noise or a mix of other speakers' voice.
Bibliographic reference. Czap, Laszlo (1998): "Audio and audio-visual perception of consonants disturbed by white noise and 'cocktail party'", In ICSLP-1998, paper 0445.