Real-time Detection of Fricatives using Audio and Video Signal Processing

In this paper, an experimental work on real time detection of fricatives using audio and video signal processing is presented. The Viola-Jones algorithm is used to detect the face and mouth of the speaker in the real-time video input taken using laptop webcam. Then the vertical and horizontal mouth openings of the speaker are estimated using MATLAB program. A filter bank method is used for estimating spectral centroid of the audio signal in real-time. The vertical and horizontal mouth openings estimated from the video data and the variations in spectral centroid estimated from the audio data are used for spotting unvoiced fricatives in /VCV/ syllables.