Contribution of oral periphery on visual speech intelligibility

Abstract

Visual speech recognition aims at improving speech recognition for human-computer interaction. Motivated by the cognitive ability of humans to lip-read, visual speech recognition systems take into account the movement of visible speech articulators to classify the spoken word. However, while most of the research has been focussed on lip movement, the contribution of other factors has not been much looked into. This paper studies the effect of the movement of the area around the lips on the accuracy of speech classification. Two sets of visual features are derived: one set corresponds to the parameters from an accurate lip contour while the other feature set takes into account the area around the lips also. The features have been classified using data mining algorithms in WEKA. It is observed from results that features incorporating the area around the lips show an improvement in the performance of machines to recognize the spoken word. © 2011 Springer-Verlag.