Facial Changes during the Pronunciation of Velar and Palatal Consonants in IR Images

Abstract

Speech is one of the most complex non stationary signals and is also the easiest way of communication between human beings. Speech signals can be exaggerated by the behavior and emotions of speakers. Facial expression can be recognized through electromyography (EMG) signals. These signals from specific facial muscles are recorded for speech recognition and system automation. The human face is an area for displaying different emotions. The temperature distribution on facial muscles can be captured through R camera, which may be regarded as texture features of images. In this research work five male subjects were taken and their corresponding IR images at the instant of utterance of velar and palatal consonants were captured and their average distances were calculated. The analysis of the results showed that the average of difference images were more speaker dependent as compared to speech content. The average distances of palatal consonants are relatively more as compared to velar consonants.