Visual Speaker Identification Using Lip and Body Movements

Speaker identification has been studied in many fields such as image processing, audio processing, artificial intelligence and speech recognition. Two of these areas are integrated together in order to identify the speaker. This research work is focused on two main approaches which are lip movements and body movements. These two approaches are combined together to achieve the speaker identification. Based on the proposed methodology, a speaker is identified in different scenarios, if there is a single speaker or if there is multiple speakers in the video or if the speaker’s lips are not in view. For evaluation purposes, we have used precision, recall and accuracy scores applied on manually created videos dataset.