SEGMENTING METHOD OF AUDIO VISUAL RECORDING SUBSTANCE, COMPUTER STORAGE MEDIUM AND COMPUTER SYSTEM

摘要

PROBLEM TO BE SOLVED: To obtain a start of speaker identification of a voice base in order to accurately segmentalize individual presentation through automatic picture recognition by extracting an audio segment corresponding to a video frame segment and applying an acoustic clustering method to the audio segment. SOLUTION: Source video 3601 is analyzed (3602) to find a slide region. The audio channel of the video 3601 is extracted (3603) for the region of the video 3601 corresponding to the slide segment. The extracted audio segment is clusterized (3604) for every speaker and each of the obtained clusters between audio segments is considered to be based on a single speaker. Audio segments of the same speaker cluster are combined (3605) and a source specific speaker model is trained for each combined audio segment (3606). The audio channel of the source video 3601 is segmentalized for every speaker by speaker recognition (3607).