Recognize the most dominant person in multi-party meetings using nontraditional features

Recognize the most dominant person in meetings is a prerequisite of human-computer interaction and artificial intelligence in meeting environment. This paper provides a novel method to recognize the most dominant person in meetings by analyzing features such as speaking length and speaking energy of each speaker in multi-party conversation scenario and uses the audio-visual meetings in the AMI Corpus to test the approach. Experiments show that although our features are simple, the results are promising. Finally, the least dominant person in meetings is also recognized with the same approach.