Analysis and Visualization of Group Discussion Based on Sound Source Angle Obtained Using Kinect

Abstract

This study proposes a method to analyze and visualize the utterance sequence and transition of a face-to-face group discussion. This method uses sound source angles obtained using the Microsoft Kinect audiovisual sensor. An instructor organizes several learner groups and conducts group discussions. When these group discussions are underway concurrently, it is difficult for the instructor to effectively monitor all groups. To solve this problem, we propose a method to analyze and visualize the utterance sequence and transition of each group automatically by using only sound source angle data that is estimated automatically using the four-microphone array in the Kinect sensor. We developed a program to extract sound source angle data from the Kinect sensor, transform it to designated speakers, and visualize the utterance sequence and transition to enable the instructor to recognize the state of a group. Equipment setup and data analysis are simple and easy in this method. We conducted an experiment and the results showed that almost all utterance sequences and transitions could be visualized successfully using the proposed method.