Developing Voice Separation System for Support Education Research: Determining Learner Reaction without Contact

Evaluating educational programs using traditional methods such as questionnaires and interviews require several trained workers. Therefore, we developed a voice separation system using the Microsoft Kinect sensor for support education research. This system records the voice of the learner in the evaluator’s absence and identifies the learner’s reaction by combining a position measurement technique with various sound source separation techniques. Through the performance evaluation and the implementation in an education program, we confirmed the effectiveness of this system.