On-Line Speaker Enrollment using Rhythmical Voices for Human Robot Interaction

In this study, we present a simple on-line speaker enrollment and identification among human-robot interaction (HRI) with intelligent service robots. For this purpose, speaker enrollment is performed through rhythmical singing voices or a simple game such as paper-scissors-rock. While the conventional enrollment methods frequently used in the security area should be cooperative, the proposed approach can be enrolled in a very natural way. After enrolling, the text-independent speaker recognition is accomplished by using the well-known mel-frequency cepstral coefficients (MFCC) and Gaussian mixture models (GMM). The experimental results reveal that the proposed approach yields better recognition performance in comparison to the results obtained by the conventional enrollment method.