Biometric characteristics such as voice could be viewed as a combination of physiological and behavioral traits. Indeed, the voice depends not only on physical features such as vibrations of vocal cords and vocal tract shape, but also on behavioral features, such as the state of mind of the person who speaks. The most significant factor affecting automatic voice biometric performance is the variation in the signal characteristics, even from the same speaker. To tackle this problem, speaker model created by the voice biometric system need to have the ability to adapt the speaker variability. In this book, the cross match technique is proposed to provide a speaker model that can adapt to variability over periods of time. In addition to that, cross match also adds the dimension of multimodality to the system at the score-level when the similarity score from cross match can be fused with the score from the default speaker modeling used in the system.
A. Kamarul Ariff Ibrahim, Kamarulafizam I., Sheikh Hussain