Speaker characterization using adult and children’s speech

[摘要] Speech signals contain important information about a speaker, such as age, gender, language, accent, and emotional/psychological state. Automatic recognition of these types of characteristics has a wide range of commercial, medical and forensic applications such as interactive voice response systems, service customization, natural human-machine interaction, recognizing the type of pathology of speakers, and directing the forensic investigation process. Many such applications depend on reliable systems using short speech segments without regard to the spoken text (text-independent). All these applications are also applicable using children’s speech.This research aims to develop accurate methods and tools to identify different characteristics of the speakers. Our experiments cover speaker recognition, gender recognition, age-group classification, and accent identification. However, similar approaches and techniques can be applied to identify other characteristics such as emotional/psychological state. The main focus of this research is on detecting these characteristics from children’s speech, which is previously reported as a more challenging subject compared to adult. Furthermore, the impact of different frequency bands on the performances of several recognition systems is studied, and the performance obtained using children’s speech is compared with the corresponding results from experiments using adults’ speech.Speaker characterization is performed by fitting a probability density function to acoustic features extracted from the speech signals. Since the distribution of acoustic features is complex, Gaussian mixture models (GMM) are applied. Due to lack of data, parametric model adaptation methods have been applied to adapt the universal background model (UBM) to the char acteristics of utterances. An effective approach involves adapting the UBM to speech signals using the Maximum-A-Posteriori (MAP) scheme. Then, the Gaussian means of the adapted GMM are concatenated to form a Gaussian mean super-vector for a given utterance. Finally, a classification or regression algorithm is used to identify the speaker characteristics. While effective, Gaussian mean super-vectors are of a high dimensionality resulting in high computational cost and difficulty in obtaining a robust model in the context of limited data. In the field of speaker recognition, recent advances using the i-vector framework have increased the classification accuracy. This framework, which provides a compact representation of an utterance in the form of a low dimensional feature vector, applies a simple factor analysis on GMM means.

[发布机构] University:University of Birmingham;Department:School of Engineering, Department of Electronic, Electrical and Systems Engineering

[关键词] T Technology;TK Electrical engineering. Electronics Nuclear engineering