The Text-Dependent Chinese Speaker Recognition System Based on the Universal Individual Identification

Abstract

In the past few decades, with the increase and popularization of smart devices, the security problem of device certification has been derived. Biometric authentication is becoming more and more popular, and contactless speaker recognition is one of its elegant forms. For the research of speaker recognition, most researchers are training based on a large amount of data. But in real life, for the use of a product, it is difficult to obtain a large amount of user data in a short time, so the setting of this user data is very important. Therefore, the purpose of the paper is to find the universal individual utterances in the Chinese system. We created a dataset of 30 speakers about the Chinese Pinyin table in the paper. The focus of this research is the use of initials and finals in Chinese Pinyin. According to the experimental results, the use of certain initials and finals in Chinese Pinyin helps to effectively determine the identity of the speaker. In the paper, we successfully found Chinese disyllable and trisyllable characters that can effectively identify the speaker. These Chinese characters can be used to develop text for any speaker-based authentication system.