A Speaker Recognition Method Based on Personal Identification Voice and Trapezoidal Fuzzy Similarity

A text-dependent speaker recognition method is proposed using trapezoidal fuzzy similarity function to measure the similarity of voice features between a test user and the registered speaker who has nearest distance. The trapezoidal fuzzy similarity function is constructed based on three-time data recorded during enrolment process as personal identification voice (PIV) and statistical data of an individual recorded many times in a long time period to cover the intra-variation. A set of acoustic voice features is also introduced to present some general speaker and text dependent characteristics that are effective for modeling PIV, thus allowing to capture the intervariation from one speaker to another. The experimental results on 24 speakers recorded in four different sessions show that, without false acceptation, the proposed system can decrease 30.05% of false rejection cases, compared to the traditional nearest neighbor approach. The focus of this work is on applications which require fast processing and few burdens for users.