System Design And Implementation Of Voice Print Recognization And Verification On The Basis Of Sex Determination

In current criminal cases’ process,usual method of identifying criminal suspects depends on witnesses identifying or police officers’ experience. This way is lagging and slow on the accuracy and the efficiency of judgement, which not only badly results in the delay on the suspect’s judgement, but also leads to more serious consequences. According to the situation, this thesis assumes everyone’s voice print features which act as a personal identity had stored in the computer. And computer acting as a safe butler and effective assistent of police or justice could fast identify and validate criminals through the system to recognize the speaker in many of existing voice print models data under the condition of that police has got criminals’ voice data.To verify the suspect’s identity, this thesis comes up with a system of the combination of text-independent speaker recognization and verification based on sex determination.First, the system recognizes speaker’s sex.Secondly speaker recognization progress outputs maximum likelihood probability of target speaker model of the suspect on the basis of gender classification results.Finally the stage of speaker verification confirms the current speaker suspects the similarity with the results from the second output, to decide whether to "accept" the identity of the suspect.The system would help realize rapid identification of the suspect, accelerate the working efficiency, speed up the danger’s elimination, ensure the safety of people’s private property and personal safety.The system is consist of five subsystems which are voice print collection subsystem, voice print training subsystem, voice print recognition subsystem, the record of verification library management subsystem and user’s voice print management subsystem.Actually, voice print recognition subsystem includes three aspects which are the sex determination, voice print recognition and voice print confirmation. These five subsystems’ combination will together fullfill the function of this system.For the technology speaking, the system adopts the Play Framework on the Java Web development, using Mysql 5.5 to store all the user’s voice print states and suspects’ recognized record results, using the microphone as an equipment for voice collection, using Matlab to deal with the voice features with related algorithm.UML method is used in the thesis’ s design process to detail requirements and design ideas.In the process of system development, test-driven development technology is used to ensure the code of scalability, maintaination, reusable and efficient development.