Study And Implementation Of Voiceprint Recognition Technology Over Concurrent Virtual Machines

In the technical field of the authentication, using the cipher to encrypt and decrypt which is the most common and simplest technique has been widely used. However, there is a serious shortcoming of this authentication methods. Simple passwords are easy to crack while complex passwords are easily forgotten. With the explosive growth of the Internet system, in-depth study of computer technology and cryptography, cipher security is facing a severe test, the traditional password authentication is difficult to meet needs of safety and convenience in modern life.In contrast, biometrics has become a very good technical innovation breakthrough. Voiceprint identification(also known as speaker recognition) is used to identify a speaker identification by voice biometrics data. People’s channel structure and speech habits are specific and long-term stable, which makes everyone’s voices are unique. Voiceprint identification is extracted voiceprint representative voice from the speaker in order to identify the speaker’s true identity. Because of voiceprint recognition economical, non-contact, accuracy characteristics that can break local and time restrictions, it is getting more and more used on confirming the identity of social security, public security, judicial case detection, in favor of a bank or securities identification, national security and other fields.However, the process of collecting the speaker’s voice is often severely affected in practice by external environment, voice channels, hardware devices and methods of processing speech signal, so that the recognition rate, the false accept rate and false reject rate of the voice recognition system which uses the distorted voice are difficult to achieve the experimental results. Meanwhile, voiceprint identification and training model involve a lot of data processing algorithms, execution in order on a single computer of these huge computing tasks will seriously affect the efficiency of the system.To solve these problems, this thesis will research, design and implementation support concurrent virtual machines voiceprint recognition system, including the following Work:1.Study on how to concurrent parallel virtual machines technology into voiceprint recognition system;2. We propose a design on concurrent virtual machines voiceprint recognition system based on the application scenario of Social Security identification over phone;3. Optimize process of voice modeling and confirmation of recognition, improve the feasibility and applicability of the system;4. Based on the application scenario phone features and voice system, establish a reasonable voice sample library;5. Design and develop a virtual machine server- responsible for the services of voiceprint identification, such as model training, voice confirmation and Web services;6. Assess the overall performance of the system, including the voiceprint recognition rate, operating results and the performance of concurrent parallel virtual machines.