Implementation Of Text-Dependent Speaker Verification System In Embedded-Linux Environment

As a kind of biometric identification technology, speaker recognition is to recognize people's identity from its voice, which contains physiological and behavioral characteristics specific to each individual. It has developed quickly within this decade and has important application in law, military affairsand bank, etc. And recently recognition technology based on PC can solve the speaker independent ,large vocabulary, continuous recognition problems. With the development of embedded systems, speaker recognition technology based on embedded systems has became new and important aspect of development in this field. To meet the wants or needs of people's increasing taste, In the process of developing embedded systems, how to achieve a friendly and attractive human-computer interface become a new tendency.In this essay, a "text-dependent speaker verification system" is designed and implemented by Qt and runs on Embedded-Linux platform based on ARM processor. And my major work is:1. Plain the whole design Plan suitable for Embedded-Linux platform.2. Use MATLAB to implement and parser all algorithm, test and analysis by real speech.3. Design and implement main formwork by Qt, can response user's request, input and output user's speech, communication with verificationsubsystem, etc.4. Transplantation all code to Embedded-Linux platform, run program by real speech.5. Discuss for Float-Point calculation in Embedded-Linux platform. The Soft-float is faster than Linux module.