Research On Speech Recognition Feature Extraction Algorithm And Soundprint Attendance System Realization

Speaker recognition technology belongs to biometric authentication technology, which gets Physiological and behavioral characteristics of human speech parameters from speech waveform to identify speaker identity automatically. With its unique convenience, economy, and remote applications and other advantages, it has been used in the Internet and telecommunications, military and national defense, public security and justice, finance, speech retrieval and other civil security certification fields widely. Speaker recognition becomes an important and popular authentication and identification methods in daily life and work,and its theoretical research and product development attaches a great attention.Characteristic extraction is one of the most important technology of speaker recognition technology,which is the key technology studied by the numerous domestic and international scholars.The practical application environment for speaker recognition puts forward higher request.It is extracting more representable,robuster and better speech feature parameters for speaker recognition that becomes the important problem to be solved.This paper based on the speaker recognition feature extraction algorithms. Above all,this paper puts forward a new characteristic parameters named NewMFCC,which is based on traditional MFCC feature.NewMFCC is better than MFCC in accuracy rate from experiments. Secondly, this paper studies the secondary feature extraction of NewMFCC to lift accuracy rate and gets the result that the combinational feature NewMFCC_LPCC combining16NewMFCC and12LPCC is the better characteristic parameter. At last,based on the above theoretical study,this paper achieve a speaker recognition application development named voiceprint attendance system, to make up for the shortcomings of the traditional punch card system.