Method for recognizing voice print
A voiceprint recognition and recognition method technology, applied in character and pattern recognition, analog computer, analog and hybrid computing, etc., can solve problems such as application difficulties and inability to adapt to requirements, and achieve the effect of low error rate
Patent Information
- Authority / Receiving Office
- CN · China
- Current Assignee / Owner
- Publication Date
- 2003-10-08
- Estimated Expiration
- Not applicable · inactive patent
Smart Images
Figure 1 Figure 2 Figure 3
Abstract
Description
technical field
[0001] The invention belongs to the technical field of computers and information services, in particular to a method for identifying and confirming identities through human voiceprint information. Background technique
[0002] Voiceprint Recognition (Speaker Recognition), that is, Speaker Recognition, is to identify who is speaking a certain voice based on the biological characteristics of the speaker contained in the human voice, which is the so-called "knowing people by hearing the sound". ". Voiceprint recognition can be used in almost all security protection fields and personalization applications that require identification or confirmation. For example: (1) voiceprint identification: criminal investigation, criminal tracking, national defense monitoring, personalized applications, etc.; (2) voiceprint confirmation: securities transactions, bank transactions, public security forensics, voice-activated locks for personal computers and cars, ID cards, cred...
Examples
Embodiment Construction
[0041] A voiceprint recognition method proposed by the present invention is described in detail as follows in conjunction with the accompanying drawings and embodiments, as well as the application:
[0042] Embodiments of the voiceprint recognition method of the present invention, such as image 3 (a)- image 3 As shown in (c), two types of voiceprint recognition methods including the model training method and voiceprint identification and voiceprint confirmation are described as follows in conjunction with the accompanying drawings:
[0043] The model training method of this embodiment is as follows: image 3 Shown in (a), its specific steps include:
[0044] 1) Take the voice data of a speaker, analyze its original voice waveform data, and throw away each silent segment therein;
[0045] 2) Take the frame width of 32 milliseconds and half of the frame width as the frame shift, extract 16-dimensional linear predictive cepstral parameters (LPCC) for each frame, and calculat...