The speaker recognition can be divided into two types of applications:
Speaker Verification and Speaker Identification.
in Speaker Verification , the person “looks” through his identification of some sort (name, password, pin, another type of code …), and l ‘application checks if the voice of those who are talking about the same as that of his voiceprint stored in the system. If the speaker’s voice and the footprint are similar, that is, if the similarity value exceeds a certain threshold, the verification is successful and the person is authenticated.
In an application of Speaker Identification , however, the person speaking “does not show up,” and you have to measure the degree of similarity of his voice with respect to a set of speech patterns available. If all potential users of the system have already been cataloged (“closed-set” ), we assume that the voice of the person must still correspond to one of the available options. Otherwise, if there is the possibility that a user has not been previously cataloged ( “open-set” ), the test result may be that the voice does not match any of those notes.
For both applications Speaker Verification or Speaker Identification is required then a training phase, in which they are prepared speech patterns of all the people whose voice will be controlled. Depending on the context and the type of application, the system can be configured to obtain a more or less “restrictive”.
Some examples of applications of Speaker Recognition:
- authentication for access to computer applications : the verification of the PIN or password is accompanied by that of the voice, to achieve a higher level of security against fraud or unauthorized access. Or you can use voice verification as the main mechanism for authentication, in a context in which the user can not use their hands (eg, surgery).
- customer identification in call centers : the identification of the client can be used by the system for addressing and prioritizing calls. The control of the customer can also serve as a measure of security for access to confidential information.
- control access to restricted areas : the voice control can be coupled to a badge reader or another biometric identification
(eg, fingerprint, face …).
- research in multimedia archives in a collection of audio / video files, you can automatically identify people with known voice prints
( audio / video indexing and retrieval ).