Definition of Speaker Verification. Another basic task in speaker recognition is speaker verification. The goal in solving this task is to determine whether two speech utterances belong to the same person (see figure 1.2). Thus, it results in one of the following two statements: ∙ true – so both speeches come from the same speaker, or ∙ false – the tested speech does not belong to the person in the reference sample. ” In some sources, for obvious reasons, the term detection“ is also used for verification. ” ” Solving this problem is also common in practical life. This task must also be addressed by, for example, security systems. The common situation is, that the person who wants to use the system do the enrollment“ by submiting his/her referential speech model (e.g. by reading his/her ID card), and then later for gaining access, he/she provides a test sample of his/her speech. Such sample is then compared with the referential model (see [5]). Also, in the area of forensic applications, it is easier to identify a speaker from a smaller group of potential suspects than identifying one of many. This is possible because mathematically this task matches the 1 to 1 “ model. From this it can be concluded that the success rate does not depend on the number of reference speakers. This thesis, in its other parts, especially in the Chapter 4, will deal mainly with the task of speaker verification.
