| Summary: | 碩士 === 中央警察大學 === 刑事警察研究所 === 87 === Now speaker recognition is on the basis of the vowels and voiced consonants that their speech signals maintain clear impulses. The major purpose of this thesis is to research the effects of voicing impulses on speaker recognition. Some findings are the following:
First, the intraspeaker's C.V. in the fundamental frequency of same speech is less than 10%. It is a nice condition of speaker recognition. If the C.V. of experiment is above 10%, speakers are not the same. Therefore the speech of speakers recognition has to be the same .
Moreover, the discernment of speech duration is not a good condition of speaker recognition.
On the other hand, the discernment of LPC cepstral coefficients based 2 impulses' frame is not good. But the LPC cepstral coefficients with 2-3 impulses in a frame can derive more information of speech than about 3-5 impulses in a frame.
Eventually, the clear impulses of speech are suitable for speaker recognition. Although some speech contain vowels or voiced consonants, they might be not suitable for speaker recognition without clear impulses.
|
|---|
