![]() | Only 14 pages are availabe for public view |
Abstract The subject of automatic speaker veriflcation has found great interest in tbe present time. Tbis tbesis is a contribut- r o n to tbe subject. A t no r-o ug n review of tbe publlsbed worK 1S presented. Tbe nature of speech signal and the different parameters characterizing it are studied. The linear prediction coding of speech signal is used for speaker verification. Usually the wbole speecb signal with all its spectrum components is used for speaker verification . In tbe present work a comparative study is made between the traditional metbod and a new proposed approach using tbe prefiltered signal version baving only tbe fourtb and fiftb formant components. Measurements were conducted for dl fferent persons at different sessions and tbe required programs for the recorded speecb signals were establisbed and speaKer verificat- ion was done. The results are encouraging witb I respect to precision. |