Speaker Verification Based Fractal Geometry

Abstract

The research shows that the fractal structure of the speech signal exhibits fractal characteristics. The encouraging analysis results indicated that the fractal dimension have good discrimination capabilities for speech signal, where it gave a speaker recognition percentage about 70% (In the field of research for a single parameter this percentage be valued), and these capabilities are strongly depends on the setting situations of the recording process. The pitch period is the another used parameter supported to the fractal dimension to strengthen the verification decision. Where the pitch period estimated by a new suggested simple method to gave a speaker verification percentage about 90%. The coalition work of the two parameters (fractal dimension and pitch period) gave a speaker verification percentage of about 85%.