Note: Your custom schedule will not be saved unless you create a new account or login to an existing account.
| Paper: | SP-P3.12 |
| Session: | Topics in Speaker and Langauge Recognition |
| Time: | Tuesday, May 18, 15:30 - 17:30 |
| Presentation: |
Poster |
| Topic: |
Speech Processing: Speaker Recognition |
| Title: |
A PITCH SYNCHRONOUS FEATURE EXTRACTION METHOD FOR SPEAKER RECOGNITION |
| Authors: |
Samuel Kim; Yonsei University | | |
| | Thomas Eriksson; Chalmers University of Technology | | |
| | Hong-Goo Kang; Yonsei University | | |
| | Dae Hee Youn; Yonsei University | | |
| Abstract: |
This paper presents a novel feature extraction method to improvethe performance of speaker identification systems. The proposedfeature has a form of a typical conventional feature, melfrequency ceptral coefficients (MFCC), but a flexiblesegmentation to reduce spectral mismatch between training andtesting processes. Specifically, the length and shift size of theanalysis frame are determined by a pitch synchronous method, pitchsynchronous MFCC (PSMFCC). To verify the performance of the newfeature, we measure the cepstral distortion between training andtesting and also perform closed set speaker identification tests.With text-independent and text-dependent experiments, theproposed algorithm provides 44.3% and 26.7% relativeimprovement respectively. |
| |
| Back | |