Abstract
Speaker verification and identification systems most often employ HMMs and GMMs as recognition engines. This paper
describes an algorithm for the optimal selection of the feature
space, suitable for these engines. In verification systems, each
speaker (target) is assigned an “individual” optimal feature space in which he/she is best discriminated against impostors. Several feature selection procedures were tested for the selection process. A Recognition Related Criterion (RRC), correlated with the recognition rate, was developed and evaluated. The algorithm was evaluated on a text-dependent database. A significant improvement (over the “standard” MFCC space) in verification results was demonstrated with the selected individual feature space. An EER of 0.7% was achieved when the feature set was the “almost standard” Mel Frequency Cepstrum Coefficients (MFCC) space (12 MFCC
+ 12 ∆MFCC). Under the same conditions, a system based on
the selected feature space yielded an EER of only 0.48%.
describes an algorithm for the optimal selection of the feature
space, suitable for these engines. In verification systems, each
speaker (target) is assigned an “individual” optimal feature space in which he/she is best discriminated against impostors. Several feature selection procedures were tested for the selection process. A Recognition Related Criterion (RRC), correlated with the recognition rate, was developed and evaluated. The algorithm was evaluated on a text-dependent database. A significant improvement (over the “standard” MFCC space) in verification results was demonstrated with the selected individual feature space. An EER of 0.7% was achieved when the feature set was the “almost standard” Mel Frequency Cepstrum Coefficients (MFCC) space (12 MFCC
+ 12 ∆MFCC). Under the same conditions, a system based on
the selected feature space yielded an EER of only 0.48%.
| Original language | English GB |
|---|---|
| State | Published - 2004 |