Speaker Recognition without Feature Extraction Process

概要

論文の詳細を見る
By employing the dual Penalized Logistic Regression Machine (dPLRM), this paper explores a speaker identification method which does not require feature extraction process depending on a prior knowledge. The induction machine can discover implicitly speaker characteristics relevant to discrimination only from a set of training data by the mechanism of the kernel regression. 0ur text-independent speaker identification experiments with training data uttered by 10 male speakers in three different sessions show that the proposed method is competitive with the conventional Gaussian mixture model (GMM)-based method with 26-dimensional Mel-frequency cepstrum (MFCC) feature even though our method handle directly coarse data of 256-dimensional log-power spectrum. It is also shown that our method outperforms the GMM-based method especially as the amount of training data becomes smaller.
2004-12-20