Speaker recognition with polynomial classifiers

被引:87
作者
Campbell, WM [1 ]
Assaleh, KT [1 ]
Broun, CC [1 ]
机构
[1] Motorola Labs, Tempe, AZ 85284 USA
来源
IEEE TRANSACTIONS ON SPEECH AND AUDIO PROCESSING | 2002年 / 10卷 / 04期
关键词
discriminative training; polynomial classification; speaker recognition;
D O I
10.1109/TSA.2002.1011533
中图分类号
O42 [声学];
学科分类号
070206 ; 082403 ;
摘要
Modern speaker recognition applications require high accuracy at low complexity. We propose the use of a polynomial-based classifier to achieve these objectives. This approach has several advantages. First, polynomial classifier scoring yields a system which is highly computationally scalable with the number of speakers. Second, a new training algorithm is proposed which is discriminative, handles large data sets, and has low memory usage. Third, the output of the polynomial classifier is easily incorporated into a statistical framework allowing it to be combined with other techniques such as hidden Markov models. Results are given for the application of the new methods to the YOHO speaker recognition database.
引用
收藏
页码:205 / 212
页数:8
相关论文
共 24 条
  • [1] [Anonymous], P INT C SPOK LANG PR
  • [2] [Anonymous], 1996, PATTERN CLASSIFICATI
  • [3] [Anonymous], CONNECTIONIST SPEECH
  • [4] Bishop C. M., 1995, NEURAL NETWORKS PATT
  • [5] Speaker recognition: A tutorial
    Campbell, JP
    [J]. PROCEEDINGS OF THE IEEE, 1997, 85 (09) : 1437 - 1462
  • [6] CAMPBELL JP, 1995, INT CONF ACOUST SPEE, P341, DOI 10.1109/ICASSP.1995.479543
  • [7] Campbell WM, 2000, INT CONF ACOUST SPEE, P1185
  • [8] CAMPBELL WM, 1999, P 5 INT S SIGN PROCE, P619
  • [9] CAMPBELL WM, P 17 INT C MACH LEAR, V200, P1015
  • [10] CAREY MJ, 1991, INT CONF ACOUST SPEE, P397, DOI 10.1109/ICASSP.1991.150360