о뷢չ
JOURNAL OF COMPUTER RESEARCH AND DEVELOPMENT
1999Vol.36No.6P.712-716



ڶ뱾ʸķ޶ı
߱Ϸ
ӿġҦѫ
ժҪͳûߵһʱڵΪѵл뱾ѵķʶϵͳȶ.ΪӦʱԣû߲ͬʱڵѵߵģͣÿ߾ж뱾.Щ뱾ǲ𽥼СʶʵŻ̵õ.Ϊ˲ͬŵϵͳʶܵӰ죬иһŵ.ͬʱһ֡ܵһصʵȡ, ʸ뱾Լ44%ʶ.ЩϵͳʶٶȺ³.бȽPLPLPC׷л߱ϵʶ.
ؼʣ߱ϣ뱾ʸŵ
ţTP391
AN APPROACH BASED ON MULTIPLE VECTOR QUANTIZATION
FOR ON-LINE TEXT-INDEPENDENT SPEAKER IDENTIFICATION
MA Ji-YongGAO WenYAO Hong-Xun
(Department of Computer Science, Harbin Institute of Technology, Harbin 150001)
AbstractThe traditional approach for training speaker codebooks only uses one session training speech samples, but the recognition system based on this approach is usually not robust. To adapt to the intraspeaker variations, the paper here introduces an approach for training speaker codebooks using multiple session training speech samples,with every speaker having multiple codebooks. These codebooks are trained based on the minimum recognition error rate.To compensate for the variations arising from transmission conditions, an approach to compensation of the variation presented. To speed up recognition speed, an on-line feature extraction method for voiced sounds and two level vector quantization and codebook index strategy are used. These techniques increase the robustness of the speech feature and speed up the training and identification procedure greatly. Finally, the identification results of comparison using the perceptually based linear predictive(PLP) analysis and the LPC cepstrum analysis are given.
Key wordson-line text-independent speaker identification, multiple codebooks quantization, transmission compensation
1
ͨý¼Ϣлʶ𣬿˵Ǵӱͳ绰ĳ־Ϳʼ.ǽ绰¼Ϥ˵ȽϣӶȷߵ.÷ȱǣʶΧСԲϤʶ׼÷ԵŵǶԺϤкܺõ³.ʶǣԤüӻߵȡȻԱ߽ʶɻߵʶ.йػʶоĽչף1ݡ2б.ף3ȡӦȨϵĿ³.ΪȡĶ̬ף4ݡ5иʸԻعģͣף5жף6еõʸԻعģͿȡĶ̬Ľ۱ʾ.ף6öͳлʶķ.ף7ݡ8˻ʶ³.ף9и˻ʶĸ˹ģ.ף1ָ˹ģͺʸ10ݼȫ·ģ͵ʶ൱.ף12иĻʶ𷽷.ʸŵѵʶٶȿ.ʶ߱Ϻͻȷ棬Ŀǰֻȷϵоб13ݣ߱ϵоδ.߱Ϊҡ๦ܸ֪оĿ, ˻ܽӿҪӦüֵ.Ŀǰ߱ҪҪҪ⣺
(1 δغαװ.߱϶Ա߲ʱ缭ﷸͥڽѵͲʱ߿ܲ빤Աʹϵͳܱϳ.
(2 ŵԿܲͬ, ̫, ʹϵͳʶʴ½, ΪԲݵʱ,ܸöŵԺѵʱŵԲͬ.
(3 ϵͳӦʱϵͳʶ֮ì.߼еĻĿ n ܴʱҪn֪ߵֽбȽϣռ˺ܶĻʱ.ԲҪʵʱԵӦóϣ缭ﷸͥһⲢҪ˻ܽӿӦãһ÷ǳҪ.
3⣬ĸһ޶ı߻߱ϵͳƷ.
2ȡ
ͨ绰ƵʷΧԼ3003400Hz֮䣬ݲƵΪ8000HzʱԱƵ׻.ϵͳΪ8λmu-lawѹȻļmu-lawѹΪ16λPCM.
źſɽƵΪһʱƽ̣еĻߵĸϢźеģʹôźзߵĸ൱ӣߵĸҪźŵĵƵһΪ401000HzķΧ,һ棬ӵ֪˶20005000HzرУ˵ߵĸϢҪ405000HzƵʷΧ.
Ҫذǿĸߵͣɫĺû.ҪȡƵʣƵԽߣԽߣѹɳɷйأɫǷӳһԣҪɸɷиִƵʾģȻͬߵĻͬÿߵʲͬ䷺ɷҲͬǸݲͬƵʳɷּǿֲָͬɫӶֲͬߵ.ͨߵ񶯷ģнǿ׼ԣӰ죬нϺõ³ԣΪʶҪź.෴ܻӰȡ³Բ.ĿǰȡķУ˲ƵסLPCס LPCϵLPCϵLPCϵLPCϵMel Frequency Cepstrum ϵ( MFCC )֪Ԥ( the perceptually based linear predictive ) PLP ϵ4ݵ.
ߵѵΪźžԤ֮(ȡPLPϵźŲԤ),˵ȷ֡,ȡ.ϵͳ LPCϵPLPϵȡ֡Ϊ32mm, ֡Ϊ16mmLPCϵĽȡΪ16PLPϵĽȡΪ12вƵȨ.

ʽУCiRCiTֱȫģ͵ĲοϵͲԵĵϵpǽsƵ׷ֵǿϵȡΪ0.6 .ֲȷӳȫģƵ׷ֵĲ.ʵ׼ȨȶԻʶнϸߵ.һ棬ֲȶŵͬʱҲ.
3뱾ʸ
ʸڻʶǽߵźŵķƽ̷ָɸƽ̣ÿһƽȹ̶Ӧһصƽȶ.ĸ38ĸ22ĸ.뱾ĴСӦĿ.ϵͳ뱾СȡΪ64.źŷȳĶ̬ΧʱֹźŵķֵСʱԽϴ.Ӧѵʶʱ⣬Ϊ뻰Ͳ̫ܷ󣬶뻰Ͳ̫ԶʱȽͣԽϴ.⣬뻰ͲϽʱοΪ沨ʱʱֵ𽥼Сֵ˥뻰뻰Ͳɷ.̼˷ӦķԣͬͲͬɵӰ죬ʹźűֳһַƽ.ΪӦʱԣû߲ͬʱڵѵߵģͣÿ߾ж뱾.Щ뱾ǲ𽥼СʶʵŻ̵õ.ѵÿһ뱾Ϊ30ңÿһߵ뱾ĿΪ5ң뱾СΪ64.
4ŵʶ𷽷
ĵΪY={y1,y2,,yT},T֡, ٶѵͲʱŵͬɵƫ룬ϵ
(1)
ʽУX={x1,x2,,xT}ѵŵµУhʶʱŵѵʱŵĲ.ٶ߼nɣٶnߵĵk뱾ĸܶΪpnk(x)N(nk,2nk)k=1,2,,M.M뱾ĴС.ƦhȻΪ
(2)
ʽSX={x1,x2,,xT}ڻģе״̬S={1,2,,M}, ģ͵Ĳ.
(3)
ʽ(3)ڦhȡƫʹΪ㣬ɵ
(4)
ʽУyjk,hnkֱyj,hnĵk.
̿ɽ
(5)
snjǺͦhnйص, ϵҪ.£óʼֵn=0ɼX={x1,x2,,xT}ģе״̬Sʽ㦤nһʽ(1)X={x1,x2,,xT}УȻģо״̬ʽе㣬տɵæn.ʶʱX={x1,x2,,xT}ÿģеȻʣѡӦȻʵģΪʶĻ߶Ӧģͣϵ

5߻߱
߻߱ϵͳҪٶȿ,Ҫϵͳнϸߵʶ.ΪҪ󣬱ϵͳȡ.һ֡ܵһص.Ϊźžжʱƽһص֡仯кǿ.Ȼȡغܿ죬ȡ̫ô߱ϵƥʱҲҪ.Խ,ַһ㷽ʶʻͬ.ȡʵʱҪ󣬲ַʹʶٶȼӿ3000౶.


ͼ1ʸ
ΪʶٶȣΪͼ1ʾ.1뱾СΪ16 뱾СΪ20õ1뱾뱾СΪ64뱾뱾СΪ20뱾ʶʱʸȾ1ҵӦ֣Ȼ뱾е2.ʶʱļ44%.
߻߱ϷΪһǴеĻߵĲļУһֱӴӻͲɼлʶ.ļʶʱÿ(һΪ35)ϵͳһεʶ.Ϊʵһܣϵͳһ黺洢μ.ÿһʱ.ΪʵֱӴӻͲʶϵͳͬźŸٹܣϵͳԵÿһʱʶ.ϵͳһ黺洢μ.ÿһʱݣԭͼͼ 2 ʾ.ȡʱƬ8ƬÿһʱƬִ󣬼ۻЧεʱ䳤Ƿڸʱ䣬ڣʶ𣬲ȻִһʱƬִе8ʱƬۻЧεʱ䳤ǲڸʱ䣬ΪЧʶ.


ͼ2ʶִй
6Խ
6.1ʵ鱳
ԭSUN Sparc II վϿһı޹صĻʶϵͳ(TISI)ϵͳṹͼ 3ʾ߼65ɣÿߵѵ 150 룬ĳȷֱȡΪ 5롢3 1룬ʵ黷һ㣬ʵͲԼʱΪ6.
6.2ȶʶʵӰ
ĳС1ʱϵͳʶʺܵͣԲȷֱΪ 531ʵڱ1 , ڲԺѵĻŵԻͬ,,ͬĻ,뻰ͲľͬһķֵĲΧͬ.


ͼ3ʸĻʶϵͳṹ
1 ʶ

ʱ(s)ļ(%)ӻͲ(%)
511
356
14041

6.3ŵԼʶʵӰ
ʱŵԼѵʱŵԼԲͬʱ,ϵͳʶ87%,ͨ뱾ѵԸʶϵͳʶ.
7ᡡ
˶뱾ʸķ޶ı߱ϷͨоõҪۣ
(1PLPLPC׷õǻʶõPLPʱ临Խϵͣʺ߻ʶ.
(2һ֡һʶ𷽷ؼʶʱ䣬ʵ÷ʶû½.
(3ŵͶ뱾ʸ˸ϵͳ³.
(4ͨʸ뱾,ʶʱļ44%,ʶӰ.
(565ɵĻ߼,ڲԺѵĻŵԻͬʱ,ϵͳʶʿɴ99%.
ۿɼ,ĿǰӰ컰߱ʶʵҪŵ.ˣѰȽĻ߸µŵһǻʶоҪ.
Ŀõҡƻ(Ŀ863-306-03-03)͹ȻѧĿ.
߼飺ӿУ19636ʿҪоΪʶͼʶ.
ģУ19564ڣʿʦйѧԺ㼼оҡƻܼר鳤(ϯר)оԺԺҪоΪ˹ܺͶý弼ڼӾģʽʶͼýѹ.
ҦѫУ196511ʦҪоΪ๦ܸ֪.
ߵλӿҵѧѧϵ150001)
ģҵѧѧϵ150001)
Ҧѫҵѧѧϵ150001)
οף
1Rosenberg A E. Soong F K. Recent research in automatic speaker recognition. In: Furui S Sondhi M M eds. Advances in Speech Signal Processing. New York: Marcel Dekker, 1992. 701738
2Furui S. Recent advances in speaker recognition. In: Bigun J, Chollet G eds. The First Int'l Conf on Audio- and Video-Based Biometric Person Authentication. Switzerland: Springer, 1997. 237252
3Khaled T A. Robust cepstral features for speaker identification. In: Lever K ed. Proc of 1994 IEEE Int'l Conf on Acoustics, Speech, and Signal Processing. South Australia, Adelaide, 1994. 129132
4Floch J L, Montacie C, Caraty M J. Speaker recognition experiments on NTIMIT database. In: Proc of 4th European Conf on Speech Communication and Technology, Vol 1. Madrid, 1995. 379382
5Magrin-Chagnolleau I, Wilke J, Bimbot F. A further investigation on AR-vector models for text-independent speaker identification. In: Vijay K M ed. Proc of 1996 IEEE Int'l Conf on Acoustics, Speech, and Signal Processing, Atalanta, USA, 1996. 300304
6Bimbot F, Margrin-Chagnolleau I, Mathan L. Second-order statistical measures for text-independent speaker identification. Speech Communication, 1995, 17(1-2): 9099
7Openshaw J P, Sun Z P, Mason J S. A comparison of composite features under degraded speech in speaker recognition. In: Sullivan B J ed. Proc of 1993 IEEE Int'l Conf on Acoustics, Speech, and Signal Processing. Minnesota, USA, 1993. 371374
8Kao Yu-Huang, Baras J S, Rajasekarn P K. Robustness study of free-text speaker identification and verification. In: Barry J S ed. Proc of 1993 IEEE Int'l Conf on Acoustics, Speech and Signal Processing. Minnesota, USA. 1993. 379382
9Reynolds D A. Speaker identification and verification using Gaussian mixture speaker models. Speech Communication, 1995, 17(1): 91108
10Soong F K. A vector quantization approach to speaker recognition. AT & T Technical Journal, 1987, 66(1): 1426
11Savic M, Gupta S K.Variable parameter speaker verification system based on hidden Markov modeling. In: Lonnie L ed. Proc of 1990 IEEE Int'l Conf on Acoustics, Speech, and Signal Processing. Albuquerque, New Mexico, USA, 1990. 281284
12Rudasi L, Zahorian S A.Text-independent talker identification with neural networks. In: Proc of 1991 IEEE Int'l Conf on Acoustics, Speech, and Signal Processing. Toronto, Canada, 1991. 409408
13Attili J B.On the development of a real-time text-independent speaker verification systemPh D dissertation. Rensseler Polytechnic Institite, Rensseler1987
ոڣ1997-09-11
޸ڣ1998-05-08
