|Table of Contents|

Speech Pitch Detection Algorithm Based on Modified Cepstrum Model


Research Field:
Publishing date:


Speech Pitch Detection Algorithm Based on Modified Cepstrum Model
ZENG Yu-min12WU Zhen-yang1
1.School of Information Science and Engineering,Southeast University,Nanjing 210096,China;2.School of Physics and Technology,Nanjing Normal University,Nanjing 210097,China
pitch cepstrum linear pred ictive cod ing pred ictive pesidua l
An improved speech pitch detect ion a lgorithm based on mod if ied cepstrum model is proposed. In the proposed algorithm, a ten-order LPC ( linear predictive coding ) analysis is performed on a segmented speech, and the segmented speech is f iltered by the inverse filter to g ive theLPC pred ict ive residua.l The cepstrum of the pred ict ive residual is calcu lated w ith the simp le method of the h igh frequency spectral components of DFT being set to zero. The pitch period o f the vo iced speech is extracted from the cepstrum of pred ict ive residua.l The simu lated p itch detection results show that the pitch extract ion error of the proposed a lgorithm is significantly low er than that of the conventional cepstrum based a lgorithm bo th for clean speech and d ifferent no isy speech. The performance o f the proposed algorithm is alsomuch better than that of the average magn itude difference funct ion based p itch detection algorithm and slightly better than that of the basic autocorrelat ion function based algorithm.


[ 1] Rab ine r L, ChengM, RosenbergA, et a .l A com pa rative perform ance study o f several pitch detection a lgorithm s [ J] . IEEE Trans on Acoustics, Speech, and S ignal Processing, 1976, 24 ( 5): 399- 417.
[ 2] No llA M. Cepstrum pitch dete rm ina tion [ J] . Journa l o f the Acoustic Soc ie ty of Ame rica, 1967, 41 ( 2 ): 293- 309.
[ 3] Kadambe S, Boudreaux-Barte ls G. App lication o f the w avelet transform for p itch detection of speech signals [ J] . IEEE Trans on Inform ation Theo ry, 1992, 38 ( 2): 917- 924.
[ 4] H uang D, L inW, Raha rdja S. Speech p itch de tection in no isy env ironm ent us ing m ult-i rate adap tive lo ssless FIR filters [ A ]. Proceed ing s of the Inte rna tiona l Symposium on C ircu its and Systems 2004 ( ISCAS . 04) [ C ]. [ S. .l ]: IEEE, 2004. Ⅲ - 429- 432.
[ 5] Xu X, M iyanag aY. A robust pitch de tection in no isy speech w ith band-pass filtering on m odulation spec tra [ A]. Proceed ings of In ternational Sym po sium on Communications and Inform ation T echno logy ( ISC IT2005) [ C] . [ S. .l ]: IEEE, 2005. 266- 269.
[ 6] Ahm adi S, Span ias A. Cepstrum-based p itch de tection using a new statistica l V /UV c lassifica tion a lgor ithm [ J]. IEEE T rans on Speech and Audio Processing, 1999, 7 ( 3): 333- 338.
[ 7] H odgson L, Je rniganM, W ills B. Nonlinearmu ltiplicative cepstra l ana lysis for pitch extraction in speech [ A] . Proceed ings o f Internationa l Con ference on Acoustics, Speech, and S ignal Pro cessing ( ICASSP-90) [ C] . A lbuquerque, USA: IEEE, 1990. 257- 260.
[ 8] Nadeu C, Pascua l J, H ernando J. P itch de term ina tion using the cepstrum of the one- sided autoco rre lation sequence [ A]. Pro ceedings o f Interna tiona l Con ference on Acoustics, Speech, and S ignal Process ing ( ICASSP- 91) [ C ]. Toronto, Canada: IEEE, 1991. 3 677- 3 680.
[ 9] Andrew sM, Picone J, Deg roat R. Robust p itch de term ina tion v ia SVD based cepstra l methods [ A ]. Proceedings o f International Conference on Acoustics, Speech, and S igna l Processing ( ICASSP-90) [ C]. A-l buquerque, USA: IEEE, 1990. 253- 256.
[ 10] 杨行峻, 迟惠生, 李爱军, 等. 语音信号数字处理 [M ]. 北京: 电子工业出版社, 2000. 54- 62.
[ 11] Ve rhelstW, Steenhaut O. A new m ode l for the shorttim e comp lex cepstrum of vo iced speech [ J] . IEEE T rans on Acoustics, Speech, and S igna l Processing, 1986, 34 ( 1): 43- 51.
[ 12] SPIB. No iseX92 no ise database [ EB /OL]. H ttp: / / spib. rice. edu /spib / select_noise. htm ,l 2002- 11- 15.


Last Update: 2007-08-30