Regularized linear prediction of speech

L. Anders Ekman; W. Bastiaan Kleijn; Manohar N. Murthi

doi:10.1109/TASL.2007.909448

Regularized linear prediction of speech

L. Anders Ekman, W. Bastiaan Kleijn, Manohar N. Murthi

Electrical and Computer Engineering

Research output: Contribution to journal › Article › peer-review

32 Scopus citations

Abstract

All-pole spectral envelope estimates based on linear prediction (LP) for speech signals often exhibit unnaturally sharp peaks, especially for high-pitch speakers. In this paper, regularization is used to penalize rapid changes in the spectral envelope, which improves the spectral envelope estimate. Based on extensive experimental evidence, we conclude that regularized linear prediction outperforms bandwidth-expanded linear prediction. The regularization approach gives lower spectral distortion on average, and fewer outliers, while maintaining a very low computational complexity.

Original language	English (US)
Article number	4378273
Pages (from-to)	65-73
Number of pages	9
Journal	IEEE Transactions on Audio, Speech and Language Processing
Volume	16
Issue number	1
DOIs	https://doi.org/10.1109/TASL.2007.909448
State	Published - Jan 2008

Keywords

Bandwidth expansion
Envelope estimation
Linear prediction (LP)
Regularization

ASJC Scopus subject areas

Electrical and Electronic Engineering
Acoustics and Ultrasonics

Access to Document

10.1109/TASL.2007.909448

Cite this

@article{98bf48d5a1cf44788b641fbbc164948f,

title = "Regularized linear prediction of speech",

abstract = "All-pole spectral envelope estimates based on linear prediction (LP) for speech signals often exhibit unnaturally sharp peaks, especially for high-pitch speakers. In this paper, regularization is used to penalize rapid changes in the spectral envelope, which improves the spectral envelope estimate. Based on extensive experimental evidence, we conclude that regularized linear prediction outperforms bandwidth-expanded linear prediction. The regularization approach gives lower spectral distortion on average, and fewer outliers, while maintaining a very low computational complexity.",

keywords = "Bandwidth expansion, Envelope estimation, Linear prediction (LP), Regularization",

author = "Ekman, {L. Anders} and Kleijn, {W. Bastiaan} and Murthi, {Manohar N.}",

note = "Funding Information: Manuscript received March 30, 2007; revised September 4, 2007. The work of L. A. Ekman and W. B. Kleijn was supported by Ericsson Research and Development. The work of M.N. Murthi was supported by the U.S. National Science Foundation under Award CCF-0347229. The associate editor coordinating the review of this manuscript and approving it for publication was Dr. Yariv Ephraim.",

year = "2008",

month = jan,

doi = "10.1109/TASL.2007.909448",

language = "English (US)",

volume = "16",

pages = "65--73",

journal = "IEEE Transactions on Speech and Audio Processing",

issn = "1558-7916",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

number = "1",

}

TY - JOUR

T1 - Regularized linear prediction of speech

AU - Ekman, L. Anders

AU - Kleijn, W. Bastiaan

AU - Murthi, Manohar N.

N1 - Funding Information: Manuscript received March 30, 2007; revised September 4, 2007. The work of L. A. Ekman and W. B. Kleijn was supported by Ericsson Research and Development. The work of M.N. Murthi was supported by the U.S. National Science Foundation under Award CCF-0347229. The associate editor coordinating the review of this manuscript and approving it for publication was Dr. Yariv Ephraim.

PY - 2008/1

Y1 - 2008/1

N2 - All-pole spectral envelope estimates based on linear prediction (LP) for speech signals often exhibit unnaturally sharp peaks, especially for high-pitch speakers. In this paper, regularization is used to penalize rapid changes in the spectral envelope, which improves the spectral envelope estimate. Based on extensive experimental evidence, we conclude that regularized linear prediction outperforms bandwidth-expanded linear prediction. The regularization approach gives lower spectral distortion on average, and fewer outliers, while maintaining a very low computational complexity.

AB - All-pole spectral envelope estimates based on linear prediction (LP) for speech signals often exhibit unnaturally sharp peaks, especially for high-pitch speakers. In this paper, regularization is used to penalize rapid changes in the spectral envelope, which improves the spectral envelope estimate. Based on extensive experimental evidence, we conclude that regularized linear prediction outperforms bandwidth-expanded linear prediction. The regularization approach gives lower spectral distortion on average, and fewer outliers, while maintaining a very low computational complexity.

KW - Bandwidth expansion

KW - Envelope estimation

KW - Linear prediction (LP)

KW - Regularization

UR - http://www.scopus.com/inward/record.url?scp=64949193677&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=64949193677&partnerID=8YFLogxK

U2 - 10.1109/TASL.2007.909448

DO - 10.1109/TASL.2007.909448

M3 - Article

AN - SCOPUS:64949193677

VL - 16

SP - 65

EP - 73

JO - IEEE Transactions on Speech and Audio Processing

JF - IEEE Transactions on Speech and Audio Processing

SN - 1558-7916

IS - 1

M1 - 4378273

ER -

Regularized linear prediction of speech

Abstract

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this