Compressive Sensing-Based Speech Enhancement

Jia Ching Wang, Yuan Shan Lee, Chang Hong Lin, Shu Fan Wang, Chih Hao Shih, Chung-Hsien Wu

Research output: Contribution to journalArticle

13 Citations (Scopus)

Abstract

This study proposes a speech enhancement method based on compressive sensing. The main procedures involved in the proposed method are performed in the frequency domain. First, an overcomplete dictionary is constructed from the trained speech frames. The atoms of this redundant dictionary are spectrum vectors that are trained by the K-SVD algorithm to ensure the sparsity of the dictionary. For a noisy speech spectrum, formant detection and a quasi-SNR criterion are first utilized to determine whether a frequency bin in the spectrogram is reliable, and a corresponding mask is designed. The mask-extracted reliable components in a speech spectrum are regarded as partial observations and a measurement matrix is constructed. The problem can therefore be treated as a compressive sensing problem. The K atoms of a K -sparsity speech spectrum are found using an orthogonal matching pursuit algorithm. Because the K atoms form the speech signal subspace, the removal of the noise projected onto these K atoms is achieved by multiplying the noisy spectrum with the optimized gain that corresponds to each selected atom. The proposed method is experimentally compared with the baseline methods and demonstrates its superiority.

Original languageEnglish
Pages (from-to)2122-2131
Number of pages10
JournalIEEE/ACM Transactions on Audio Speech and Language Processing
Volume24
Issue number11
DOIs
Publication statusPublished - 2016 Nov 1

Fingerprint

Speech Enhancement
Speech enhancement
Compressive Sensing
Atoms
Glossaries
dictionaries
augmentation
dictionary
Sparsity
atoms
Mask
Masks
Partial Observation
Spectrogram
masks
Matching Pursuit
Speech Signal
Bins
Singular value decomposition
spectrograms

All Science Journal Classification (ASJC) codes

  • Signal Processing
  • Media Technology
  • Instrumentation
  • Acoustics and Ultrasonics
  • Linguistics and Language
  • Electrical and Electronic Engineering
  • Speech and Hearing

Cite this

Wang, Jia Ching ; Lee, Yuan Shan ; Lin, Chang Hong ; Wang, Shu Fan ; Shih, Chih Hao ; Wu, Chung-Hsien. / Compressive Sensing-Based Speech Enhancement. In: IEEE/ACM Transactions on Audio Speech and Language Processing. 2016 ; Vol. 24, No. 11. pp. 2122-2131.
@article{f9826d835a7e4f05ab699682b1cc4bfa,
title = "Compressive Sensing-Based Speech Enhancement",
abstract = "This study proposes a speech enhancement method based on compressive sensing. The main procedures involved in the proposed method are performed in the frequency domain. First, an overcomplete dictionary is constructed from the trained speech frames. The atoms of this redundant dictionary are spectrum vectors that are trained by the K-SVD algorithm to ensure the sparsity of the dictionary. For a noisy speech spectrum, formant detection and a quasi-SNR criterion are first utilized to determine whether a frequency bin in the spectrogram is reliable, and a corresponding mask is designed. The mask-extracted reliable components in a speech spectrum are regarded as partial observations and a measurement matrix is constructed. The problem can therefore be treated as a compressive sensing problem. The K atoms of a K -sparsity speech spectrum are found using an orthogonal matching pursuit algorithm. Because the K atoms form the speech signal subspace, the removal of the noise projected onto these K atoms is achieved by multiplying the noisy spectrum with the optimized gain that corresponds to each selected atom. The proposed method is experimentally compared with the baseline methods and demonstrates its superiority.",
author = "Wang, {Jia Ching} and Lee, {Yuan Shan} and Lin, {Chang Hong} and Wang, {Shu Fan} and Shih, {Chih Hao} and Chung-Hsien Wu",
year = "2016",
month = "11",
day = "1",
doi = "10.1109/TASLP.2016.2598306",
language = "English",
volume = "24",
pages = "2122--2131",
journal = "IEEE/ACM Transactions on Speech and Language Processing",
issn = "2329-9290",
publisher = "IEEE Advancing Technology for Humanity",
number = "11",

}

Compressive Sensing-Based Speech Enhancement. / Wang, Jia Ching; Lee, Yuan Shan; Lin, Chang Hong; Wang, Shu Fan; Shih, Chih Hao; Wu, Chung-Hsien.

In: IEEE/ACM Transactions on Audio Speech and Language Processing, Vol. 24, No. 11, 01.11.2016, p. 2122-2131.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Compressive Sensing-Based Speech Enhancement

AU - Wang, Jia Ching

AU - Lee, Yuan Shan

AU - Lin, Chang Hong

AU - Wang, Shu Fan

AU - Shih, Chih Hao

AU - Wu, Chung-Hsien

PY - 2016/11/1

Y1 - 2016/11/1

N2 - This study proposes a speech enhancement method based on compressive sensing. The main procedures involved in the proposed method are performed in the frequency domain. First, an overcomplete dictionary is constructed from the trained speech frames. The atoms of this redundant dictionary are spectrum vectors that are trained by the K-SVD algorithm to ensure the sparsity of the dictionary. For a noisy speech spectrum, formant detection and a quasi-SNR criterion are first utilized to determine whether a frequency bin in the spectrogram is reliable, and a corresponding mask is designed. The mask-extracted reliable components in a speech spectrum are regarded as partial observations and a measurement matrix is constructed. The problem can therefore be treated as a compressive sensing problem. The K atoms of a K -sparsity speech spectrum are found using an orthogonal matching pursuit algorithm. Because the K atoms form the speech signal subspace, the removal of the noise projected onto these K atoms is achieved by multiplying the noisy spectrum with the optimized gain that corresponds to each selected atom. The proposed method is experimentally compared with the baseline methods and demonstrates its superiority.

AB - This study proposes a speech enhancement method based on compressive sensing. The main procedures involved in the proposed method are performed in the frequency domain. First, an overcomplete dictionary is constructed from the trained speech frames. The atoms of this redundant dictionary are spectrum vectors that are trained by the K-SVD algorithm to ensure the sparsity of the dictionary. For a noisy speech spectrum, formant detection and a quasi-SNR criterion are first utilized to determine whether a frequency bin in the spectrogram is reliable, and a corresponding mask is designed. The mask-extracted reliable components in a speech spectrum are regarded as partial observations and a measurement matrix is constructed. The problem can therefore be treated as a compressive sensing problem. The K atoms of a K -sparsity speech spectrum are found using an orthogonal matching pursuit algorithm. Because the K atoms form the speech signal subspace, the removal of the noise projected onto these K atoms is achieved by multiplying the noisy spectrum with the optimized gain that corresponds to each selected atom. The proposed method is experimentally compared with the baseline methods and demonstrates its superiority.

UR - http://www.scopus.com/inward/record.url?scp=84990925734&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84990925734&partnerID=8YFLogxK

U2 - 10.1109/TASLP.2016.2598306

DO - 10.1109/TASLP.2016.2598306

M3 - Article

VL - 24

SP - 2122

EP - 2131

JO - IEEE/ACM Transactions on Speech and Language Processing

JF - IEEE/ACM Transactions on Speech and Language Processing

SN - 2329-9290

IS - 11

ER -