speech enhancement using excitation source information
DESCRIPTION
Speech Enhancement using Excitation Source Information. B. Yegnanarayana, S.R. Mahadeva Prasanna & K. Sreenivasa Rao Department of Computer Science & Engineering Indian Institute of Technology Madras, India Email: {yegna,prasanna,ksr}@cs.iitm.ernet.in http://speech.cs.iitm.ernet.in. - PowerPoint PPT PresentationTRANSCRIPT
![Page 1: Speech Enhancement using Excitation Source Information](https://reader035.vdocuments.us/reader035/viewer/2022062518/56814740550346895db47dab/html5/thumbnails/1.jpg)
Speech Enhancement using Excitation Source
Information
B. Yegnanarayana, S.R. Mahadeva Prasanna & K. Sreenivasa Rao
Department of Computer Science & Engineering Indian Institute of Technology Madras, India
Email: {yegna,prasanna,ksr}@cs.iitm.ernet.in
http://speech.cs.iitm.ernet.in
1
![Page 2: Speech Enhancement using Excitation Source Information](https://reader035.vdocuments.us/reader035/viewer/2022062518/56814740550346895db47dab/html5/thumbnails/2.jpg)
• To enhance speech degraded by noise & reverberation using multiple microphone data
• Approach based on excitation source information
• Time-delay estimation using source information
• Coherent addition of Hilbert envelopes of LP residuals
• Derivation of weighted LP residual
• Synthesis of enhanced speech (Demonstration)
Objective & Organization
2
![Page 3: Speech Enhancement using Excitation Source Information](https://reader035.vdocuments.us/reader035/viewer/2022062518/56814740550346895db47dab/html5/thumbnails/3.jpg)
Excitation Source Characteristics
3
(LP residual & Hilbert Envelope (HE) of LP residual)
(a) LP residual, (b) Hilbert Transform & (c) Hilbert Envelope
![Page 4: Speech Enhancement using Excitation Source Information](https://reader035.vdocuments.us/reader035/viewer/2022062518/56814740550346895db47dab/html5/thumbnails/4.jpg)
HEs of LP Residuals of Speech
4
LP residual & its HE of (a)-(b) Clean speech, (c)-(d) Degraded speech
![Page 5: Speech Enhancement using Excitation Source Information](https://reader035.vdocuments.us/reader035/viewer/2022062518/56814740550346895db47dab/html5/thumbnails/5.jpg)
Time-Delay Estimation
• Computing the HEs of two-microphone data
• Computing their cross-correlation
• Estimating time-delay
5
![Page 6: Speech Enhancement using Excitation Source Information](https://reader035.vdocuments.us/reader035/viewer/2022062518/56814740550346895db47dab/html5/thumbnails/6.jpg)
Coherent Addition of HEs of LP Residuals
6
(a) Microphone-1, (b) Microphone-2, (c) Microphone-3, (d) Coherently added & (e) Incoherently added
![Page 7: Speech Enhancement using Excitation Source Information](https://reader035.vdocuments.us/reader035/viewer/2022062518/56814740550346895db47dab/html5/thumbnails/7.jpg)
Basis for Speech Enhancement
• Nature of the coherently added Hilbert envelope is exploited to weight the residual
• Weighting of the LP residual e1(n) is done using
n e1(n) êc(n)e1M(n) =-------------------
n êc(n)
where, êc(n) is the coherently added HE & e1M(n) is the modified residual
7
![Page 8: Speech Enhancement using Excitation Source Information](https://reader035.vdocuments.us/reader035/viewer/2022062518/56814740550346895db47dab/html5/thumbnails/8.jpg)
Results of Enhancement
8
(a) LP residual of degraded speech, (b) Weighted LP residual, (c) Degraded speech & (d) Enhanced speech
![Page 9: Speech Enhancement using Excitation Source Information](https://reader035.vdocuments.us/reader035/viewer/2022062518/56814740550346895db47dab/html5/thumbnails/9.jpg)
Results of Enhancement
9
Spectrogram of (a) Degraded speech, (b) Enhanced speech by proposed approach & (c) Coherently added signal
![Page 10: Speech Enhancement using Excitation Source Information](https://reader035.vdocuments.us/reader035/viewer/2022062518/56814740550346895db47dab/html5/thumbnails/10.jpg)
Summary & Conclusions • New approach for speech enhancement
• Using excitation source information in LP residual
• Coherent addition of HEs of LP residuals
• Enhancement using modified residual
• Need to derive suitable weight function for improvement in the quality of enhancement
10
![Page 11: Speech Enhancement using Excitation Source Information](https://reader035.vdocuments.us/reader035/viewer/2022062518/56814740550346895db47dab/html5/thumbnails/11.jpg)
Paper #1582
Speech Enhancement
using Excitation
Source Information
![Page 12: Speech Enhancement using Excitation Source Information](https://reader035.vdocuments.us/reader035/viewer/2022062518/56814740550346895db47dab/html5/thumbnails/12.jpg)
B. Yegnanarayana
S.R. Mahadeva Prasanna &
K. Sreenivasa Rao
![Page 13: Speech Enhancement using Excitation Source Information](https://reader035.vdocuments.us/reader035/viewer/2022062518/56814740550346895db47dab/html5/thumbnails/13.jpg)
Department of Computer Science & Engineering
Indian Institute of Technology Madras, India
Email: {yegna,prasanna,ksr}@cs.iitm.ernet.in
http://speech.cs.iitm.ernet.in
![Page 14: Speech Enhancement using Excitation Source Information](https://reader035.vdocuments.us/reader035/viewer/2022062518/56814740550346895db47dab/html5/thumbnails/14.jpg)
ABSTRACTThis paper proposes an approach for processing speech from multiple microphones to enhance speech degraded by noise and reverberation. The approach is based on exploiting the features of excitation source in speech production. In particular, the characteristics of voiced speech can be used to derive a coherently added signal from the Linear prediction (LP) residuals of the degraded speech data from different microphones. A weight function is derived from the coherently added signal. For coherent addition the time delay between a pair of microphones is estimated using the knowledge of the source information present in the LP residual. The enhanced speech is generated by exciting the time varying all-pole filter with the weighted LP residual.