Speech Enhancement - Audio Samples

homeHome: Sharon's Homepage


Single Microphone

1.      S. Gannot, D. Burshtein and E. Weinstein, "Iterative and Sequential Kalman Filter-Based Speech Enhancement Algorithms,"
IEEE Trans. on Speech and Audio Proc., vol. 6, no. 4, pp. 373-385, Jul. 1998. [KEM]

2.      D. Burshtein and S. Gannot, "Speech Enhancement Using a Mixture-Maximum Model,"
IEEE Trans. on Signal Processing, Vol. 49, No. 8, pp. 1614-1626, Aug. 2001. [MixMaX]

3.       Y. Ephraim, D. Malah and B. H. Juang, "On the Application of Hidden Markov Models for Enhancing Noisy Speech,"
IEEE Trans. Acoust., Speech and Sig. Proc., vol. 37, pp. 1846-1856, 1989. [HMM]

4.      Y. Ephraim, "A Bayesian Estimation Approach for Speech Enhancement Using Hidden Markov Models,"
IEEE Trans. on Sig. Proc., vol. 40, pp. 725-735, 1992. [HMM]


Remark:

The implementation of the HMM algorithm was made for comparison
by the authors of the KEM and MixMax algorithms and on their own responsibility.

Databases

  • TIMIT
  • Amsterdam Free University
  • NOISEX-92

SOURCE

CLEAN

NOISY

HMM

KEM

MIXMAX

ENGLISH

CAR noise, Female, 6dB

CAR noise, Male, -3dB

ROOM noise, Female, 9dB

Speech noise, Male, 3dB

Speech noise, Female, 3dB

White noise, Male, 9dB

Factory noise, Female, 6dB

Factory noise, Male, 6dB

DUTCH

Factory noise, Male, 6dB

Factory noise, Female, 6dB

Car noise, Female, 0dB

CAR noise, Male, -6dB

Speech noise, Male, 3dB

Speech noise, Female, 3dB

ENGLISH, No Post Processing

CAR noise, Female, -9dB

Speech noise, Female, 0dB

White noise, Female, 6dB

CAR noise, Male, -6dB

Single Microphone

Microphone Array

1.      S. Gannot, D. Burshtein and E. Weinstein, "Signal Enhancement Using Beamforming and Non-Stationarity with application to Speech,"
IEEE Trans. on Sig. Proc., vol. 49, no. 8, pp. 1614-1626, Aug. 2001.
[TF-GSC]

2.       L. J. Griffiths and C. W. Jim,  "An Alternative Approach to Linearly Constrained Adaptive Beamforming,"
IEEE Trans. on Antennas and Propagation, vol. 30, no. 1, pp. 27-35, Jan. 1982. [D-GSC]

 

SOURCE

CLEAN

NOISY

D-GSC

TF-GSC

+MIXMAX

Fan Noise,Room,-3dB

Microphone Array

 

 


Microphone Array with Postfiltering

 

  1. I. Cohen and B. Berdugo, "Speech Enhancement for Nonstationary noise environments,"
    Signal Processing, vol. 81, pp. 2403-2418, Aug. 2001.
    [OM-LSA]
     
  2. Sharon Gannot and Israel Cohen, "Speech Enhancement Based on the General Transfer Function GSC and Postfiltering".
    IEEE Trans. on Speech and Audio Processing, vol. 12, No. 6, pp. 561-571, Nov. 2004.
    [MULTI]
     
  3. I. Cohen, S. Gannot and B. Berdugo,  "An Integrated Real-Time Beamforming and Postfiltering System for Non-Stationary Noise Environments,"
    EURASIP Journal on Applied Signal Processing, special issue on signal processing for acoustic communication systems, Vol. 2003, No. 11, pp. 1064-1073, Oct. 2003.
    [MULTI] 


SOURCE

CLEAN

NOISY

TF-GSC

+MIXMAX

+OMLSA

+MULTI

Car,-6dB

Car,3dB

Room,Direct.,-6dB

Room,Direct.,3dB

Room,Diffused,3dB

Room,Diff., NonSt.,3dB

Microphone Array and Postfiltering

Joint Noise Reduction and Acoustic Echo Cancellation

  1. Gal Reuven, Sharon Gannot and Israel Cohen, "Joint Noise Reduction and Acoustic Echo Cancellation using the Transfer-Function Generalized Sidelobe Canceller". Speech Communication, special issue on Speech Enhancement, Volume 49, Issues 7-8, July-August 2007, Pages 623-635. [ETF-GSC]


SOURCE

Noisy

ETF-GSC

Directional noise
SNR=SER=5dB
T60=200ms



Noise and Echo 

 

Dual Source

 

  1. Gal Reuven, Sharon Gannot and Israel Cohen, "Dual Source Transfer-Function Generalized Sidelobe Canceller", IEEE Transactions on Speech and Audio Processing,
    may, 2008.
    [DTF-GSC ]

 

SOURCE

Noisy

TF-GSC

DTF-GSC

Directional noise
SNR=SIR=5dB
T60=40ms

 

Diffused noise
SNR=SIR=5dB
T60=40ms

 

Directional noise
SNR=SIR=5dB
T60=300ms

Dual Source



Multiple Sources

 

  1. S. Markovich, S. Gannot and I. Cohen , "Multichannel Eigenspace Beamforming in a Reverberant Noisy Environment with Multiple Interfering Speech Signals", submitted to IEEE Transactions on Speech and Audio Processing,
    Sep., 2008.

 

 

SOURCE

Noisy

Proposed

2 desired sources

2 non-stationary interference signals (SIR=6dB)

1 stationary noise (SIR=13dB)

T60=300ms, 11 microphones

Multiple Sources


CTF-GSC

 

  1. Ronen Talmon, Israel Cohen and Sharon Gannot, "Convolutive Transfer Function Generalized Sidelobe Canceler", IEEE Transactions on Audio, Speech and Language Processing. Accepted for publication, Mar. 2009.

SOURCE

Noisy

TF-GSC

CTF-GSC

Known RTF
SNR=0dB, T60=500ms

Fig. 4 (b-d)

Estimated RTF
SNR=5dB, T60=500ms

Fig. 6 (b-d)

CTF-GSC

 

 

Transient Noise Reduction Using Nonlocal Diffusion Filters (Submitted to IEEE TSAL)

Ronen Talmon, Israel Cohen and Sharon Gannot

 

SOURCE

Noisy

Enhanced

Household 1

 

Household 2

Metronome

 

Transient NR


Sharon Gannot

 

hit counter
hit counter