Demo Title:
Blind Speech Extraction Based on MMSE STSA Estimator and
ICA-Based Noise Estimation

Authors:
- Masanobu Go (e-mail: masanobu-g@is.naist.jp)
- Hiroshi Saruwatari (e-mail: sawatari@is.naist.jp)
- Takayuki Inoue
- Kiyohiro Shikano
(Go and Saruwatari are corresponding persons.)

Affiliation:
- Nara Institute of Science and Technology, JAPAN

----------------------------------------------
Abstract:

We propose a new method for robust blind speech extraction
under realistic noise environments. We use a combination
method of ICA-based noise estimation [1], channel-wise
nonlinear denoising processing [2] based on Dynamic minimum
mean-square error (MMSE) short-time spectral amplitude (STSA)
estimator [3], DOA estimator by Minimum Variance method,
and (4) Delay-and-Sum (DS) beamformer steered to the estimated DOA.
First, Infomax-type ICA-based noise estimator can cancel out
the target point-source speech and extract the noise estimate
with high accuracy. Next, we apply channel-wise denoising
processing [2] to the observed signals based on the estimated
noise, in which we use a Dynamic MMSE STSA estimator [3]
with momentarily estimated instantaneous a posteriori SNR.
Finally, based on the DOA estimated from the channel-wise
outputs, we summarize the channel-wise MMSE STSA estimator
outputs via DS. Such a "channel-wise nonlinear processing
with post-DS" structure leads to less musical noise [2].

References:
[1] Y. Takahashi, T. Takatani, K. Osako, H. Saruwatari, and
K. Shikano, ``Blind spatial subtraction array for speech
enhancement in noisy environment,'' IEEE Transactions on Audio,
Speech and Language Processing, vol.17, no.4, pp.650--664, 2009.

[2] Y. Takahashi, Y. Uemura, H. Saruwatari, K. Shikano, K. Kondo,
``Musical noise analysis based on higher order statistics for
microphone array and nonlinear signal processing,"
Proc. IEEE International Conference on Acoustics, Speech, and
Signal Processing (ICASSP2009), pp.229--232, 2009.

[3] R. Okamoto, Y. Takahashi, H. Saruwatari, K. Shikano,
``MMSE STSA estimator with nonstationary noise estimation
based on ICA for high-quality speech enhancement,"
Proc. ICASSP 2010, pp.4778--4781, 2010.