Lae-Hoon Kim

Generalized Optimal Multi-Microphone Speech Enhancement Using Sequential Minimum Variance Distortionless Response(MVDR) Beamforming and Postfiltering

A theoretical basis for optimal multichannel speech enhancements presented, sufficient, flexible to be used with any assumed statistical model and optimality criterion. Any Bayesian optimal one-channel estimator for speech enhancement can be generalized to the multichannel case as a sequentially constructed minimum variance distortionless response (MVDR) beamformer followed by an optimal one-channel postfilter. We present experimental results using the minimum mean-square error log-spectral amplitude (MMSE-logSA) optimality criterion, applied to a statistical model with simplified channel but realistic inter-microphone noise coherence. Word error rate in the audio-visual speech in a car (AVICAR) corpus (moving car, windows open) is reduced from 18% to 9%

Reverberated speech signal separation based on regularized subband feedforward ICA and instantaneous direction of arrival

In this paper, independent component analysis (ICA) in a subband domain has been extended into a feed-forward network. The feed-forward network maximizes mutual independence of separated current frames using information from the both current and previous multi-channel frames of speech signals captured by a microphone array. To guide into a proper separation preventing permutation and arbitrary scaling, we not only rely on the steered response for the first tap of the demixing filter but also penalize on the direction thus drastically increasing the mean squared error with the spatial filtered output. After convergence, by applying instantaneous direction of arrival (IDOA) based post-processing, we can additionally suppress the leakage of the interference as well as the reverberated target signal. The signal to interference ratio (SIR) is improved more than 20 dBC for distances up to 2.7 m and angle differences down to 26°.

Equalization of low frequency response in automobile

In automobile space, we experience sound coloring of reproduced sound quite differently from large spaces such as concert halls. This is assumed to be due to the well-separated acoustic modes in the low frequency range up to the relatively high crossover frequency. Such unwanted sound coloring can be reduced through equalization of the well-separated modes. However, it is not a simple process as binaural responses are different for every person and drivers are likely to move their head during driving. We introduce a novel approach, based on minimum phase inversion, to the equalization of the low frequency response in order to compensate for the coloring. We then compare the proposed approach with the conventional least squares based inversion and attempt to show the superiority of this approach through experimental and listening test results.

Optimal Multi-Microphone Speech Enhancement in Cars

Hands-free speech telephony and speech recognition in cars suffer from additive noise and reverberation. We propose an iterative blind room impulse response (RIR) estimation algorithm based on an analysis-by-synthesis loop closed around a multi-path generalized sidelobe canceller (GSC). By combining a post-filter with the proposed scheme, optimal speech enhancement in practical situations can be achieved. The algorithm is tested using simulated data and real speech recordings from the AVICAR database.

Optimal speech estimator considering room response as well as additive noise: Different approaches in low and high frequency range

This paper proposes minimum mean squared error (MMSE) speech signal estimation in a reverberant space using different optimal estimators in the low and high frequency ranges. At low frequencies, an MMSE spectral amplitude estimator divided by the spectral amplitude of a representative impulse response produces optimal performance. In the high frequency range, the MMSE estimator is computed based on its sufficient statistic: the maximum likelihood (ML) estimate. Inference is factored using a two- step algorithm: the maximum likelihood value of the source spectrum is first estimated using expectation-maximization (EM) under the assumption of the hidden room response with complex Gaussian pdf, then the MMSE source spectral estimate is computed.

Acoustic model for robustness analysis of optimal multipoint room equalization

In this paper, an acoustic model for the robustness analysis of optimal multipoint room equalization is proposed. The optimal multipoint equalization aims to have the optimal performance in a least-squares sense for all measured points. The model can be used for theoretical robustness estimation depending on the critical design parameters such as the number of measurement points, the distance between measurements, or the frequency before applying real equalization system. The analysis results show that it is important to set the appropriate number of measurement points and the distances between measurement points to ensure the enlarged equalization region at a specific frequency.

A novel approach for the equalization of low frequency response in the automotive space

We experience the coloring of reproduced sound in an automotive space in a different way than in a large space such as a concert hall. It comes from the well-separated acoustic modes in the low frequency range up to a relatively high crossover frequency. The unwanted sound coloring can be reduced by equalization. However, this is not a simple matter because the binaural response is different in each person and drivers are likely to move their heads while driving. We introduce a novel approach, based on minimum phase inversion, for the equalization of the low frequency response to compensate the coloring. We compare the proposed approach with the conventional least square based inversion and show the superiority of our approach by experiment. This is confirmed by the results of listening tests.

Joint estimation of DOA and speech based on EM beamforming

In this paper, we propose a multi-microphone joint optimal estimation of the direction of arrival (DOA) and the source speech signal through newly introduced EM beamforming. This produces a posterior PDF for the DOA, based only on the reliable speech spectrum. By maximizing over the posterior PDF of the DOA, we achieve maximum a posteriori DOA estimation. After convergence, the estimated source spectrum through weighted sum in the Bayesian sense is a maximum likelihood estimate (MLE). This is a sufficient statistic for minimum mean square error (MMSE) optimal estimation using a subsequent single channel MMSE filter.

Toward overcoming fundamental limitation in frequency-domain blind source separation for reverberant speech mixtures

Blind source separation can be implemented in the frequency domain using one-tap multiplication operation in each frequency bin, but only when the frame length is long enough to disregard temporal aliasing effects. If we take a short-time frequency transformation with a window shorter than a room reverberation time, the justification above does not hold anymore. In this paper, we present an appropriate representation in the short-time frequency domain. The suitability is justified by showing the equivalence with the original time domain approach under the overlap-add context. Experimental validation using a corpus synthesized by convolution with measured sets of room impulse responses is also provided.

Speech enhancement beyond minimum mean squared error with perceptual noise shaping.

Residual error signal after speech enhancement through linear filtering can be decomposed into two disjoint portions: speech signal distortion and background noise suppression. Speech is known to follow a super‐Gaussian probabilistic distribution function (PDF) such as Laplacian, while background noise follows Gaussian PDF. Minimum mean squared error estimation requires only second order statistics not only for the noise but also for the speech. Therefore higher‐order dependence of observed speech on the original speech may cause leakage of speech information into the error residual. This talk will formulate an optimization problem minimizing higher‐order statistics (HOS) as well as energy of the signal distortion constrained by a limit on the maximum audibility of the residual noise. Note that due to the non‐stationary nature of speech, we perform the speech enhancement in short overlapping frames. Minimizing HOS of the speech distortion ensures that the speech distortion includes only noise terms, with ...