Room Impulse Response Estimation Research Articles

Efficient Implementation of Kernel Regularization Based on ADMM and Its Application to Room Impulse Response Estimation

Efficient Implementation of Kernel Regularization Based on ADMM and Its Application to Room Impulse Response Estimation

DEchorate: a calibrated room impulse response dataset for echo-aware signal processing

This paper presents a new dataset of measured multichannel room impulse responses (RIRs) named dEchorate. It includes annotations of early echo timings and 3D positions of microphones, real sources, and image sources under different wall configurations in a cuboid room. These data provide a tool for benchmarking recent methods in echo-aware speech enhancement, room geometry estimation, RIR estimation, acoustic echo retrieval, microphone calibration, echo labeling, and reflector position estimation. The dataset is provided with software utilities to easily access, manipulate, and visualize the data as well as baseline methods for echo-related tasks.

Correction of room impulse response truncation based on a nonlinear decay model

Serious effects of noise present in room impulse responses (RIRs) on evaluation of room acoustical parameters can be compensated by applying different methods. One of the preferred methods is based on the noise subtraction, RIR truncation and correction for truncation. This paper presents a procedure for calculation of the correction term that should be added to the backward integrated truncated RIR. The procedure can be applied in an automated way, and it is based on a nonlinear decay model consisting of an exponential decay plus stationary noise. Unknown parameters of the model are calculated by an optimization procedure where the model is fitted to the measurements. The proposed procedure for correction term calculation is tested on both synthesized and measured RIRs. The effects of the procedure parameters related to the RIR estimation range used for fitting the model to the measurements are analyzed. This procedure is compared with the one often applied in practice as a reference. The comparison shows that the proposed procedure outperforms the reference one. In addition, the impact of changing the correction term on the reverberation time estimation is investigated.

Adaptive estimation and reshaping of room impulse response

In this paper, we focus on the reverberation depression in scenarios such as hand-free telephone and teleconference system applications. The combination of cross-relation based blind room impulse response (RIR) estimation and the p-norm based channel reshaping leads to this adaptive reverberation depression scheme. The normalized multi-channel frequency-domain least-mean-square (NMCFLMS) algorithm is used for RIR estimation and the p-norm optimization approach for channel reshaping. Simulations show that it is possible to establish such a post-processing system for reverberation depression. Listening tests verified that reverberation is hardly heard after channel reshaping.

Regularization for Partial Multichannel Equalization for Speech Dereverberation

Acoustic multichannel equalization techniques such as the multiple-input/output inverse theorem (MINT), which aim to equalize the room impulse responses (RIRs) between the source and the microphone array, are known to be highly sensitive to RIR estimation errors. To increase robustness, it has been proposed to incorporate regularization in order to decrease the energy of the equalization filters. In addition, more robust partial multichannel equalization techniques such as relaxed multichannel least-squares (RMCLS) and channel shortening (CS) have recently been proposed. In this paper, we propose a partial multichannel equalization technique based on MINT (P-MINT) which aims to shorten the RIR. Furthermore, we investigate the effectiveness of incorporating regularization to further increase the robustness of P-MINT and the aforementioned partial multichannel equalization techniques, i.e., RMCLS and CS. In addition, we introduce an automatic non-intrusive procedure for determining the regularization parameter based on the L-curve. Simulation results using measured RIRs show that incorporating regularization in P-MINT yields a significant performance improvement in the presence of RIR estimation errors, whereas a smaller performance improvement is observed when incorporating regularization in RMCLS and CS. Furthermore, it is shown that the intrusively regularized P-MINT technique outperforms all other investigated intrusively regularized multichannel equalization techniques in terms of perceptual speech quality (PESQ). Finally, it is shown that the automatic non-intrusive regularization parameter in regularized P-MINT leads to a very similar performance as the intrusively determined optimal regularization parameter, making regularized P-MINT a robust, perceptually advantageous, and practically applicable multichannel equalization technique for speech dereverberation.

Convex Regularizations for the Simultaneous Recording of Room Impulse Responses

We propose to acquire large sets of room impulse responses (RIRs) by simultaneously playing known source signals on multiple loudspeakers. We then estimate the RIRs via a convex optimization algorithm using convex penalties promoting sparsity and/or exponential amplitude envelope. We validate this approach on real-world recordings. The proposed algorithm makes it possible to estimate the RIRs to a reasonable accuracy even when the number of recorded samples is smaller than the number of RIR samples to be estimated, thereby leading to a speedup of the recording process compared to state-of-the-art RIR acquisition techniques. Moreover, the penalty promoting both sparsity and exponential amplitude envelope provides the best results in terms of robustness to the choice of its parameters, thereby consolidating the evidence in favor of sparse regularization for RIR estimation. Finally, the impact of the choice of the emitted signals is analyzed and evaluated.

On the quality-assessment of reverberated speech

This paper addresses the problem of quantifying the reverberation effect in speech signals. The perception of reverberation is assessed based on a new measure combining the characteristics of reverberation time, room spectral variance, and direct-to-reverberant energy ratio, which are estimated from the associated room impulse response (RIR). The practical aspects behind a robust RIR estimation are underlined, allowing an effective feature extraction for reverberation evaluation. The resulting objective metric achieves a correlation factor of about 90% with the subjective scores of two distinct speech databases, illustrating the system’s ability to assess the reverberation effect in a reliable manner.

A hybrid MLS technique for room impulse response estimation

The measurement of room impulse response (RIR) when there are high background noise levels frequently means one must deal with very low signal-to-noise ratios (SNR). If such is the case, the measurement might yield unreliable results, even when synchronous averaging techniques are used. Furthermore, if there are non-linearities in the apparatus or system time variances, the final SNR can be severely degraded. The test signals used in RIR measurement are often disturbed by non-stationary ambient noise components. A novel approach based on the energy analysis of ambient noise – both in the time and in frequency – was considered. A modified maximum length sequence (MLS) measurement technique, referred to herein as the hybrid MLS technique, was developed for use in room acoustics. The technique consists of reducing the noise energy of the captured sequences before applying the averaging technique in order to improve the overall SNRs and frequency response accuracy. Experiments were conducted under real conditions with different types of underlying ambient noises. Results are shown and discussed. Advantages and disadvantages of the hybrid MLS technique over standard MLS technique are evaluated and discussed. Our findings show that the new technique leads to a significant increase in the overall SNR.

Segmented swept sine technique for room impulse response estimation

The swept sine technique is very popular for the measurement of room impulse response due to its large signal-to-noise ratio (SNR) and immunity against subtle time-variance and nonlinearity of the room under test. However, the acoustical measurements in the presence of high background noise levels leads usually to unsatisfactory results. For situations of high level nonstationary noise, the mean square of the overall sequence must be minimized in order to increase the SNR. The new technique named segmented swept sine consists in exciting the room with a set of M swept sine signals with application of a dedicated weighting average method and a noise pattern recognition procedure. Each swept sine signal is segmented followed by the estimation of the mean square (MS) value of the respective segment. A weighting procedure is applied to each segment according to the MS value and the likelihood between them. The average technique is then applied. This procedure ensures that the resulting swept sine signal has the highest SNR value. Several examples are presented to compare the standard swept sine and the segmented swept sine techniques, giving advantages and disadvantages of each technique.

Bayesian regularization and nonnegative deconvolution for room impulse response estimation

This paper proposes Bayesian Regularization And Nonnegative Deconvolution (BRAND) for accurately and robustly estimating acoustic room impulse responses for applications such as time-delay estimation and echo cancellation. Similar to conventional deconvolution methods, BRAND estimates the coefficients of convolutive finite-impulse-response (FIR) filters using least-square optimization. However, BRAND exploits the nonnegative, sparse structure of acoustic room impulse responses with nonnegativity constraints and L/sub 1/-norm sparsity regularization on the filter coefficients. The optimization problem is modeled within the context of a probabilistic Bayesian framework, and expectation-maximization (EM) is used to derive efficient update rules for estimating the optimal regularization parameters. BRAND is demonstrated on two representative examples, subsample time-delay estimation in reverberant environments and acoustic echo cancellation. The results presented in this paper show the advantages of BRAND in high temporal resolution and robustness to ambient noise compared with other conventional techniques.