Relative Transfer Function Estimation Research Articles

A recursive expectation-maximization algorithm for speaker tracking and separation

The problem of blind and online speaker localization and separation using multiple microphones is addressed based on the recursive expectation-maximization (REM) procedure. A two-stage REM-based algorithm is proposed: (1) multi-speaker direction of arrival (DOA) estimation and (2) multi-speaker relative transfer function (RTF) estimation. The DOA estimation task uses only the time frequency (TF) bins dominated by a single speaker while the entire frequency range is not required to accomplish this task. In contrast, the RTF estimation task requires the entire frequency range in order to estimate the RTF for each frequency bin. Accordingly, a different statistical model is used for the two tasks. The first REM model is applied under the assumption that the speech signal is sparse in the TF domain, and utilizes a mixture of Gaussians (MoG) model to identify the TF bins associated with a single dominant speaker. The corresponding DOAs are estimated using these bins. The second REM model is applied under the assumption that the speakers are concurrently active in all TF bins and consequently applies a multichannel Wiener filter (MCWF) to separate the speakers. As a result of the assumption of the concurrent speakers, a more precise TF map of the speakers’ activity is obtained. The RTFs are estimated using the outputs of the MCWF-beamformer (BF), which are constructed using the DOAs obtained in the previous stage. Next, using the linearly constrained minimum variance (LCMV)-BF that utilizes the estimated RTFs, the speech signals are separated. The algorithm is evaluated using real-life scenarios of two speakers. Evaluation of the mean absolute error (MAE) of the estimated DOAs and the separation capabilities, demonstrates significant improvement w.r.t. a baseline DOA estimation and speaker separation algorithm.

Sequential Estimation of Relative Transfer Function in Application of Acoustic Beamforming

In this paper, a sequential approach is proposed to estimate the relative transfer functions (RTF) used in developing a generalized sidelobe canceller (GSC). The latency in calibrating microphone arrays for GSC, often suffered by conventional approaches involving batch operations, is significantly reduced in the proposed sequential method. This is accomplished by an immediate generation of the RTF from initial input segments and subsequent updates of the RTF as the input stream continues. From the experimental results via the mean square error (MSE) criterion, it has been shown that the proposed method exhibits improved performance over the conventional batch approach as well as over recently introduced least mean squares approaches.

Cognitive-Driven Binaural Beamforming Using EEG-Based Auditory Attention Decoding

Identifying the target speaker in hearing aid applications is an essential ingredient to improve speech intelligibility. Recently, a least-squares-based auditory attention decoding (AAD) method has been proposed to identify the target speaker from single-trial EEG recordings in an acoustic scenario with two competing speakers. Aiming at enhancing the target speaker and suppressing the interfering speaker and ambient noise, in this article, we propose a cognitive-driven speech enhancement system, consisting of a binaural beamformer which is steered based on AAD and estimated relative transfer function (RTF) vectors, which require estimates of the direction-of-arrivals (DOAs) of both speakers. For binaural beamforming and to generate reference signals for AAD, we consider either minimum-variance-distortionless-response (MVDR) beamformers or linearly-constrained-minimum-variance (LCMV) beamformers. Contrary to the binaural MVDR beamformer, the binaural LCMV beamformer allows to preserve the spatial impression of the acoustic scene and to control the suppression of the interfering speaker, which is important when intending to switch attention between speakers. The speech enhancement performance of the proposed system is evaluated in terms of the binaural signal-to-interference-plus-noise ratio ( $\text {SINR}$ ) improvement in anechoic and reverberant conditions. Furthermore, we investigate the impact of RTF and DOA estimation errors and AAD errors on the speech enhancement performance. The experimental results show that the proposed system using LCMV beamformers yields a larger decoding performance and binaural $\text {SINR}$ improvement compared to using MVDR beamformers.

Individualized Interaural Feature Learning and Personalized Binaural Localization Model

The increasing importance of spatial audio technologies has demonstrated the need and importance of correctly adapting to the individual characteristics of the human auditory system, and illustrates the crucial need for humanoid localization systems for testing these technologies. To this end, this paper introduces a novel feature analysis and selection approach for binaural localization and builds a probabilistic localization mapping model, especially useful for the vertical dimension localization. The approach uses the mutual information as a metric to evaluate the most significant frequencies of the interaural phase difference and interaural level difference. Then, by using the random forest algorithm and embedding the mutual information as a feature selection criteria, the feature selection procedures are encoded with the training of the localization mapping. The trained mapping model is capable of using interaural features more efficiently, and, because of the multiple-tree-based model structure, the localization model shows robust performance to noise and interference. By integrating the direct path relative transfer function estimation, we propose to devise a novel localization approach that has improved performance in the presence of noise and reverberation. The proposed mapping model is compared with the state-of-the-art manifold learning procedure in different acoustical configurations, and a more accurate and robust output can be observed.

A multichannel subspace approach with signal presence probability for speech enhancement

For the last few decades, speech enhancement based on microphone arrays has primarily utilized prior information about system models, e.g., array geometry and source location. However, estimation of the time delay to align microphone inputs is largely affected by reverberation and microphone mismatch. Preprocessing time aligning, e.g., fixed beamforming (the first branch of the generalized sidelobe canceller), is not desirable in general applications. Recently, interest has shifted to linear filtering, which works with only second-order statistics of noisy input and estimated noise. This paper proposes a linear filter design based on a multichannel subspace approach for speech enhancement. The contribution of the proposed multichannel subspace methods is threefold. First, a linear filter is applied to the multichannel frequency domain using a spatiospectral correlation matrix. Next, three types of multichannel signal presence probability (MC-SPP) are derived in the subspace domain. Third, incorporating the MC-SPPs into the gain modification of the linear filter achieves further improved noise reduction performance. Of the gain modifications, the proposed gain modification with subspace probability related to the eigenvector corresponding to the maximum eigenvalue realized the best noise reduction performance. The evaluation on average improved the proposed subspace-based methods by approximately 4 dB in overall SNR while maintaining a similar cepstral distance measured over the minimum variance distortionless response with the state-of-the-art relative transfer function estimation in adverse noisy environments.

Relative transfer function (RTF) estimation utilising peaks in time‐domain RTF

An effective algorithm for a more accurate estimation of relative transfer function (RTF) to improve the directional-blind source separation-based RTF estimation is presented. The peaks in time-domain RTF (PTDR) as a feature of time-domain RTF to improve the conventional algorithm are utilised. The PTDRs contain important characteristics of RTF; however, they are smoothed due to the geometric constraint in the conventional algorithm. In the proposed algorithm, the PTDRs are recovered by employing the semi-non-negative matrix factorisation algorithm to effectively find the PTDRs from other peaks. A series of experiments using interfere and various noises demonstrates effectiveness of the proposed RTF estimation algorithm utilising the PTDR.