Abstract:Reverberation may severely degrade the quality of speech signals recorded using microphones in a room. For compact microphone arrays, the choice of the reference microphone for multi-microphone dereverberation typically does not have a large influence on the dereverberation performance. In contrast, when the microphones are spatially distributed, the choice of the reference microphone may significantly contribute to the dereverberation performance. In this paper, we propose to perform reference microphone selection for the weighted prediction error (WPE) dereverberation algorithm based on the normalized $\ell_p$-norm of the dereverberated output signal. Experimental results for different source positions in a reverberant laboratory show that the proposed method yields a better dereverberation performance than reference microphone selection based on the early-to-late reverberation ratio or signal power.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is: in multi - microphone dereverberation, how to select the best reference microphone to improve the dereverberation performance, especially in spatially - distributed microphone arrays. Specifically, when microphones are distributed in a large space, there may be significant differences in the early - to - late reverberation ratio (ELR) and signal power among different microphones, which makes the selection of the reference microphone have an important impact on the dereverberation effect. ### Problem Background 1. **The Influence of Reverberation**: - Reverberation (i.e., multiple reflections of sound waves in a room) can seriously affect the quality of speech signals, reducing speech clarity and the performance of automatic speech recognition. - In a compact microphone array, the selection of the reference microphone has less impact on the dereverberation performance; but in a spatially - distributed microphone array, this selection may significantly affect the dereverberation effect. 2. **Limitations of Existing Methods**: - Existing reference microphone selection methods are mainly based on the early - to - late reverberation ratio (ELR) or signal power, but these methods fail to fully consider the differences in signal power among different microphones. ### The Method Proposed in the Paper To overcome the above problems, the author proposes a reference microphone selection method based on the normalized ℓp - norm. The specific steps are as follows: 1. **Initial Attempt**: - Based on the cost function of the weighted prediction error (WPE) algorithm, initially attempt to define the reference microphone selection problem as an ℓp - norm minimization problem: \[ \hat{r}^{(I)}_{\ell_p} = \arg\min_r \sum_{f = 1}^F \| x_r(f)-X_{\tau,r}(f)\hat{g}_r^{(I)}(f)\|_p \] where $\hat{g}_r^{(I)}$ is the prediction filter after the $I$ - th iteration. 2. **Normalization Processing**: - Since there are large differences in signal power among different microphones, directly using the ℓp - norm may lead to the selection of the output with the smallest power rather than the output with the highest quality. - Therefore, the author proposes normalization processing and uses the normalized ℓp - norm for selection: \[ \hat{r}^{(I)}_{\ell_p/\ell_2} = \arg\min_r \sum_{f = 1}^F \frac{\| x_r(f)-X_{\tau,r}(f)\hat{g}_r^{(I)}(f)\|_p}{\| x_r(f)-X_{\tau,r}(f)\hat{g}_r^{(I)}(f)\|_2} \] ### Experimental Results The experimental results show that the reference microphone selection method based on the normalized ℓp - norm is superior to the methods based on ELR or signal power. In addition, even with only a small number of WPE iterations, this method can achieve a similar performance improvement. ### Conclusion The paper proposes a new reference microphone selection method based on the normalized ℓp - norm, which effectively improves the dereverberation performance in spatially - distributed microphone arrays. Future research directions include considering more complex scenarios such as noise, moving sound sources, and different microphone sensitivities.

Reference Microphone Selection for the Weighted Prediction Error Algorithm using the Normalized L-p Norm

Microphone Subset Selection for the Weighted Prediction Error Algorithm using a Group Sparsity Penalty

Distributed Speech Dereverberation Using Weighted Prediction Error

Microphone array processing via joint wideband angle-of-arrival estimation and speech feature enhancement

Integrating Plug-and-Play Data Priors with Weighted Prediction Error for Speech Dereverberation

Adaptive Dereverberation, Noise and Interferer Reduction Using Sparse Weighted Linearly Constrained Minimum Power Beamforming

Noise Robust Speech Recognition Using Multi-Channel Based Channel Selection And ChannelWeighting.

Integrated Minimum Mean Squared Error Algorithms for Combined Acoustic Echo Cancellation and Noise Reduction

A neural network-supported two-stage algorithm for lightweight dereverberation on hearing devices

Multichannel Linear Prediction-Based Speech Dereverberation Considering Sparse and Low-Rank Priors

Improved direction of arrival estimations with a wearable microphone array for dynamic environments by reliability weighting

Comparison of Reference Microphone Selection Algorithms for Distributed Microphone Array Based Speech Enhancement in Meeting Recognition Scenarios

Joint DOA Estimation and Dereverberation Based on Multi-Channel Linear Prediction Filtering and Azimuth Sparsity

Robust parameter design for Wiener-based binaural noise reduction methods in hearing aids

Diffusion Posterior Sampling for Informed Single-Channel Dereverberation

Dereverberation using joint estimation of dry speech signal and acoustic system

Predicting Preferred Dialogue-to-Background Loudness Difference in Dialogue-Separated Audio

Run-Time Adaptation of Neural Beamforming for Robust Speech Dereverberation and Denoising

A two-stage speaker extraction algorithm under adverse acoustic conditions using a single-microphone

Speech Dereverberation and Noise Reduction for both diffusive noise field and point noise source in Binaural Hearing Aids: Preliminary Version

Steered Response Power-Based Direction-of-Arrival Estimation Exploiting an Auxiliary Microphone