Search papers, labs, and topics across Lattice.
This paper explores deep learning techniques for estimating the Relative Transfer Matrix (ReTM) in the context of speech enhancement using multiple microphones and sources. By leveraging time and short-time frequency transform domain convolutional networks alongside Long Short-Term Memory networks, the authors present three supervised learning frameworks that significantly outperform traditional covariance-based methods in ReTM estimation. Experimental results indicate that these models not only enhance estimation accuracy across five objective metrics but also deliver competitive performance in speech enhancement applications.
Deep learning models can dramatically improve the accuracy of Relative Transfer Matrix estimation, outperforming traditional methods in noisy environments.
The Relative Transfer Matrix (ReTM), recently introduced as a generalization of the relative transfer function for multiple receivers and sources, shows promising performance when applied to speech enhancement in noisy environments. Estimating the ReTM of sound sources by exploiting the covariance matrices of multichannel recordings is highly beneficial for practical applications and, to date, remains the only proposed approach. This paper investigates deep learning-based ReTM estimation. We propose three novel supervised learning frameworks using time and short-time frequency transform domain convolutional networks, and a Long Short-Term Memory-based recurrent neural network. Experimental results demonstrate that the proposed models achieve more accurate estimation of the ReTM using five objective metrics compared to the covariance-based method. We also show the effectiveness of the proposed frameworks for speech enhancement, achieving performance on par with the baseline method.
