This DE measure was an extension of Perceptual Evaluation of Audio Quality, and required reference and test signals.
Motivated by this, we propose the residual-dense lattice network (RDL-Net), which is a new CNN for speech enhancement that employs both residual and dense aggregations without over-allocating parameters for feature re-usage.
Ranked #10 on Speech Enhancement on DEMAND
Deep learning has achieved substantial improvement on single-channel speech enhancement tasks.
MMSE approaches utilising the proposed a priori SNR estimator are able to achieve higher enhanced speech quality and intelligibility scores than recent masking- and mapping-based deep learning approaches.