Abstract: In this work, we propose CleanMel, a single-channel Mel-spectrogram denoising and dereverberation network for improving both speech quality and automatic speech recognition (ASR) performance ...

Abstract: Acoustic features play an important role in improving the quality of the synthesised speech. Currently, the Mel spectrogram is a widely employed acoustic feature in most acoustic models.

All the datasets must be located in the datasets folder. This folder should contain the following subfolders after downloading the datasets: GTZAN Speech_Music: Contains the GTZAN Speech Music dataset ...

Autotunable parameters with direct physical interpretation. Easy visualization of all intermediate workflow steps. Collected cluster statistics allow for fine-grained QC and classification of signals.