Audio samples of simulated speech data with different noise types (UPDATE: Nov/2020).
NOISE TYPE
NOISY
ENHANCED DNN
Stadium
City
Airport
Callcenter
Babies
Barks
Audio samples of real speech data from different scenarios in the wild provided by the found data Youtube dataset known as VOXCELEB dataset (Feb/2020)
NOISY
ENHANCED STATISTICAL
ENHANCED DNN (first)
*These are results of a collaboration SONOC-ViVoLab that proposes how to properly obtaining a robust DNN-based SNR estimate by focusing on the accuracy-sensitive block for SNR estimation in the framework of the classical spectral-domain speech estimator algorithm MORE.