Fig. 3From: Front-end technologies for robust ASR in reverberant environments—spectral enhancement-based dereverberation and auditory modulation filterbank featuresDistribution of (T 60,DRR) values for the RIRs provided by the REVERB challenge (for ASR multi-condition training). Each dot represents one single-channel RIR with its (T 60,DRR). Twenty-eight classes are defined based on this distribution (labels 0 to 27 inside the class boundary boxes)Back to article page