Figure 4From: A perceptual masking approach for noise robust speech recognition The ATH, original PSD, noise and tones, and global masking threshold for part of the utterance “three six seven” corrupted with subway noise of 5 dB SNR. Back to article page