Skip to main content
. Author manuscript; available in PMC: 2009 May 7.
Published in final edited form as: J Acoust Soc Am. 2007 Aug;122(2):1165–1172. doi: 10.1121/1.2749454

Figure 2.

Figure 2

Top panel shows the spectrogram of a sentence in quiet from the IEEE corpus. Middle panel shows the spectrogram of the sentence embedded in multi-talker babble at −5 dB SNR. Bottom panel shows the ideal binary mask using an SNR threshold of 0 dB, with white pixels indicating a 0 (target weaker than the masker) and black pixels indicating a 1 (target stronger than the masker).