Figure 9From: Enhancing the magnitude spectrum of speech features for robust speech recognition The MMSE log-STSA–processed spectrogram of an utterance under two SNR levels: (a) clean, (b) 5 dB. Back to article page