Table 3. Dataset used for experimenting multi-channel speech enhancement.

Speech enhancement	Speech database	Training and test data	SNR/dB	Types of noises
The existing BAV-MCSE	AURORA	• 25 utterances of clean speech signals for training • 42 sets of noise mixed signals used for testing	−10 dB, −5 dB, 0 dB, 5 dB, 10 dB, 15 dB	Airport, Babble, Car, Exhibition, restaurant and white Gaussian noise
The proposed DWT-CNN-MCSE	AURORA	• 80% of signals used for training to train the CNN model. • 20% of signals used for testing	−10 dB, −5 dB, 0 dB, 5 dB, 10 dB, 15 dB	Airport, Babble, Car, Exhibition, restaurant and white Gaussian noise
	LibriSpeech	• 80% of signals used for training to train the CNN model. • 20% of signals used for testing	−10 dB, −5 dB, 0 dB, 5 dB, 10 dB, 15 dB	Dog bark noise, Fan noise, Ambulance noise