Table 2. Experimental design.
| Technique | Techniques used in the multi-channel speech enhancement system | Speech database | SNR/dB | Noises | |
|---|---|---|---|---|---|
| The existing BAV-MCSE | Beamforming, ANR and VAD | AURORA | −10 dB, −5 dB, 0 dB, 5 dB, 10 dB, 15 dB |
Stationary Noise: White Gaussian noise |
Non-stationary Noises: Airport, Babble, Car, Exhibition, Restaurant |
| Proposed DWT-CNN-MCSE | Convolution neural network (CNN) and discrete wavelet feature extraction (DWT) technique | AURORA | −10 dB, −5 dB, 0 dB, 5 dB, 10 dB, 15 dB |
Stationary Noise: White Gaussian noise |
Non-stationary Noises: Airport, Babble, Car, Exhibition, Restaurant |
| LibriSpeech | −10 dB, −5 dB, 0 dB, 5 dB, 10 dB, 15 dB |
N/A | Dog bark noise, Fan noise, Ambulance noise | ||