Speech enhancement based on the integration of fully convolutional network, temporal lowpass filtering and spectrogram masking.
Kuan-Yi LiuSyu-Siang WangYu TsaoJeih-Weih HungPublished in: ROCLING (2019)
Keyphrases
- low pass
- high pass
- speech enhancement
- single channel
- wavelet decomposition
- speech signal
- bandpass
- frequency domain
- convolutional network
- subband
- high frequency
- convolutional neural networks
- filter bank
- multi channel
- signal to noise ratio
- multiresolution
- noisy environments
- noise reduction
- wavelet transform
- discrete wavelet transform
- multiscale
- coarse to fine
- speech recognition
- low frequency
- human visual system
- wiener filter
- computer vision
- bit rate
- contrast enhancement
- image fusion
- frequency band
- linear prediction
- image coding
- non stationary
- image quality