A complex-valued multichannel speech enhancement learning algorithm for optimal tradeoff between noise reduction and speech distortion.
Jingxian TuYoushen XiaSongchuan ZhangPublished in: Neurocomputing (2017)
Keyphrases
- speech enhancement
- noise reduction
- complex valued
- single channel
- learning algorithm
- noisy environments
- linear prediction
- signal to noise ratio
- multi channel
- real valued
- edge preserving
- speech signal
- edge detection
- recurrent neural networks
- wiener filter
- image coding
- independent component analysis
- median filter
- activation function
- smoothing algorithm
- vocal tract
- pattern recognition
- frequency domain
- machine learning
- artificial neural networks
- wavelet decomposition
- background noise
- lossless compression
- automatic speech recognition
- audio visual
- learning rate