Incorporating Intra-Spectral Dependencies with a Recurrent Output Layer for Improved Speech Enhancement.
Khandokar Md. NayemDonald S. WilliamsonPublished in: MLSP (2019)
Keyphrases
- speech enhancement
- feed forward
- linear prediction
- output layer
- hidden layer
- noisy environments
- noise reduction
- recurrent neural networks
- neural network
- speech signal
- signal to noise ratio
- single channel
- spectral analysis
- pattern recognition
- learning rules
- machine learning
- smoothing algorithm
- vocal tract
- similarity measure
- background noise
- training data
- feedforward neural networks
- artificial neural networks
- image coding
- edge detection
- multiresolution