Time-Frequency Masking Strategies for Single-Channel Low-Latency Speech Enhancement Using Neural Networks.
Mikko ParviainenPasi PertiläTuomas VirtanenPeter GroschePublished in: IWAENC (2018)
Keyphrases
- single channel
- speech enhancement
- low latency
- neural network
- multi channel
- frequency domain
- high speed
- signal processing
- high throughput
- independent component analysis
- noise reduction
- highly efficient
- real time
- signal to noise ratio
- prior information
- wavelet decomposition
- virtual machine
- noisy environments
- pattern recognition
- brain activity
- wavelet transform
- speech signal
- spatial domain
- human visual system
- feature extraction
- wiener filter
- linear prediction
- multimedia