Sound Source Separation by Spectral Subtraction Based on Instantaneous Estimation of Noise Spectrum.
Kenji OzawaMasanori MoriseShuichi SakamotoKanji WatanabePublished in: ICSAI (2019)
Keyphrases
- spectral subtraction
- source separation
- speech signal
- hybrid method
- blind source separation
- speech recognition
- single channel
- speech enhancement
- signal to noise ratio
- sound source
- automatic speech recognition
- multi channel
- audio features
- background noise
- independent component analysis
- hidden markov models
- additive noise
- non stationary
- noisy environments
- noise reduction
- language model
- temporal structure
- frequency band
- denoising
- frequency domain
- pattern recognition
- maximum likelihood
- multi modal
- low level
- face recognition
- feature extraction