Speech Enhancement with Natural Sounding Residual Noise Based on Connected Time-Frequency Speech Presence Regions.
Karsten Vandborg SørensenSøren Vang AndersenPublished in: EURASIP J. Adv. Signal Process. (2005)
Keyphrases
- speech enhancement
- noisy environments
- noise reduction
- signal to noise ratio
- speech signal
- noisy speech
- single channel
- linear prediction
- vocal tract
- speech recognition
- background noise
- spectral subtraction
- wiener filter
- additive noise
- smoothing algorithm
- multi channel
- edge preserving
- automatic speech recognition
- image sequences
- information retrieval
- image coding
- frequency domain
- non stationary
- edge detection
- wavelet transform
- lossless compression
- image structure
- speaker identification
- signal processing
- speech synthesis
- feature vectors
- pattern recognition