Overlapped Speech Detection in Meeting Using Cross-Channel Spectral Subtraction and Spectrum Similarity.
Ryo YokoyamaYu NasuKoichi ShinodaKoji IwanoPublished in: INTERSPEECH (2012)
Keyphrases
- spectral subtraction
- speech signal
- speech recognition
- signal to noise ratio
- cross channel
- hybrid method
- automatic speech recognition
- noisy environments
- speech enhancement
- vocal tract
- similarity measure
- background noise
- power spectrum
- noise reduction
- hidden markov models
- speech synthesis
- noise model
- non stationary
- multiscale
- single channel
- linear prediction
- free riding
- hybrid algorithm
- natural images