Frequency-domain auditory suppression modelling (FASM) - a WDFT-based anthropomorphic noise-robust feature extraction algorithm for speech recognition.
Alexei V. IvanovMarek ParfieniukAlexander A. PetrovskyPublished in: INTERSPEECH (2005)
Keyphrases
- frequency domain
- speech recognition
- feature extraction
- noisy environments
- power spectrum
- spatial domain
- similarity measure
- fast fourier transform
- computational complexity
- low frequency
- image processing
- feature vectors
- fourier transform
- expectation maximization
- speech synthesis
- gaussian noise
- natural images
- neural network
- image compression
- language model
- denoising
- multiscale
- machine learning