Spatial histogram equalization of complex-valued acoustic spectra in modulation domain for noise-robust speech recognition.
Hsin-Ju HsiehBerlin ChenJeih-weih HungPublished in: APSIPA (2014)
Keyphrases
- speech recognition
- noisy environments
- complex valued
- noisy speech
- histogram equalization
- speech recognition systems
- speaker independent
- speech recognizers
- hidden markov models
- real valued
- image enhancement
- language model
- speech signal
- speech recognizer
- pattern recognition
- background noise
- automatic speech recognition
- contrast enhancement
- acoustic models
- noise reduction
- recurrent neural networks
- speech synthesis
- speaker identification
- gray scale
- speech enhancement
- acoustic features
- activation function
- multiscale
- spatial information
- principal component analysis
- computer vision
- denoising
- bayesian networks
- gray level
- blind source separation
- neural network
- signal to noise ratio