The MERL/SRI system for the 3RD CHiME challenge using beamforming, robust feature extraction, and advanced speech recognition.
Takaaki HoriZhuo ChenHakan ErdoganJohn R. HersheyJonathan Le RouxVikramjit MitraShinji WatanabePublished in: ASRU (2015)
Keyphrases
- speech recognition
- feature extraction
- noisy environments
- pattern recognition
- speaker identification
- hidden markov models
- language model
- speech enhancement
- speech recognizer
- cepstral coefficients
- speech signal
- automatic speech recognition
- frequency domain
- speech processing
- speech synthesis
- speech recognizers
- speech understanding
- speech recognition technology
- feature selection
- speech recognition systems
- principal component analysis
- image processing
- isolated word
- broadcast news
- background noise
- image classification
- feature vectors
- digit recognition
- feature space
- bayesian networks
- face recognition
- computer vision