DNN-based speech bandwidth expansion and its application to adding high-frequency missing features for automatic speech recognition of narrowband speech.
Kehuang LiZhen HuangYong XuChin-Hui LeePublished in: INTERSPEECH (2015)
Keyphrases
- automatic speech recognition
- high frequency
- speech signal
- speech recognition
- low frequency
- speech segments
- broadcast news
- hidden markov models
- word error rate
- conversational speech
- noisy environments
- vocal tract
- spontaneous speech
- recognition errors
- speech corpus
- speaker identification
- wavelet transform
- speech sounds
- high resolution
- speech retrieval
- spoken words
- non stationary
- feature extraction
- speech recognition systems
- word recognition
- image processing
- subband
- discrete wavelet transform
- feature vectors
- acoustic features
- wavelet coefficients
- high frequency components
- phoneme recognition
- low pass
- image segmentation
- computer vision
- machine learning
- speaker adaptation
- high quality
- multiscale
- probabilistic model
- feature set
- language model
- image compression
- frequency band