Compensating the Speech Features via Discrete Cosine Transform for Robust Speech Recognition (基於離散餘弦轉換之語音特徵的強健性補償法).
Hsin-Ju HsiehWen-Hsiang TuJeih-Weih HungPublished in: ROCLING (2011)
Keyphrases
- speech recognition
- discrete cosine transform
- speech recognition systems
- noisy environments
- speech synthesis
- speech signal
- speech processing
- hidden markov models
- cepstral coefficients
- automatic speech recognition
- speech recognition technology
- pattern recognition
- speech recognizer
- language model
- feature vectors
- feature set
- mel frequency cepstral coefficients
- multiresolution
- speaker identification
- classification accuracy
- speaker recognition
- feature extraction
- isolated word
- vocal tract
- machine learning
- transform domain
- image compression
- video sequences
- image processing
- computer vision
- dct coefficients
- filter bank
- computationally efficient
- speaker diarization
- probabilistic model
- feature space
- multiscale
- speech recognition errors