Boosting Speech/Non-speech Classification Using Averaged Mel-Frequency Cepstrum Coefficients Features.
Ziyou XiongThomas S. HuangPublished in: IEEE Pacific Rim Conference on Multimedia (2002)
Keyphrases
- cepstral coefficients
- linear predictive coding
- feature set
- speech recognition
- speech signal
- mel frequency cepstral coefficients
- feature vectors
- classification accuracy
- feature selection
- linear prediction
- pattern recognition
- spectral features
- emotion classification
- automatic speech recognition
- audio signal
- feature space
- classification models
- speaker identification
- feature extraction
- classification method
- speaker recognition
- hidden markov models
- visual speech
- lexical features
- extracting features
- acoustic signals
- image classification
- classification process
- weak learners
- feature selection algorithms
- benchmark datasets
- extracted features
- machine learning methods
- acoustic features
- support vector machine svm
- class labels
- gaussian mixture model
- audio features
- emotion recognition
- decision trees
- feature values
- base classifiers
- machine learning
- training set
- wavelet coefficients
- support vector
- multiresolution
- linear combination
- learning algorithm
- text to speech
- discriminative classifiers
- svm classifier
- cost sensitive
- noisy environments
- discriminative features
- weak classifiers