A hybrid SVM/HMM acoustic modeling approach to automatic speech recognition.
Jan StadermannGerhard RigollPublished in: INTERSPEECH (2004)
Keyphrases
- automatic speech recognition
- speech recognition
- hidden markov models
- acoustic features
- speech recognizers
- speech signal
- speech sounds
- acoustic models
- speech segments
- word error rate
- speaker independent
- support vector machine svm
- spontaneous speech
- discriminative training
- speech recognition systems
- support vector
- broadcast news
- speech recognizer
- conversational speech
- speech retrieval
- support vector machine
- spoken words
- speaker adaptation
- noisy environments
- speech synthesis
- feature vectors
- recognition errors
- speech corpus
- machine learning
- sound source
- knn
- computer vision
- word recognition
- non stationary
- pattern recognition
- feature selection
- neural network
- formant frequencies