Minimum trajectory error training for deep neural networks, combined with stacked bottleneck features.
Zhizheng WuSimon KingPublished in: INTERSPEECH (2015)
Keyphrases
- neural network
- training process
- feature vectors
- feature set
- feature extraction
- pattern recognition
- low level
- genetic algorithm
- feed forward neural networks
- training samples
- prior knowledge
- feature space
- co occurrence
- fuzzy logic
- training examples
- false positives
- error bounds
- feed forward
- neural nets
- extracted features
- discriminative classifiers