Ultra-low-power voice-activity-detector through context- and resource-cost-aware feature selection in decision trees.
Steven LauwereinsWannes MeertJort F. GemmekeMarian VerhelstPublished in: MLSP (2014)
Keyphrases
- decision trees
- feature selection
- naive bayes
- text categorization
- bayes classifier
- support vector
- feature set
- machine learning
- constructive induction
- feature construction
- information gain
- feature subset
- high dimensionality
- resource allocation
- predictive accuracy
- decision rules
- multi task
- total cost
- selected features
- detection method
- dimensionality reduction
- real time
- resource availability