Frequency map selection using a RBFN-based classifier in the MVDR beamformer for speaker localization in reverberant rooms.
Daniele SalvatiCarlo DrioliGian Luca ForestiPublished in: INTERSPEECH (2015)
Keyphrases
- minimum variance
- map building
- training data
- selection criterion
- feature maps
- mobile robot
- loop closing
- selection scheme
- speech recognition
- speech signal
- multipath
- speaker identification
- learning algorithm
- support vector machine
- feature selection
- decision trees
- maximum a posteriori
- audio visual
- svm classifier
- rbf network
- multilayer perceptron
- classification method
- training set
- support vector
- radial basis function network
- radial basis function neural network
- robot localization
- pattern recognition
- feature space
- linear prediction
- selection algorithm
- multi layer perceptron
- linear classifiers
- classification algorithm