Using unlabeled data to improve classification of emotional states in human computer interaction.
Martin SchelsMarkus KächeleMichael GlodekDavid HrabalSteffen WalterFriedhelm SchwenkerPublished in: J. Multimodal User Interfaces (2014)
Keyphrases
- human computer interaction
- unlabeled data
- improve the classification accuracy
- emotion recognition
- labeled data
- semi supervised learning
- semi supervised classification
- co training
- labeled and unlabeled data
- class labels
- semi supervised
- supervised learning
- text classification
- training set
- supervised learning algorithms
- classification accuracy
- emotional state
- active learning
- training data
- label information
- user interface
- human computer
- decision boundary
- labeled examples
- learning algorithm
- training examples
- labeled training data
- unsupervised learning
- data points
- semi supervised learning algorithms
- affect recognition
- decision trees
- labeled instances
- training samples
- facial expressions
- feature selection
- machine learning
- physiological signals
- affective computing
- select relevant features
- instance selection
- multi modal
- feature set
- support vector machine
- prior knowledge
- feature space
- natural language
- video sequences
- information retrieval