Random discriminative projection-based feature selection for computational paralinguistics
- Global styles
- Apa
- Bibtex
- Chicago Fullnote
- Help
Abstract
Computational paralinguistics deals with the underlying meaning of the verbalmessages. Understanding the meaning of verbal messages provides interpreting spokencontent and behaving accordingly like humans. It allows us to develop human likemachines. Hence, paralinguistic area is attracting increasing attention for research.Paralinguistic analysis involves extracting features from raw speech data, chunking,selecting relevant features and training the model. In this thesis, the focus is on thefeature selection step. Feature selection aims at finding a relevant and necessary set offeatures to train generalizable models. The main challenge for feature selection methodsis the greedy-search nature of them. One major motivation for this study to develop anefficient feature selection technique is the success of a recently developed discriminativeprojection based feature selection method. Here, the method is enhanced by applyingthe power of stochasticity to overcome traps in local minimum while reducing thecomputational complexity. The proposed approach assigns weights both to groupsand to features individually in many randomly selected contexts and then combinesthem for a final ranking. The efficacy of the proposed method is shown in two recentchallenge corpora to detect level of depression severity and conflict.
Collections