Syntactic learning for ESEDA.1, a tool for enhanced speech emotion detection and analysis

This paper presents the second version of ESEDA, a speech emotion recognition tool. The current version of the tool has a number of novel capabilities as compared to the first version ESEDA.0 and other speech emotion recognition systems: firstly, it incorporates a novel classification method TGI+, and secondly, it includes a module for high level feature extraction. These two functionalities have the same underlying idea: calculate how consistent a given sample is with a set of rules derived for each recognition class, which is measured with edit distances to tree automata representing recognition classes. Then the sample is either classified with a decision tree working on these distance to automaton features (as in the case of TGI+) or we do early fusion of the low-level and the high-level features before the classification step (as in the case of high-level features).