English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools annual

Global: Zendy Free Plan

Global: Zendy Tools monthly

Global: Zendy Plus monthly

The automatic analysis of speech to detect affectivestates may improve the way users interact with electronicdevices. However, the analysis only at the acoustic level could benot enough to determine the emotion of a user in a realisticscenario. In this paper we analyzed the spontaneous speechrecordings of the FAU Aibo Corpus at the acoustic and linguisticlevels to extract two sets of features. The acoustic set was reducedby a greedy procedure selecting the most relevant features tooptimize the learning stage. We compared two versions of thisgreedy selection algorithm by performing the search of therelevant features forwards and backwards. We experimentedwith three classification approaches: Naive-Bayes, a supportvector machine and a logistic model tree, and two fusion schemes:decision-level fusion, merging the hard-decisions of the acousticand linguistic classifiers by means of a decision tree; and featurelevelfusion, concatenating both sets of features before thelearning stage. Despite the low performance achieved by thelinguistic data, a dramatic improvement was achieved after itscombination with the acoustic information, improving the resultsachieved by this second modality on its own. The results achievedby the classifiers using the parameters merged at feature leveloutperformed the classification results of the decision-level fusionscheme, despite the simplicity of the scheme. Moreover, theextremely reduced set of acoustic features obtained by the greedyforward search selection algorithm improved the results providedby the full set.

Comparative Study on Feature Selection and Fusion Schemes for Emotion Recognition from Speech