Design and evaluation of reduced-size feature sets for the assessment of sincerity in speech

ENRIQUE M. ALBORNOZ; CÉSAR E. MARTINEZ

Budapest

Congreso; 17th IEEE International Symposium on Computational Intelligence and Informatics; 2016

Óbuda University, Budapest - IEEE

The recognition of states and traits of speakers isa significant issue to investigate, to be able to achieve moreuseful interactive systems. The sincerity of a speaker is a relevantparalinguistic phenomenon, which has not received too muchattention from the affective computing community. In this work,we tackle the problem using novel feature sets proposed foremotion recognition. In addition, bioinspired features (using anauditory signal representation) and other spectral features arealso evaluated. Finally, diverse combinations of these reduced-sizefeature sets are built. The provided standard, complete setwith 6373 features is used for comparison purposes. Results showthat using the combination of the proposed representations andstate-of-art features, it is possible to obtain very small feature sets(less than 3% of the original size) that get comparable correlationmeasure with respect to the baseline.