Emotion Recognition: A Statistical Learning Approach

Projektinformationen

LearningEmotions

ID Finanzhilfevereinbarung: 887283

DOI

10.3030/887283

Projekt abgeschlossen

EK-Unterschriftsdatum 27 Februar 2020

Startdatum 4 Januar 2021

Enddatum 3 Januar 2023

Finanziert unter

• EXCELLENT SCIENCE - Marie Skłodowska-Curie Actions

Gesamtkosten

€ 212 933,76

EU-Beitrag

€ 212 933,76

Koordiniert durch

THE UNIVERSITY OF READING
United Kingdom

Periodic Reporting for period 1 - LearningEmotions (Emotion Recognition: A Statistical Learning Approach)

Berichtszeitraum: 2021-01-04 bis 2023-01-03

Zusammenfassung vom Kontext und den Gesamtzielen des Projekts

The LearningEmotions project attempted to address two questions: (i) Why are some emotions harder to recognise than others? and (ii) Why do some individuals, such as autistic individuals, have more difficulty recognising emotions generally? Specifically, I examined whether the two questions may actually arise from a common source, that is, difficulties with learning patterns and their associations (i.e. statistical learning).

Examining these two questions is key to further our understanding of emotion processing and may provide valuable insight in designing interventions for autistic individuals.

Arbeit, die ab Beginn des Projekts bis zum Ende des durch den Bericht erfassten Berichtszeitraums geleistet wurde, und die wichtigsten bis dahin erzielten Ergebnisse

I compared various computational measures to quantify variability in audio-video recordings of emotion expression and found one measure in particular best fit human data. Essentially, this measure is computed by reducing the audio-visual clip of the expression into a matrix and determining its “center”. We do this for all the clips of the same emotion and obtain the center of the centers.

Across various tasks, we found that autistic individuals showed comparable statistical learning as non-autistic individuals when the cue-outcome associations are strong, but groups differences emerged when the associations are weak, with poorer performance among autistic individuals. Moreover, autistic individuals were more likely to incorporate feedback — even those that should be ignored because it is ‘noise’– to their subsequent decision, which may lead to differences in their performance on the weak cue-outcome associations.

In one emotion perception study conducted as part of the project, we found that autistic individuals who frequently encountered others wearing face coverings showed improvement in recognising emotions from just the eyes over a 10-month period during which there was a face covering mandate. This suggests that long-term passive exposure can modify our emotion perception ability. In another study, we found that autistic individuals did not differ from non-autistic individuals in recognising spoken and sung emotions, but autistic individuals were slower for some emotions, particularly those that involve the Theory of Mind or the ability to understand one’s mental states to infer their feelings or thoughts, which have been reported to be atypical among autistic individuals.

Fortschritte, die über den aktuellen Stand der Technik hinausgehen und voraussichtliche potenzielle Auswirkungen (einschließlich der bis dato erzielten sozioökonomischen Auswirkungen und weiter gefassten gesellschaftlichen Auswirkungen des Projekts)

The results of the project on statistical learning and emotion recognition among autistic individuals deepen our understanding of autism and have potential implications for developing appropriate interventions in the future.

Learning cue-outcome associations and results of a probabilistic learning task