INTERSPEECH 2020 Computational Paralinguistics Challenges

Utrecht researchers win two out of three grand challenges in computational paralinguistics

Researchers from the Social and Affective Computing group have won two of the three Interspeech Computational Paralinguistics grand challenges this year, out of 65 competing teams. The challenges deal with paralinguistics: the parts of spoken communication that are not about words, but about other aspects such as pitch, volume, and intonation.

The teams from Utrecht, both led by Heysem Kaya (Computer Science), won the first place in two categories: breathing detection from speech and elderly emotion estimation from speech.  The Utrecht researchers collaborated with research groups from Russia and Germany.

The Breathing Sub-Challenge Prize was awarded to Maxim Markitantov, Denis Dresvyanskiy, Danila Mamontov, Heysem Kaya, Wolfgang Minker, and Alexey Karpov, for their paper Ensembling End-to-End Deep Models for Computational Paralinguistics Tasks.

The Elderly Emotion Sub-Challenge Prize was awarded to Gizem Sogancioglu, Oxana Verkholyak, Heysem Kaya, Dmitrii Fedotov, Tobias Cadee, Albert Ali Salah, and Alexey Karpov, for their paper Is Everything Fine, Grandma? Acoustic and Linguistic Modeling for Robust Elderly Speech Emotion Recognition.