AUTHOR=Diemerling Hannes , Stresemann Leonie , Braun Tina , von Oertzen Timo 

TITLE=Implementing machine learning techniques for continuous emotion prediction from uniformly segmented voice recordings

JOURNAL=Frontiers in Psychology

VOLUME=Volume 15 - 2024

YEAR=2024

URL=https://www.frontiersin.org/journals/psychology/articles/10.3389/fpsyg.2024.1300996

DOI=10.3389/fpsyg.2024.1300996

ISSN=1664-1078

ABSTRACT=This article presents a novel approach to recognizing emotions from generated 1.5-second audio
recordings, utilizing 1510 unique samples sourced from two distinct databases in German
and English. Features extracted from these recordings are employed in Deep Neural Networks
(DNN), spectrogram features for Convolutional Neural Networks (CNN), and a hybrid model
combining both (C-DNN) for emotion prediction. The study underscores challenges in merging
datasets due to their heterogeneity, language disparities, and the intricacies of trimming
audio recordings. Nevertheless, the results demonstrate that these models achieve an accuracy
considerably above random guessing and are on par with human ratings. Given the shortness
of the used sample length, this method holds promise for estimating emotional shifts in continuous,
longer speech data.