Recognising realistic emotions and affect in speech: State of the art and lessons learnt from the first challenge

By Björn Schuller, Anton Batliner, Stefan Steidl, Dino Seppi. Speech Communication, 2011

788 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Experimental and Cognitive Psychology, Signal Processing

View Paper: https://www.semanticscholar.org/paper/9c79aeea15e95b257e1d7cadb6c4ef7b5a733d68

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Experimental and Cognitive Psychology

In Signal Processing

Citations in the corpus

Features and classifiers for emotion recognition from speech: a survey from 2000 to 2011 (2012, background): “A detailed analysis on the Emotion Challenge can be found in Schuller et al. (2011) .”
A review of depression and suicide risk assessment using speech analysis (2015, background): “…and long-term information is captured either statically through the use of utterance level statistics/functionals or dynamically through frame-based delta (D) and delta–delta (DD) coefficients, reflecting differences between the adjacent vectors’ feature coefficients (Schuller et al., 2013, 2011).”
The INTERSPEECH 2016 Computational Paralinguistics Challenge: Deception, Sincerity & Native Language (2016, methodology): “In this INTERSPEECH 2016 COMPUTATIONAL PARALINGUISTICS CHALLENGE (COMPARE) – the eighth since 2009 [1], we address three new problems within the field of Computational Paralinguistics [2] in a challenge setting as will be outlined one by one in the following.”
Speech emotion recognition using deep neural network and extreme learning machine (2014, background): “Speech emotion recognition is a very challenging task of which extracting effective emotional features is an open question [1, 2].”
Introducing the RECOLA multimodal corpus of remote collaborative and affective interactions (2013, background): “However, this approach was criticized for including (non-realistic) forced traits of emotion, which are claimed to be much more subtle when the emotion arises from a real-life context [16].”
A survey of multimodal sentiment analysis (2017, background): “In the meanwhile, a rich body of literature has been established, including a range of recent surveys such as [84].”
A Survey on perceived speaker traits: Personality, likability, pathology, and the first challenge (2015, methodology): “In 2009 to 2012, challenges (Schuller et al., 2009, 2010, 2011b, 2012, 2013a,b) were organised at the INTERSPEECH conferences that featured several different aspects of paralinguistics: topics of interest were not what the speaker said, i. e., word recognition, or the semantics behind word…”
MMGCN: Multimodal Fusion via Deep Graph Convolution Network for Emotion Recognition in Conversation (2021, methodology): “The acoustic raw features are extracted using the OpenS-mile toolkit with IS10 conﬁguration (Schuller et al., 2011).”
The INTERSPEECH 2013 computational paralinguistics challenge: social signals, conflict, emotion, autism (2013, methodology): “With the INTERSPEECH’s 2009 Emotion Challenge [1], 2010 Paralinguistic Challenge [2], 2011 Speaker State Challenge [3], and the recent 2012 Speaker Trait Challenge [4] we organised challenges and official exchange fora as exist for many more ‘traditional’ speech tasks, comparable to the NIST evaluat”
Categorical and dimensional affect analysis in continuous input: Current trends and future directions (2013, background): “Details on ‘earlier’ databases, automatic analyzers and various related work can be found in [39,8,31], and [40].”
Efficient Emotion Recognition from Speech Using Deep Learning on Spectrograms (2017, methodology): “Traditional methods were based on short-time frame-level feature extraction, followed by utterance-level information extraction, and classification or regression as required [3-8].”
The INTERSPEECH 2020 Computational Paralinguistics Challenge: Elderly Emotion, Breathing & Masks (2020, methodology): “As evaluation measure, we employ: (1) in the ESC and the MSC, Unweighted Average Recall (UAR) as used since the first Challenge from 2009 [1], especially because it is more adequate for (unbalanced) multi-class classifications than Weighted Average Recall (i.”
Emotion recognition from spontaneous speech using Hidden Markov models with deep belief networks (2013, background): “5% absolute); see [7] for the complete list.”