The INTERSPEECH 2020 Computational Paralinguistics Challenge: Elderly Emotion, Breathing &amp; Masks

Schuller, Björn W.; Batliner, Anton; Bergler, Christian; Meßner, Eva-Maria; Hamilton, Antonia F. de C.; Amiriparian, Shahin; Baird, Alice; Rizos, Georgios; Schmitt, Maximilian; Stappen, Lukas; Baumeister, Harald; MacIntyre, Alexis Deighton; Hantke, Simone

doi:10.21437/interspeech.2020-0032

Cited by 86 publications

(133 citation statements)

References 26 publications

(10 reference statements)

Supporting

Mentioning

130

Contrasting

Order By: Relevance

“…[±] A surge in machine learning research, has come from international challenges (Schuller et al, 2013 ; Ringeval et al, 2019 )—driving improvements in accuracy across multiple machine learning domains (Meer et al, 2000 ). However, this fast-paced environment often leaves less time for interpreting how particular features may have explicitly impacted a result, or for an explanation of a models decision-making process.…”

Section: Methodology: Ethical Data Considerationsmentioning

confidence: 99%

Considerations for a More Ethical Approach to Data in AI: On Data Representation and Infrastructure

Baird

Schuller

2020

Front. Big Data

Self Cite

View full text Add to dashboard Cite

Data shapes the development of Artificial Intelligence (AI) as we currently know it, and for many years centralized networking infrastructures have dominated both the sourcing and subsequent use of such data. Research suggests that centralized approaches result in poor representation, and as AI is now integrated more in daily life, there is a need for efforts to improve on this. The AI research community has begun to explore managing data infrastructures more democratically, finding that decentralized networking allows for more transparency which can alleviate core ethical concerns, such as selection-bias. With this in mind, herein, we present a mini-survey framed around data representation and data infrastructures in AI. We outline four key considerations (auditing, benchmarking, confidence and trust, explainability and interpretability) as they pertain to data-driven AI, and propose that reflection of them, along with improved interdisciplinary discussion may aid the mitigation of data-based AI ethical concerns, and ultimately improve individual wellbeing when interacting with AI.

show abstract

Section: Methodology: Ethical Data Considerationsmentioning

confidence: 99%

Considerations for a More Ethical Approach to Data in AI: On Data Representation and Infrastructure

Baird

Schuller

2020

Front. Big Data

Self Cite

View full text Add to dashboard Cite

show abstract

“…In the ML experiments, we employ the ComParE feature set (Schuller et al 2013), comprising 6373 acoustic features (Eyben et al 2015) computed by applying statistical functions to 65 Low-Level Descriptors (LLDs), extracted by the OPENSMILE feature extractor (Eyben et al 2010), and a Support Vector Machine (SVM) classifier with a linear kernel from the open-source toolkit LIBLINEAR (Fan et al 2008). Even though Deep Neural Networks (DNNs) are prevalent nowadays for ML tasks, in affective computing research their performance is not yet superior to rather classic ML procedures such as SVMs.…”

Section: Methodsmentioning

confidence: 99%

“…Even though Deep Neural Networks (DNNs) are prevalent nowadays for ML tasks, in affective computing research their performance is not yet superior to rather classic ML procedures such as SVMs. This can be seen in the series of Interspeech Challenges, from Schuller et al (2013) to Schuller et al (2018) and might simply be due to the sparse data problem: DNNs need very large databases; such databases do not exist for emotion modelling. Therefore, we chose an SVM classifier as it has only few hyperparameters, compared to recent deep learning approaches, and thus gives more reliable results in terms of robustness during training; our approach is more focused on understanding and less on optimising classification.…”

Section: Methodsmentioning

confidence: 99%

DEMoS: an Italian emotional speech corpus

Parada-Cabaleiro

Costantini

Batliner

et al. 2019

Lang Resources & Evaluation

Self Cite

View full text Add to dashboard Cite

We present DEMoS (Database of Elicited Mood in Speech), a new, large database with Italian emotional speech: 68 speakers, some 9 k speech samples. As Italian is under-represented in speech emotion research, for a comparison with the state-of-theart, we model the 'big 6 emotions' and guilt. Besides making available this database for research, our contribution is three-fold: First, we employ a variety of mood induction procedures, whose combinations are especially tailored for specific emotions. Second, we use combinations of selection procedures such as an alexithymia test and self-and external assessment, obtaining 1,5 k (proto-) typical samples; these were used in a perception test (86 native Italian subjects, categorical identification and dimensional rating). Third, machine learning techniques-based on standardised brute-forced openSMILE ComParE features and support vector machine classifiers-were applied to assess how emotional typicality and sample size might impact machine learning efficiency. Our results are three-fold as well: First, we show that appropriate induction techniques ensure the collection of valid samples, whereas the type of self-assessment employed turned out not to be a meaningful measurement. Second, emotional typicality-which shows up in an acoustic analysis of prosodic main features-in contrast to sample size is not an essential feature for successfully training machine learning models. Third, the perceptual findings demonstrate that the confusion patterns mostly relate to cultural rules and to ambiguous emotions.

show abstract

“…Extracting both hand-crafted acoustic features and deep representations of the audio signal on the frame-level of all sessions. We decided to extract both acoustic and D eep S pectrum features, due to their previous performance and proven ability in capturing characteristics of speech (Schuller et al, 2013 ; Amiriparian et al, 2016 , 2018 ; Eyben, 2016 ). Both feature sets are different in their nature; C om P ar E is a hand-crafted, expert-designed feature set which can cover time-dependent frame-level information for the input signals, and D eep S pectrum is based on the spectrograms of audio signals, focusing mostly on the time-frequency properties of the speech.…”

Section: Dataset and Featuresmentioning

confidence: 99%

Synchronization in Interpersonal Speech

et al. 2019

Self Cite

View full text Add to dashboard Cite

During both positive and negative dyadic exchanges, individuals will often unconsciously imitate their partner. A substantial amount of research has been made on this phenomenon, and such studies have shown that synchronization between communication partners can improve interpersonal relationships. Automatic computational approaches for recognizing synchrony are still in their infancy. In this study, we extend on previous work in which we applied a novel method utilizing hand-crafted low-level acoustic descriptors and autoencoders (AEs) to analyse synchrony in the speech domain. For this purpose, a database consisting of 394 in-the-wild speakers from six different cultures, is used. For each speaker in the dyadic exchange, two AEs are implemented. Post the training phase, the acoustic features for one of the speakers is tested using the AE trained on their dyadic partner. In this same way, we also explore the benefits that deep representations from audio may have, implementing the state-of-the-art Deep Spectrum toolkit. For all speakers at varied time-points during their interaction, the calculation of reconstruction error from the AE trained on their respective dyadic partner is made. The results obtained from this acoustic analysis are then compared with the linguistic experiments based on word counts and word embeddings generated by our word2vec approach. The results demonstrate that there is a degree of synchrony during all interactions. We also find that, this degree varies across the 6 cultures found in the investigated database. These findings are further substantiated through the use of 4,096 dimensional Deep Spectrum features.

show abstract

The INTERSPEECH 2020 Computational Paralinguistics Challenge: Elderly Emotion, Breathing & Masks

Cited by 86 publications

References 26 publications

Considerations for a More Ethical Approach to Data in AI: On Data Representation and Infrastructure

Considerations for a More Ethical Approach to Data in AI: On Data Representation and Infrastructure

DEMoS: an Italian emotional speech corpus

Synchronization in Interpersonal Speech

Contact Info

Product

Resources

About