News

[12/18/2018] The annotation of the laughter events are now publicly available; cf. paper at ICMI'18.

[04/03/2018] The RECOLA database will be used for the ACM MM AVEC 2018 Challenge.

[01/09/2016] Special section on Multimedia Computing and Applications of Socio-Affective Behaviors in the Wild, ACM TOMM.

[04/03/2016] The RECOLA database will be used for the ACM MM AVEC 2016 Challenge.

[10/12/2015] Contributions of the AV+EC 2015 Challenge have been added to the publication list.

[14/02/2015] The RECOLA database will be used for the ACM MM AV+EC 2015 Challenge.

[11/02/2015] Time stamps of the features and event probabilities have been added in the data set.

[02/02/2015] Multimodal feature sets have been added to the repository.

[15/11/2013] The RECOLA database is available for download!


Hall of fame on the RECOLA database

We list in the following table studies that performed machine learning experiments and reported performance, in terms of Concordance Correlation Coefficient, on the test partition of the publicly available version of the RECOLA database, which includes 27 subjects, with 9 subjects in each training, development, and (blinded) test partition.

Models
  • BLSTM: Bidirectional Long Short-Term Memory
  • CNN: Convolutional Neural Network
  • ELM: Extreme Learning Machine
  • ESN: Echo State Network
  • FCNN: Fully-Convolutional Neural Network
  • GRU: Gated Recurrent Unit
  • LinReg: Linear Regression
  • LSTM: Long Short-Term Memory
  • RVM: Relevance Vector Machine
  • SVM: Support Vector Machine
 Audio Features
  • BoAW: Bag of Audio Words
  • DeepSpec: Deep Spectrum
  • eGeMAPS: extended Geneva Minimalistic Acoustic Parameter Set
  • LFPC: Log Frequency Power Coefficient
  • LPC: Linear Predictive Coefficient
  • MFB: Mel Filter Bank
  • MFCC: Mel Frequency Cepstral Coefficient
Ref. Year Arousal (audio) Valence (audio) Arousal (mixed or other modalities) Valence (mixed or other modalities) Models Audio Features
AlBadawy, E. A., et al. 2018 0.697 0.555 0.699 0.622 BLSTM MFB
Khorram, S, et al. 2018 0.684 0.502 - - CNN MFB
Khorram, S, et al. 2019 0.688 0.492 - - CNN MFB
Le, D, et al. 2017 0.682 0.460 - - BLSTM MFB
Zhang, Z, et al. 2018 0.699 0.449 0.766 0.660 GRU eGeMAPS
Han, J, et al. 2019 0.644 0.439 0.475 0.529 GRU eGeMAPS
Han, J, et al. 2019 0.611 0.395 0.413 0.529 GRU eGeMAPS
Valstar, M, et al. 2016 0.648 0.375 0.683 0.639 SVM eGeMAPS
Ringeval, F, et al. 2018 0.651 0.346 0.657 0.515 SVM eGeMAPS - BoAW - DeepSpec
Han, J, et al. 2017 0.664 0.303 0.685 0.554 BLSTM - SVM eGeMAPS
Tzirakis, P, et al. 2018 0.669 0.286 0.672 0.521 CNN - GRU Raw
Zhang, Z, et al. 2016 0.732 0.278 - - LSTM MFCC
Ringeval, F, et al. 2015 0.228 0.068 0.444 0.382 BLSTM - SVM - FCNN eGeMAPS
Brady, K, et al. 2016 - - 0.770 0.687 SVM - FCNN - CNN - LSTM MFCC
Somandepalli, K, et al. 2016 - - 0.703 0.681 SVM MFCC - eGeMAPS
Chen, S, et al. 2016 - - - 0.664 LSTM eGeMAPS
Sun, B, et al. 2016 - - 0.683 0.642 SVM - CNN - LinReg eGeMAPS
Amirian, M, et al. 2016 - - 0.762 0.624 SVM - CNN - LinReg ESN - LinReg - LFPC
Chao, L, et al. 2015 - - 0.716 0.618 LSTM eGeMAPS
He, L, et al. 2015 - - 0.747 0.609 BLSTM eGeMAPS
Povolny, F, et al. 2016 - - 0.719 0.596 LinReg eGeMAPS - Word2Vec
huang, Z, et al. 2015 - - 0.740 0.588 SVM eGeMAPS
Kaya, H, et al. 2017 - - 0.465 0.566 SVM LinReg - ELM - eGeMAPS
Huang, Z, et al. 2016 - - 0.770 0.545 RVM eGeMAPS
Amiriparian, S, et al. 2017 - - 0.714 0.536 SVM eGeMAPS
Khorrami, P, et al. 2017 - - - 0.506 CNN - RNN -
Popkova, A, et al. 2016 - - 0.660 0.504 LinReg - FCNN eGeMAPS
Milchevski, A, et al. 2015 - - 0.644 0.501 LinReg eGeMAPS
Kachele, M, et al. 2015 - - 0.620 0.479 FCNN MFCC - LPC - LFPC
Weber, R, et al. 2016 - - 0.682 0.468 SVM -
Keren, G, et al. 2017 - - 0.430 0.407 CNN - LSTM -
Manandhar, A, et al. 2016 - - 0.444 0.398 RVM eGeMAPS
Cardinal, P, et al. 2015 - - 0.293 0.290 FCNN - LinReg eGeMAPS

We strongly encourage researchers working on the RECOLA Database to inform us on their scientific publications: please send an email to diuf-recola[at]unifr.ch with the full citation of your paper in the core of the message and, if possible, with the PDF file attached. After checking, publications will be listed here to ease access to other researchers interested in this corpus, while facilitating its bibliography.