DREAMER: A Database for Emotion Recognition Through EEG and ECG Signals From Wireless Low-cost Off-the-Shelf Devices

Abstract

In this paper, we present DREAMER, a multimodal database consisting of electroencephalogram (EEG) and electrocardiogram (ECG) signals recorded during affect elicitation by means of audio-visual stimuli. Signals from 23 participants were recorded along with the participants self-assessment of their affective state after each stimuli, in terms of valence, arousal, and dominance. All the signals were captured using portable, wearable, wireless, low-cost, and off-the-shelf equipment that has the potential to allow the use of affective computing methods in everyday applications. A baseline for participant-wise affect recognition using EEG and ECG-based features, as well as their fusion, was established through supervised classification experiments using support vector machines (SVMs). The self-assessment of the participants was evaluated through comparison with the self-assessments from another study using the same audio-visual stimuli. Classification results for valence, arousal, and dominance of the proposed database are comparable to the ones achieved for other databases that use nonportable, expensive, medical grade devices. These results indicate the prospects of using low-cost devices for affect recognition applications. The proposed database will be made publicly available in order to allow researchers to achieve a more thorough evaluation of the suitability of these capturing devices for affect recognition applications.

A Bayesian Deep Learning Framework for End-To-End Prediction of Emotion From Heartbeat

“…The output from both streams is then concatenated before passing through a dense layer to output a regression estimate for valence,ŷ. higher accuracy across both datasets than previously reported [40], [43] (Table II).…”

“…This allowed us to exploit existing high-quality ECG datasets for this study. We developed an end-to-end neural network capable of modelling temporal structure in the IBI time series, which outperformed previous classifiers on this task [40], [43].…”

“…Instead, they employ 'static' classifiers that process global features from the input time series (or for a small number of segments). Such approaches include Naive Bayes (NB) [40], [41], linear discriminant analysis (LDA) [42], and support vector machine (SVM) [43], [44], [45]. A summary can be found in Table I. A number of studies have sought to model temporal information within EEG signals, using hidden Markov models [46], Gaussian Process models [47], continuous conditional random fields [48], and long short-term memory (LSTM) neural networks [49].…”

“…We applied our Bayesian deep learning framework for endto-end prediction of emotion using heartbeat (IBI) data from two established datasets − AMIGOS [40] and DREAMER [43]. In this section, we provide details on these data, which were chosen for their quality, clarity, and close comparability.…”

End-To-End Prediction of Emotion from Heartbeat Data Collected by a Consumer Fitness Tracker

“…Additionally, those studies that have explored unimodal heartbeat models for emotion detection tend to ignore temporal structures of the signal. Instead, they use 'static' classification methods that analyse global features of the input time-series, such as Naive Bayes (NB), [16], [18], linear discrimant analysis (LDA) [22], and support vector machine (SVM) [15], [19], [21]. A summary of these studies can be found in Table I.…”