Examining Human-Horse Interaction by Means of Affect Recognition via Physiological Signals

Abstract

For some time, equine-assisted therapy (EAT), i.e., the use of horse-related activities for therapeutic reasons, has been recognised as a useful approach in the treatment of many mental health issues such as post-traumatic stress disorder (PTSD), depression, and anxiety. However, despite the interest in EAT, few scientific studies have focused on understanding the complex emotional response that horses seem to elicit in human riders and handlers. In this work, the potential use of affect recognition techniques based on physiological signals is examined for the task of assessing the interaction between humans and horses in terms of the emotional response of the humans to this interaction. Electroencephalography (EEG), electrocardiography (ECG), and electromyography (EMG) signals were captured from humans interacting with horses, and machine learning techniques were applied in order to predict the self-reported emotional states of the human subjects in terms of valence and arousal. Supervised classification experiments demonstrated the potential of this approach for affect recognition during human-horse interaction, reaching an F1-score of 78.27% for valence and 65.49% for arousal.

On the use of ECG and EMG Signals for Question Difficulty Level Prediction in the Context of Intelligent Tutoring Systems

“…The SHIMMER ECG and EMG devices ( Fig. 1) have been previously used successfully for affective computing studies [18], [21] and were selected due to their small form-factor, low weight, portability, and wireless characteristics that limited the discomfort of the participants due to the presence of equipment. A typical laptop computer (Intel i5-5300U @2.3 GHz CPU, 4.0 GB of DDR3 RAM, Windows ® 10 OS) was used for signal recording and monitoring.…”

“…2) ECG-based features: Eighty four features, related to the raw ECG signal and the heart rate variability (HRV), that have been commonly used in affective computing studies (e.g. [17], [18], [21]) were extracted from the ECG signals using the AuBT [27]. The features included the mean, median, standard deviation, minima, maxima, and range of the HRV histogram, the number of intervals with latency > 50 ms from HRV, the power spectral density ( 3) Fusion of ECG and EMG features: Future fusion was also examined since it has been shown to lead to increased performance in affective computing studies [19], [28].…”

“…The second experiment focused on creating single-participant personalised models by creating a separate classification model for each participant. To simplify the examined problems, both problems were converted to binary classification problems, as commonly practised in affective computing studies [17]- [19], [21]. To achieve this, samples annotated as Very easy or Easy were classed as referring to Low difficulty, and samples annotated as Hard or Very hard were classed as referring to High difficulty.…”

“…Affective Computing refers to computing that relates to emotions [15] and emotion/affect recognition is one of its main focuses [16]. Within this field, multiple studies [17]- [21] explored the relation between physiological signals, such as electroencephalography (EEG), electrocardiography (ECG), electromyography (EMG), electrodermal activity (EDA), and others, and the affective state of humans in terms of the Valence and Arousal dimensions, as defined by Russel in his Circumplex Model of Affect [22].…”

ECG-based affective computing for difficulty level prediction in Intelligent Tutoring Systems

“…Researchers have been trying to observe and model the behaviour of the human brain through diverse areas of research, such as psychology [11], [12], medical imaging, such as functional Magnetic Resonance Imaging (fMRI) [13]), or by analysing bio-signals, such as physiological signals [14], [15]. Various studies [14]- [18] have shown that there is correlation between physiological signals and the emotions felt by an individual, as defined in Russel's Circumplex Model of Affect [19]. These studies relied on pattern recognition and machine learning techniques in order to map features extracted from physiological signals to a relevant emotional state in terms of the Valence and Arousal dimensions of an emotion.…”