Feature Extraction and Selection for Emotion Recognition from Electrodermal Activity

Shukla, Jainendra; Barreda-Ángeles, Miguel; Oliver, Joan Guix; Nandi, Gora Chand; Puig, Domènec

doi:10.1109/taffc.2019.2901673

Cited by 173 publications

(96 citation statements)

References 43 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In Table 13, two classes of emotions represent either high/low levels of valence or arousal, while four classes represent HVHA, HVLA, LVHA, and LVLA. For instance, the study [36] with statistical features and SVM using AMIGOS dataset results in 68.8% and 67% of accuracy for arousal and valence respectively, while another study [37] improved valence results by using SVM-RBF. The recognition results from [12,13,35,39] and attention-based LSTM-RNN study [40] improved with deep-learning algorithms for either AMIGOS or DREAMER dataset for two classes of emotions.…”

Section: Discussionmentioning

confidence: 99%

“…They reported a recognition rate of 67% and 68.8% for valence and arousal respectively using an SVM classifier. Another recent GSR-based framework [37] using AMIGOS dataset proposed temporal and spectral features with SVM (RBF kernel) to report recognition performance of 83.9% and 65% for valence and arousal, respectively. For the AMIGOS dataset, the significance of DNN can be explained by two similar studies, where one of the studies [34] reported 55.1% and 54.4% F1 scores for valence and arousal, respectively, using Gaussian Naive Bayes, while another study [38] reported 71% and 81% accuracy for valence and arousal, respectively, using convolutional neural networks.…”

Section: Related Workmentioning

confidence: 99%

“…[63] with their corresponding positioning of channels to a 1D topological vector of size 81. Each corresponding sample of 14 channels is mapped to 1D topological vector using the matching sequence of set of channels of AF3, F7, F3, FC5, T7, P7, O1, O2, P8, T8, FC6, F4, F8, AF4 to the set of indices of 4, 13,19,21,29,31,37,39,47,49,55,57,67,76 respectively while keeping all other indices of 1D topological vector as zero. After mapping, a feature matrix of 81×128 dimension is obtained as single EEG example of specific trial of a specific person.…”

Section: Z-score Normalizationmentioning

confidence: 99%

See 2 more Smart Citations

CNN and LSTM-Based Emotion Charting Using Physiological Signals

Dar

Akram

Khawaja

et al. 2020

Sensors

View full text Add to dashboard Cite

Novel trends in affective computing are based on reliable sources of physiological signals such as Electroencephalogram (EEG), Electrocardiogram (ECG), and Galvanic Skin Response (GSR). The use of these signals provides challenges of performance improvement within a broader set of emotion classes in a less constrained real-world environment. To overcome these challenges, we propose a computational framework of 2D Convolutional Neural Network (CNN) architecture for the arrangement of 14 channels of EEG, and a combination of Long Short-Term Memory (LSTM) and 1D-CNN architecture for ECG and GSR. Our approach is subject-independent and incorporates two publicly available datasets of DREAMER and AMIGOS with low-cost, wearable sensors to extract physiological signals suitable for real-world environments. The results outperform state-of-the-art approaches for classification into four classes, namely High Valence—High Arousal, High Valence—Low Arousal, Low Valence—High Arousal, and Low Valence—Low Arousal. Emotion elicitation average accuracy of 98.73% is achieved with ECG right-channel modality, 76.65% with EEG modality, and 63.67% with GSR modality for AMIGOS. The overall highest accuracy of 99.0% for the AMIGOS dataset and 90.8% for the DREAMER dataset is achieved with multi-modal fusion. A strong correlation between spectral- and hidden-layer feature analysis with classification performance suggests the efficacy of the proposed method for significant feature extraction and higher emotion elicitation performance to a broader context for less constrained environments.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Section: Z-score Normalizationmentioning

confidence: 99%

See 1 more Smart Citation

CNN and LSTM-Based Emotion Charting Using Physiological Signals

Dar

Akram

Khawaja

et al. 2020

Sensors

View full text Add to dashboard Cite

show abstract

“…Many physiological modalities and features have been evaluated for ER, namely Electroencephalography (EEG) [ 28 , 29 , 30 ], Electrocardiography (ECG) [ 31 , 32 , 33 ], Electrodermal Activity (EDA) [ 34 , 35 , 36 ], Respiration (RESP) [ 26 ], Blood Volume Pulse (BVP) [ 26 , 35 ] and Temperature (TEMP) [ 26 ]. Multi-modal approaches have prevailed; however, there is still no clear evidence of which feature combinations and physiological signals are the most relevant.…”

Section: State Of the Artmentioning

confidence: 99%

Emotion Assessment Using Feature Fusion and Decision Fusion Classification Based on Physiological Data: Are We There Yet?

Bota

Wang²,

Fred

et al. 2020

Sensors

View full text Add to dashboard Cite

Emotion recognition based on physiological data classification has been a topic of increasingly growing interest for more than a decade. However, there is a lack of systematic analysis in literature regarding the selection of classifiers to use, sensor modalities, features and range of expected accuracy, just to name a few limitations. In this work, we evaluate emotion in terms of low/high arousal and valence classification through Supervised Learning (SL), Decision Fusion (DF) and Feature Fusion (FF) techniques using multimodal physiological data, namely, Electrocardiography (ECG), Electrodermal Activity (EDA), Respiration (RESP), or Blood Volume Pulse (BVP). The main contribution of our work is a systematic study across five public datasets commonly used in the Emotion Recognition (ER) state-of-the-art, namely: (1) Classification performance analysis of ER benchmarking datasets in the arousal/valence space; (2) Summarising the ranges of the classification accuracy reported across the existing literature; (3) Characterising the results for diverse classifiers, sensor modalities and feature set combinations for ER using accuracy and F1-score; (4) Exploration of an extended feature set for each modality; (5) Systematic analysis of multimodal classification in DF and FF approaches. The experimental results showed that FF is the most competitive technique in terms of classification accuracy and computational complexity. We obtain superior or comparable results to those reported in the state-of-the-art for the selected datasets.

show abstract

“…In the first method, hand-crafted features can be extracted in the time domain, the frequency domain, the time-frequency domain, etc. [17]. After that, the hand-crafted features are fed into classifiers such as KNN [18] and SVM [19].…”

mentioning

confidence: 99%

A Systematic Exploration of Deep Neural Networks for EDA-Based Emotion Recognition

Sun

2020

Information

View full text Add to dashboard Cite

Subject-independent emotion recognition based on physiological signals has become a research hotspot. Previous research has proved that electrodermal activity (EDA) signals are an effective data resource for emotion recognition. Benefiting from their great representation ability, an increasing number of deep neural networks have been applied for emotion recognition, and they can be classified as a Convolutional Neural Network (CNN), a Recurrent Neural Network (RNN), or a combination of these (CNN+RNN). However, there has been no systematic research on the predictive power and configurations of different deep neural networks in this task. In this work, we systematically explore the configurations and performances of three adapted deep neural networks: ResNet, LSTM, and hybrid ResNet-LSTM. Our experiments use the subject-independent method to evaluate the three-class classification on the MAHNOB dataset. The results prove that the CNN model (ResNet) reaches a better accuracy and F1 score than the RNN model (LSTM) and the CNN+RNN model (hybrid ResNet-LSTM). Extensive comparisons also reveal that our three deep neural networks with EDA data outperform previous models with handcraft features on emotion recognition, which proves the great potential of the end-to-end DNN method.Information 2020, 11, 212 2 of 16 different subjects. Until now, researchers still haven't achieved a satisfying recognition accuracy [8][9][10][11]. Circumventing this problem, we make great attempts on subject-independent emotion recognition in this work.Many studies in the past years have focused on physiological signals, such as EEG [12,13], ECG [14,15], and EDA [9,10]. Compared with other physiological signals, EDA can be measured on the skin surfaces of hands and wrists in a non-invasive way. Benefiting from this easy and efficient acquisition method, the EDA-based emotion recognition algorithm has a broad application prospect in the development of sensors, Internet of Things (IoT), and intelligent wearable devices. Moreover, EDA is controlled by the autonomic nervous system, which corresponds to the arousal state of people [16]. On the other aspect, EDA has fewer channels and data, compared to EEG signals. Therefore, making full use of the limited EDA data is a great challenge in the field of EDA-based emotion recognition.The methods of physiological-signal-based emotion recognition can be classified into two types based on how features are extracted: Hand-crafted feature selection and auto feature extraction. In the first method, hand-crafted features can be extracted in the time domain, the frequency domain, the time-frequency domain, etc. [17]. After that, the hand-crafted features are fed into classifiers such as KNN [18] and SVM [19]. However, the formula of feature extraction is established manually, which means that it cannot extract other unknown important features. The method based on auto feature extraction can solve the defect of hand-crafted feature selection. It utilize deep learning networks, which can extract implicit and c...

show abstract

Feature Extraction and Selection for Emotion Recognition from Electrodermal Activity

Cited by 173 publications

References 43 publications

CNN and LSTM-Based Emotion Charting Using Physiological Signals

CNN and LSTM-Based Emotion Charting Using Physiological Signals

Emotion Assessment Using Feature Fusion and Decision Fusion Classification Based on Physiological Data: Are We There Yet?

A Systematic Exploration of Deep Neural Networks for EDA-Based Emotion Recognition

Contact Info

Product

Resources

About