Spatial–Temporal Recurrent Neural Network for Emotion Recognition

Zhang, Tong; Zheng, Wenming; Cui, Zhen; Zong, Yuan; Li, Yang

doi:10.1109/tcyb.2017.2788081

Cited by 430 publications

(194 citation statements)

References 36 publications

Supporting

Mentioning

193

Contrasting

Order By: Relevance

“…2). All achieved accuracy improvements over handcrafted approaches, but a lack of training data was also mentioned [145], [146]. For example, Yanagimoto and Sugimoto [145] divided the raw 16-channel EEG data into 1s segments and used a seven-layer CNN with 10 ms kernels on the first layer, leading to accuracy improvements of over 20%.…”

Section: Learning Spatial Features From Physiologymentioning

confidence: 99%

See 1 more Smart Citation

Deep Learning for Human Affect Recognition: Insights and New Developments

Rouast

Adam

Chiong

2021

IEEE Trans. Affective Comput.

158

View full text Add to dashboard Cite

Automatic human affect recognition is a key step towards more natural human-computer interaction. Recent trends include recognition in the wild using a fusion of audiovisual and physiological sensors, a challenging setting for conventional machine learning algorithms. Since 2010, novel deep learning algorithms have been applied increasingly in this field. In this paper, we review the literature on human affect recognition between 2010 and 2017, with a special focus on approaches using deep neural networks. By classifying a total of 950 studies according to their usage of shallow or deep architectures, we are able to show a trend towards deep learning. Reviewing a subset of 233 studies that employ deep neural networks, we comprehensively quantify their applications in this field. We find that deep learning is used for learning of (i) spatial feature representations, (ii) temporal feature representations, and (iii) joint feature representations for multimodal sensor data. Exemplary state-of-the-art architectures illustrate the progress. Our findings show the role deep architectures will play in human affect recognition, and can serve as a reference point for researchers working on related applications.

show abstract

Section: Learning Spatial Features From Physiologymentioning

confidence: 99%

“…5). RNNs have been used to learn temporal context from EEG features to improve recognition accuracies [146], [55]. Brady et al [18] found that learning temporal context with LSTM and handcrafted features leads to improvements over shallow baseline models.…”

Section: Learning Temporal Features From Physiological Datamentioning

confidence: 99%

Deep Learning for Human Affect Recognition: Insights and New Developments

Rouast

Adam

Chiong

2021

IEEE Trans. Affective Comput.

158

View full text Add to dashboard Cite

show abstract

“…These features are theta (4-8 Hz), slow alpha (8-10 Hz), alpha (8-12 Hz), beta (12)(13)(14)(15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29)(30), and gamma (30+ Hz), spectral power for 32 electrodes, and the difference between the spectral powers of all the symmetrical pairs of electrodes. For feature elimination, Fisher's linear discriminant was used and the Gaussian naive Baye's is used for the classification.…”

Section: Related Workmentioning

confidence: 99%

“…Their RNN consists of fully connected two LSTM layers, a dropout layer, and a dense layer. Zhang et al [15] presented a deep learning framework called spatiotemporal recurrent neural network (STRNN) in order to combine the learning of spatiotemporal features for emotion recognition using the SJTU Emotion EEG Dataset (SEED).…”

Section: Related Workmentioning

confidence: 99%

EEG-Based Emotion Recognition using 3D Convolutional Neural Networks

Salama¹,

El-Khoribi²,

Shoman³

et al. 2018

ijacsa

160

View full text Add to dashboard Cite

Abstract-Emotion recognition is a crucial problem in Human-Computer Interaction (HCI).Various techniques were applied to enhance the robustness of the emotion recognition systems using electroencephalogram (EEG) signals especially the problem of spatiotemporal features learning. In this paper, a novel EEG-based emotion recognition approach is proposed. In this approach, the use of the 3-Dimensional Convolutional Neural Networks (3D-CNN) is investigated using a multi-channel EEG data for emotion recognition. A data augmentation phase is developed to enhance the performance of the proposed 3D-CNN approach. And, a 3D data representation is formulated from the multi-channel EEG signals, which is used as data input for the proposed 3D-CNN model. Extensive experimental works are conducted using the DEAP (Dataset of Emotion Analysis using the EEG and Physiological and Video Signals) data. It is found that the proposed method is able to achieve recognition accuracies 87.44% and 88.49% for valence and arousal classes respectively, which is outperforming the state of the art methods.

show abstract

“…Tong et.al. [9], proposed a novel deep learning framework to recognize emotion states for SEED database using differential entropy feature which is calculated for five bands. Their algorithm classified for 4 kinds of emotion states like anger, happiness, sadness and surprise with an accuracy rate of more than 90%.…”

Section: Related Workmentioning

confidence: 99%

Detection of human emotions using features based on discrete wavelet transforms of EEG signals

S¹,

Ravikumar²

2018

IJET

View full text Add to dashboard Cite

Affective computing is an emerging area of research in human computer interaction where researchers have developed automated assessment of human emotion states using physiological signals to establish affective human compute interactions. In this paper wepresent an efficient algorithm for emotion recognition using EEG signals for the data acquired by audio-video stimuli. The desired frequency bands are extracted using discrete wavelet transforms. The Statistical features, Hjorth parameters, differential entropy and wavelet features are extracted. Artificial neural networks, Support Vector Machine (SVM) and K-nearest neighbor are used on the extracted feature set to develop prediction models and to classify into four emotion states like clam, happy, fear and sad . These Artificial neural network models are evaluated on the acquired dataset and emotions are classified into four different states with over all accuracy of 86.36%. The classification rate of calm, happy, fear and sad states are 90.9%, 63.63%, 90.90 and 100 % respectively.

show abstract

Spatial–Temporal Recurrent Neural Network for Emotion Recognition

Cited by 430 publications

References 36 publications

Deep Learning for Human Affect Recognition: Insights and New Developments

Deep Learning for Human Affect Recognition: Insights and New Developments

EEG-Based Emotion Recognition using 3D Convolutional Neural Networks

Detection of human emotions using features based on discrete wavelet transforms of EEG signals

Contact Info

Product

Resources

About