The recognition of human emotions remains a challenging task for social media images. This is due to distortions created by different social media conflict with the minute changes in facial expression. This study presents a new model called the Global Spectral-Spatial Attention Network (GSSAN), which leverages both local and global information simultaneously. The proposed model comprises a shallow Convolutional Neural Network (CNN) with an MBResNext block, which integrates the features extracted from MobileNet, ResNet, and DenseNet for extracting local features. In addition, to strengthen the discriminating power of the features, GSSAN incorporates Fourier features, which provide essential cues for minute changes in the face images. To test the proposed model for emo-tion recognition using social media images, we conduct experiments on two widely-used datasets: FER-2013 and AffectNet. The same benchmark datasets are uploaded and downloaded to create a distorted social media image dataset to test the proposed model. Experiments on distorted social media images dataset show that the model surpasses the accuracy of SOTA models by 0.69% for FER-2013 and 0.51% for AffectNet social mediad datasets. The same inference can be drawn from the experiments on standard datasets.

This paper proposes a framework using artificial intelligence for recognizing human mental stress based on physiological data. It comprises three stages; at first, emotion detection has been carried out from the facial image using a deep learning-based framework employing convolutional neural networks (CNN). Secondly, electrocardiogram (ECG) signals have been acquired with the help of the developed sensor-based system integrating Arduino UNO with AD8232 sensor. Finally, preprocessing has been performed and extracted features from the ECG signals are fed to support vector machines (SVM) for identification of arrhythmia and hence diagnosis of mental stress. All three aspects are connected together via the overall output calculation. A score is added in accordance to the output for each of these three stages. The range of the final score (i.e., the overall output) decides the level of mental stress suffered by the person. The proposed method has been trained on FER-2013 for emotion detection and on MIT-BIH for arrhythmia detection. It is found that the performance outperforms other machine learning and deep learning models without much compromise in accuracy. The proposed method can perform classification based on the patient’s current condition and gives an insight of the degree and time duration of stress of the patient.