Deep Learning Framework for Categorical Emotional States Assessment Using Electrodermal Activity Signals.

In this study, we attempted to classify categorical emotional states using Electrodermal Activity (EDA) signals and a configurable Convolutional Neural Network (cCNN). The EDA signals from the publicly available, Continuously Annotated Signals of Emotion dataset were down-sampled and decomposed into phasic components using the cvxEDA algorithm. The phasic component of EDA was subjected to Short-Time Fourier Transform-based time-frequency representation to obtain spectrograms. These spectrograms were input to the proposed cCNN to automatically learn the prominent features and discriminate varied emotions such as amusing, boring, relaxing, and scary. Nested k-Fold cross-validation was used to evaluate the robustness of the model. The results indicated that the proposed pipeline could discriminate the considered emotional states with a high average classification accuracy, recall, specificity, precision, and F-measure scores of 80.20%, 60.41%, 86.8%, 60.05%, and 58.61%, respectively. Thus, the proposed pipeline could be valuable in examining diverse emotional states in normal and clinical conditions.

Comparative Analysis of Electrodermal Activity Decomposition Methods in Emotion Detection Using Machine Learning.

Electrodermal activity (EDA) reflects sympathetic nervous system activity through sweating-related changes in skin conductance. Decomposition analysis is used to deconvolve the EDA into slow and fast varying tonic and phasic activity, respectively. In this study, we used machine learning models to compare the performance of two EDA decomposition algorithms to detect emotions such as amusing, boring, relaxing, and scary. The EDA data considered in this study were obtained from the publicly available Continuously Annotated Signals of Emotion (CASE) dataset. Initially, we pre-processed and deconvolved the EDA data into tonic and phasic components using decomposition methods such as cvxEDA and BayesianEDA. Further, 12 time-domain features were extracted from the phasic component of EDA data. Finally, we applied machine learning algorithms such as logistic regression (LR) and support vector machine (SVM), to evaluate the performance of the decomposition method. Our results imply that the BayesianEDA decomposition method outperforms the cvxEDA. The mean of the first derivative feature discriminated all the considered emotional pairs with high statistical significance (p<0.05). SVM was able to detect emotions better than the LR classifier. We achieved a 10-fold average classification accuracy, sensitivity, specificity, precision, and f1-score of 88.2%, 76.25%, 92.08%, 76.16%, and 76.15% respectively, using BayesianEDA and SVM classifiers. The proposed framework can be utilized to detect emotional states for the early diagnosis of psychological conditions.

Differential Gene Expression Data Analysis of ASD Using Random Forest.

Autism spectrum disorder (ASD) is a developmental disability caused by differences in the brain regions. Analysis of differential expression (DE) of transcriptomic data allows for genome-wide analysis of gene expression changes related to ASD. De-novo mutations may play a vital role in ASD, but the list of genes involved is still far from complete. Differentially expressed genes (DEGs) are treated as candidate biomarkers and a small set of DEGs might be identified as biomarkers using either biological knowledge or data-driven approaches like machine learning and statistical analysis. In this study, we employed a machine learning-based approach to identify the differential gene expression between ASD and Typical Development (TD). The gene expression data of 15 ASD and 15 TD were obtained from the NCBI GEO database. Initially, we extracted the data and used a standard pipeline to pre-process the data. Further, Random Forest (RF) was used to discriminate genes between ASD and TD. We identified the top 10 prominent differential genes and compared them with the statistical test results. Our results show that the proposed RF model yields 5-fold cross-validation accuracy, sensitivity and specificity of 96.67%. Further, we obtained precision and F-measure scores of 97.5% and 96.57%, respectively. Moreover, we found 34 unique DEG chromosomal locations having influential contributions in identifying ASD from TD. We have also identified chr3:113322718-113322659 as the most significant contributing chromosomal location in discriminating ASD and TD. Our machine learning-based method of refining DE analysis is promising for finding biomarkers from gene expression profiles and prioritizing DEGs. Moreover, our study reported top 10 gene signatures for ASD may facilitate the development of reliable diagnosis and prognosis biomarkers for screening ASD.

Automated Emotion Recognition System Using Blood Volume Pulse and XGBoost Learning.

In this study, a new method for detecting emotions using Blood Volume Pulse (BVP) signals and machine learning was presented. The BVP of 30 subjects from the publicly available CASE dataset was pre-processed, and 39 features were extracted from various emotional states, such as amusing, boring, relaxing, and scary. The features were categorized into time, frequency, and time-frequency domains and used to build an emotion detection model with XGBoost. The model achieved the highest classification accuracy of 71.88% using the top 10 features. The most significant features of the model were computed from time (5 features), time-frequency (4 features), and frequency (1 feature) domains. The skewness calculated from the time-frequency representation of the BVP was ranked highest and played a crucial role in the classification. Our study suggests the potential of using BVP recorded from wearable devices to detect emotions in healthcare applications.

Identifying the Optimal Location of Facial EMG for Emotion Detection Using Logistic Regression.

In this study, we analyzed the utility of electromyogram (EMG) signals recorded from the zygomaticus major (zEMG), the trapezius (tEMG), and the corrugator supercilii (cEMG) for emotion detection. We computed eleven-time domain features from the EMG signals to classify the emotions such as amusing, boring, relaxing, and scary. The features were fed to the logistic regression, support vector machine, and multilayer perceptron classifiers, and model performance was evaluated. We achieved an average 10-fold cross-validation classification accuracy of 67.29%. 67.92% and 64.58% by LR using the features extracted from the EMG signals recorded from the zEMG, tEMG, and cEMG, respectively. The classification accuracy improved to 70.6% while combining features from the zEMG and cEMG for the LR model. However, the performance dropped while including the features of EMG from all three locations. Our study shows the importance of utilizing the zEMG and cEMG combination for emotion recognition.

Low Valence Low Arousal Stimuli: An Effective Candidate for EEG-Based Biometrics Authentication System.

Electroencephalography (EEG) has recently gained popularity in user authentication systems since it is unique and less impacted by fraudulent interceptions. Although EEG is known to be sensitive to emotions, understanding the stability of brain responses to EEG-based authentication systems is challenging. In this study, we compared the effect of different emotion stimuli for the application in the EEG-based biometrics system (EBS). Initially, we pre-processed audio-visual evoked EEG potentials from the 'A Database for Emotion Analysis using Physiological Signals' (DEAP) dataset. A total of 21 time-domain and 33 frequency-domain features were extracted from the considered EEG signals in response to Low valence Low arousal (LVLA) and High valence low arousal (HVLA) stimuli. These features were fed as input to an XGBoost classifier to evaluate the performance and identify the significant features. The model performance was validated using leave-one-out cross-validation. The pipeline achieved high performance with multiclass accuracy of 80.97% and a binary-class accuracy of 99.41% with LVLA stimuli. In addition, it also achieved recall, precision and F-measure scores of 80.97%, 81.58% and 80.95%, respectively. For both the cases of LVLA and LVHA, skewness was the stand-out feature. We conclude that boring stimuli (negative experience) that fall under the LVLA category can elicit a more unique neuronal response than its counterpart the LVHA (positive experience). Thus, the proposed pipeline involving LVLA stimuli could be a potential authentication technique in security applications.