There is currently no standard or widely accepted subset of features to effectively classify different emotions based on electroencephalogram (EEG) signals. While combining all possible EEG features may improve the classification performance, it can lead to high dimensionality and worse performance due to redundancy and inefficiency. To solve the high-dimensionality problem, this paper proposes a new framework to automatically search for the optimal subset of EEG features using evolutionary computation (EC) algorithms. The proposed framework has been extensively evaluated using two public datasets (MAHNOB, DEAP) and a new dataset acquired with a mobile EEG sensor. The results confirm that EC algorithms can effectively support feature selection to identify the best EEG features and the best channels to maximize performance over a four-quadrant emotion classification problem. These findings are significant for informing future development of EEG-based emotion classification because low-cost mobile EEG sensors with fewer electrodes are becoming popular for many new applications.\ud \ud Keywords\ud \ud Emotion classification; EEG signals; Feature selection; Evolutionary computation algorith

Typical commercial lease agreements in Australia stipulate 22.5 ± 1.5 °C in summer as the acceptable thermal condition that buildings have to meet, even though the overcooling incurs excessive and unnecessary energy use, gas emissions and financial expense. An argument that backs up this practice asserts that office workers' productivity and comfort will be jeopardised outside this temperature range. This paper investigated whether the office environments with a practical higher temperature setpoint can still be cognitively efficient and comfortable for office workers. In a controlled climate chamber, 26 office workers experienced the typical summer indoor temperature condition in Australia (22 °C) followed by the condition with a higher temperature setpoint (25 °C). In both conditions, subjects were required to fulfil Cambridge Brain Science (CBS) cognitive performance tests, NASA Task Load Index (NASA-TLX), and thermal comfort and air quality questionnaires. Meanwhile, participants' electroencephalogram (EEG) and heart rate (HR) were recorded under three different difficulty levels of Paced Auditory Serial Addition Tests (PASAT). Results showed that CBS test scores were not significantly affected by temperature; a higher temperature of 25 °C incurred a significantly reduced cognitive load for subjects, as has been observed by NASA-TLX, but probably due to the learning effect; the comparison between EEG and HR features during different temperatures did not show any significant difference. Participants' thermal comfort was not significantly jeopardized by the 3 °C temperature setback either. Results from this study favourably support a practical setback of temperature setpoints in Australian office buildings during summer

Recently, emotion recognition using low-cost wearable sensors based on electroencephalogram and blood volume pulse has received much attention. Long short-term memory (LSTM) networks, a special type of recurrent neural networks, have been applied successfully to emotion classification. However, the performance of these sequence classifiers depends heavily on their hyperparameter values, and it is important to adopt an efficient method to ensure the optimal values. To address this problem, we propose a new framework to automatically optimize LSTM hyperparameters using differential evolution (DE). This is the first systematic study of hyperparameter optimization in the context of emotion classification. In this paper, we evaluate and compare the proposed framework with other state-of-the-art hyperparameter optimization methods (particle swarm optimization, simulated annealing, random search, and tree of Parzen estimators) using a new dataset collected from wearable sensors. Experimental results demonstrate that optimizing LSTM hyperparameters significantly improve the recognition rate of four-quadrant dimensional emotions with a 14% increase in accuracy. The best model based on the optimized LSTM classifier using the DE algorithm achieved 77.68% accuracy. The results also showed that evolutionary computation algorithms, particularly DE, are competitive for ensuring optimized LSTM hyperparameter values. Although DE algorithm is computationally expensive, it is less complex and offers higher diversity in finding optimal solutions.

Stress is a major concern in daily life, as it imposes significant and growing health and economic costs on society every year. Stress and driving are a dangerous combination and can lead to life-threatening situations, evidenced by the large number of road traffic crashes that occur every year due to driver stress. In addition, the rate of general health issues caused by work-related chronic stress in drivers who work in public and private transport is greater than in many other occupational groups. An in-vehicle warning system for driver stress levels is needed to continuously predict dangerous driving situations and proactively alert drivers to ensure safe and comfortable driving. As a result of the recent developments in ambient intelligence, such as sensing technologies, pervasive devices, context recognition, and communications, driver stress can be automatically detected using multimodal measurements. This critical review investigates the state of the art of techniques and achievements for automatic driver stress level detection based on multimodal sensors and data. In this work, the most widely used data followed by frequent and highly performed selected features to detect driver stress levels are analyzed and presented. This review also discusses key methodological issues and gaps that hinder the implementation of driver stress detection systems and offers insights into future research directions.

Emotion recognition using miniaturised wearable physiological sensors has emerged as a revolutionary technology in various applications. However, detecting emotions using the fusion of multiple physiological signals remains a complex and challenging task. When fusing physiological signals, it is essential to consider the ability of different fusion approaches to capture the emotional information contained within and across modalities. Moreover, since physiological signals consist of time-series data, it becomes imperative to consider their temporal structures in the fusion process. In this study, we propose a temporal multimodal fusion approach with a deep learning model to capture the non-linear emotional correlation within and across electroencephalography (EEG) and blood volume pulse (BVP) signals and to improve the performance of emotion classification. The performance of the proposed model is evaluated using two different fusion approaches – early fusion and late fusion. Specifically, we use a convolutional neural network (ConvNet) long short-term memory (LSTM) model to fuse the EEG and BVP signals to jointly learn and explore the highly correlated representation of emotions across modalities, after learning each modality with a single deep network. The performance of the temporal multimodal deep learning model is validated on our dataset collected from smart wearable sensors and is also compared with results of recent studies. The experimental results show that the temporal multimodal deep learning models, based on early and late fusion approaches, successfully classified human emotions into one of four quadrants of dimensional emotions with an accuracy of 71.61% and 70.17%, respectively.