Electroencephalogram (EEG) signals have been widely used in emotion recognition. However, the current EEG-based emotion recognition has low accuracy of emotion classification, and its real-time application is limited. In order to address these issues, in this paper, we proposed an improved feature selection algorithm to recognize subjects’ emotion states based on EEG signal, and combined this feature selection method to design an online emotion recognition brain-computer interface (BCI) system. Specifically, first, different dimensional features from the time-domain, frequency domain, and time-frequency domain were extracted. Then, a modified particle swarm optimization (PSO) method with multi-stage linearly-decreasing inertia weight (MLDW) was purposed for feature selection. The MLDW algorithm can be used to easily refine the process of decreasing the inertia weight. Finally, the emotion types were classified by the support vector machine classifier. We extracted different features from the EEG data in the DEAP data set collected by 32 subjects to perform two offline experiments. Our results showed that the average accuracy of four-class emotion recognition reached 76.67%. Compared with the latest benchmark, our proposed MLDW-PSO feature selection improves the accuracy of EEG-based emotion recognition. To further validate the efficiency of the MLDW-PSO feature selection method, we developed an online two-class emotion recognition system evoked by Chinese videos, which achieved good performance for 10 healthy subjects with an average accuracy of 89.5%. The effectiveness of our method was thus demonstrated.

Objective and accurate detection of Parkinson’s disease (PD) is crucial for timely intervention and treatment. Electroencephalography (EEG) has been proven to characterize PD by measuring brain activity. In recent years, deep learning methods have gained great attention in automated PD detection, but their performance is limited by insufficient data samples. In this article, we propose a novel PD automated detection model named the multiscale convolutional prototype network (MCPNet), which integrates and improves upon multiscale convolutional neural networks (CNNs) and prototype learning. On the one hand, it employs multiscale CNNs to extract brain features from different scales, enhancing feature diversity and utilization. On the other hand, a prototype calibration strategy is introduced to mitigate the effect of data noise on prototype generation, improving the generalization performance of model. Multiple within-dataset and cross-dataset experiments on three different datasets demonstrate the effectiveness of our model in PD detection. The leave-one-subject-out (LOSO) results of within-dataset experiments show that MCPNet achieves an accuracy of 92.5%, a sensitivity of 93.1%, a specificity of 91.9%, and an AUC of 92.4% in cross-subject classification between PD patients and healthy controls. In the cross-dataset classification, the performance of MCPNet is somewhat weakened due to dataset variations. However, this weakening is partially compensated by introducing the prototype calibration strategy. With the introduction of the calibration strategy, the accuracy of cross-dataset classification increases to 90.2%, a 4.0% improvement compared to when it is not used. These results indicate that the proposed model may be a promising tool for automated PD diagnosis.

At present, emotion recognition based on electroencephalograms (EEGs) has attracted much more attention. Current studies of affective brain-computer interfaces (BCIs) focus on the recognition of happiness and sadness using brain activation patterns. Fear recognition involving brain activities in different spatial distributions and different brain functional networks has been scarcely investigated. In this study, we propose a multifeature fusion method combining energy activation, spatial distribution, and brain functional connection network (BFCN) features for fear emotion recognition. The affective brain pattern was identified by not only the power activation features of differential entropy (DE) but also the spatial distribution features of the common spatial pattern (CSP) and the EEG phase synchronization features of phase lock value (PLV). A total of 15 healthy subjects took part in the experiment, and the average accuracy rate was 85.00% ± 8.13%. The experimental results showed that the fear emotions of subjects were fully stimulated and effectively identified. The proposed fusion method on fear recognition was thus validated and is of great significance to the development of effective emotional BCI systems.

In the realm of pervasive smart media technology, human activity recognition (HAR) stands as a critical functionality, particularly for applications within smart homes, virtual reality and mobile devices. These data can be collected through accelerometers and gyroscopes embedded in smart devices such as smartphones and smartwatches. Traditional HAR algorithms are often highly specialized, being developed for specific datasets and tasks. These methods typically rely on large amounts of high-quality labeled data to achieve optimal performance, but such labeled data is notoriously difficult and expensive to obtain in practice. In contrast, the daily use of smart devices generates a vast amount of unlabeled data as people engage in various activities. If the unlabeled data is properly and effectively utilized that could significantly advance the development of human activity recognition. In this work, we propose a transformer-based framework that first converts continuous human activity data into discrete tokens, which shortens the sequence length while preserving local information. We then construct a somatosensory lexicon, where the model facilitates self-supervised learning by predicting masked somatosensory tokens as a proxy task. After learning useful features through unlabeled proxy tasks, the model requires only a small amount of labeled data for fine-tuning to achieve performance comparable to supervised learning. Experiments on three public datasets (UCI HAR, MHEALTH, UniMiB, SHAR), show that our approach can achieve performance comparable to fully labeled supervised learning, even with minimal labeling. This reduces the need for extensive labeling while maintaining high accuracy, providing a more efficient and scalable solution for human activity recognition in the age of widespread smart technology.