Previous electroencephalogram (EEG) emotion recognition relies on single-task learning, which may lead to overfitting and learned emotion features lacking generalization. In this paper, a graph-based multi-task self-supervised learning model (GMSS) for EEG emotion recognition is proposed. GMSS has the ability to learn more general representations by integrating multiple self-supervised tasks, including spatial and frequency jigsaw puzzle tasks, and contrastive learning tasks. By learning from multiple tasks simultaneously, GMSS can find a representation that captures all of the tasks thereby decreasing the chance of overfitting on the original task, i.e., emotion recognition task. In particular, the spatial jigsaw puzzle task aims to capture the intrinsic spatial relationships of different brain regions. Considering the importance of frequency information in EEG emotional signals, the goal of the frequency jigsaw puzzle task is to explore the crucial frequency bands for EEG emotion recognition. To further regularize the learned features and encourage the network to learn inherent representations, contrastive learning task is adopted in this work by mapping the transformed data into a common feature space. Experiments on SEED, SEED-IV, and MPED datasets show that the proposed model has remarkable advantages in learning more discriminative and general features for EEG emotional signals.

Objective: Electroencephalogram (EEG) based brain–computer interfaces (BCI) in motor imagery (MI) have developed rapidly in recent years. A reliable feature extraction method is essential because of a low signal-to-noise ratio (SNR) and time-dependent covariates of EEG signals. Because of efficient application in various fields, deep learning has been adopted in EEG signal processing and has obtained competitive results compared with the traditional methods. However, designing and training an end-to-end network to fully extract potential features from EEG signals remains a challenge in MI. Approach: In this study, we propose a parallel multiscale filter bank convolutional neural network (MSFBCNN) for MI classification. We introduce a layered end-to-end network structure, in which a feature-extraction network is used to extract temporal and spatial features. To enhance the transfer learning ability, we propose a network initialization and fine-tuning strategy to train an individual model for inter-subject classification on small datasets. We compare our MSFBCNN with the state-of-the-art approaches on open datasets. Results: The proposed method has a higher accuracy than the baselines in intra-subject classification. In addition, the transfer learning experiments indicate that our network can build an individual model and obtain acceptable results in inter-subject classification. The results suggest that the proposed network has superior performance, robustness, and transfer learning ability.

Neuroscience studies have demonstrated the phase-locked characteristics of some early event-related potential (ERP) components evoked by stimuli. In this study, we propose a phase preservation neural network (PPNN) to learn phase information to improve the Electroencephalography (EEG) classification in a rapid serial visual presentation (RSVP) task. The PPNN consists of three major modules that can produce spatial and temporal representations with the high discriminative ability of the EEG features for classification. We first adopt a stack of dilated temporal convolution layers to extract temporal dynamics while avoiding the loss of phase information. Considering the intrinsic channel dependence of the EEG data, a spatial convolution layer is then applied to obtain the spatial-temporal representation of the input EEG signal. Finally, a fully connected layer is adopted to extract higher-level features for the final classification. The experiments are conducted on two public and one collected EEG datasets from the RSVP task, in which we evaluated the performance and explored the capability of phase preservation of our PPNN model and visualized the extracted features. The experimental results indicate the superiority of the proposed PPNN when compared with previous methods, suggesting the PPNN is a robust model for EEG classification in RSVP task.

In remote sensing image scene classification (RSISC) tasks, downsampling is crucial for reducing computational complexity and cache demands, enhancing the model’s generalization capability of deep neural networks. Traditional methods, such as regular fixed lattice approaches (pooling in CNN and token merging in transformers), often flatten distinguishing texture features, impacting classification performance. To address this, we propose an adaptive token merging transformer (ATMformer) that preserves essential local features by estimating the importance score of each token. This allows significant tokens to be isolated during merging, mitigating the risk of feature blurring. Our experiments on three widely used RSISC datasets (NWPU-RESISC45, Aerial Image Dataset, and EuroSAT) demonstrate that ATMformer achieves state-of-the-art performance across all datasets. These findings underscore the importance of effective downsampling techniques in maintaining classification accuracy while efficiently processing large-scale data.

Object detection in UAV aerial images faces domain-adaptive challenges, such as changes in shooting height, viewing angle, and weather. These changes constitute a large number of fine-grained domains that place greater demands on the network’s generalizability. To tackle these challenges, we initially decompose image features into domain-invariant and domain-specific features using practical imaging condition parameters. The composite feature can improve domain generalization and single-domain accuracy compared to the conventional fine-grained domain-detection method. Then, to solve the problem of the overfitting of high-frequency imaging condition parameters, we mixed images from different imaging conditions in a balanced sampling manner as input for the training of the detection network. The data-augmentation method improves the robustness of training and reduces the overfitting of high-frequency imaging parameters. The proposed algorithm is compared with state-of-the-art fine-grained domain detectors on the UAVDT and VisDrone datasets. The results show that it achieves an average detection precision improvement of 5.7 and 2.4, respectively. The airborne experiments validate that the algorithm achieves a 20 Hz processing performance for 720P images on an onboard computer with Nvidia Jetson Xavier NX.

Depth sensing with single-shot structured light is widely employed for depth measurement of dynamic scenes. For conventional single-shot structured light techniques, the projected pattern is coded with regards to color, shape, and intensity, aiming at assigning each pixel a unique label. However, the features of the projected pattern are extremely susceptible to the surface texture of the measured object, resulting in errors during performing decoding. In this study, a robust and simple binary pattern is proposed for high-accuracy depth sensing. Only two gray values are utilized in the designed pattern, which is insensitive to environmental noise. The pseudo two-dimensional De Bruijn sequence coding method is employed to reduce the size of the decoding window for robust matching. Short slashes are adopted as features of the pattern that greatly simplifies the feature extraction and pattern decoding. Compared with the Kinect and time-of-flight (TOF) depth cameras, our proposed method can acquire high-precision depth in quantitative and qualitative experiments.