Facial expression recognition (FER) in uncontrolled environment is challenging due to various un-constrained conditions. Although existing deep learning-based FER approaches have been quite promising in recognizing frontal faces, they still struggle to accurately identify the facial expressions on the faces that are partly occluded in unconstrained scenarios. To mitigate this issue, we propose a transformer-based FER method (TFE) that is capable of adaptatively focusing on the most important and unoccluded facial regions. TFE is based on the multi-head self-attention mechanism that can flexibly attend to a sequence of image patches to encode the critical cues for FER. Compared with traditional transformer, the novelty of TFE is two-fold: (i) To effectively select the discriminative facial regions, we integrate all the attention weights in various transformer layers into an attention map to guide the network to perceive the important facial regions. (ii) Given an input occluded facial image, we use a decoder to reconstruct the corresponding non-occluded face. Thus, TFE is capable of inferring the occluded regions to better recognize the facial expressions. We evaluate the proposed TFE on the two prevalent in-the-wild facial expression datasets (AffectNet and RAF-DB) and the their modifications with artificial occlusions. Experimental results show that TFE improves the recognition accuracy on both the non-occluded faces and occluded faces. Compared with other state-of-the-art FE methods, TFE obtains consistent improvements. Visualization results show TFE is capable of automatically focusing on the discriminative and non-occluded facial regions for robust FER.

Aiming at the problem of insufficient details of retinal blood vessel segmentation in current research methods, this paper proposes a multiscale feature fusion residual network based on dual attention. Specifically, a feature fusion residual module with adaptive calibration weight features is designed, which avoids gradient dispersion and network degradation while effectively extracting image details. The SA module and ECA module are used many times in the backbone feature extraction network to adaptively select the focus position to generate more discriminative feature representations; at the same time, the information of different levels of the network is fused, and long-range and short-range features are used. This method aggregates low-level and high-level feature information, which effectively improves the segmentation performance. The experimental results show that the method in this paper achieves the classification accuracy of 0.9795 and 0.9785 on the STARE and DRIVE datasets, respectively, and the classification performance is better than the current mainstream methods.

In this paper, we discuss general data-framing and optimization, parallel frequency-phase estimation, and powerful frequency estimator for satellite communications. Based on the standard pilot-symbol-assisted-modulation (S-PSAM) frame, we design a general PSAM (G-PSAM) frame, whose corresponding data-aided & non-data-aided Cramér-Rao bound (DA&NDA CRB) is derived with respect to carrier frequency offset, time delay and phase offset, followed by its approximate form. Also, we provide a simple solution to the G-PSAM frame optimization by means of the approximate DA&NDA CRB and the control-variate method. With an unbiased auto-correlation (AC) value, we then propose a frequency-phase decoupler and develop a parallel frequency-phase estimation (PFPE) mode in terms of the serial frequency-phase estimation (SFPE) mode, where a serial-to-parallel estimation conversion can be implemented. Furthermore, we extend the AC frequency estimator in the PFPE mode as a powerful stepwise correlation-based one aiming at the target performance. Theoretical analysis and numerical results are provided to show the superiority of the optimized frames, the robustness of the PFPE mode, and the adaptability of the proposed frequency estimator with the optimized frames under different transmission systems.

The rapid growth of social media platforms has led to an unprecedented increase in user-generated content and social interactions, posing significant challenges for recommendation systems. This article addresses the challenges of recommendation in large-scale dynamic social environments, where user interactions and preferences evolve rapidly across vast networks. In large-scale dynamic social networks, knowledge discovery requires methods to efficiently process vast amounts of data while capturing the evolving nature of user interactions and preferences. Multigraph neural networks offer a promising approach for this task, as they can model complex relationships and temporal dynamics in these environments. This article proposes a novel social-aware multigraph neural network for the session-based recommendation (SAMGNN-SR) model that leverages dynamic social information and multigraph neural networks to enhance recommendation accuracy and knowledge discovery in complex social computing environments. The model constructs a global social-aware interaction graph from all user session sequences and employs an adaptive subgraph sampling strategy to extract relevant collaborative signals efficiently. A dynamic interest extraction module utilizing dual-direction information propagation captures users’ evolving preferences, while a social information fusion network based on graph attention mechanisms models the dynamic nature of social influences. Experiments on three real-world datasets (Douban, Delicious, and Yelp) demonstrate the superiority of SAMGNN-SR over nine state-of-the-art baselines, with improvements of up to 6.79% in NDCG@20 and 6.19% in Hit@20. Ablation studies validate the effectiveness of each model component in capturing complex social dynamics and session-based user behaviors.

Considering limited pilot and storage resources in satellite communications, we propose an efficient correlation-based reception scheme consisting of a “parallel” frequency-phase estimation (PFPE) mode and a data blockwise compensation (DBC) scheme. In specific, the PFPE mode provides an autocorrelation (AC) frequency offset estimator and a maximum-likelihood (ML) phase offset estimator using a special but heuristic AC value; the DBC scheme uses a small amount of cross-correlation (CC) based compensation values to eliminate the impact of residual offsets on each data block. Simulation results show that with a small pilot overhead the proposed reception scheme can weaken “frequency-to-phase coupling” and achieve good error performance with a small demand for storage.

In this paper, we study a class of nonbinary LDPC (NBLDPC) codes whose parity‐check matrices have column weight 2, called NBLDPC cycle codes. We propose a design framework of (2, ρ)‐regular binary quasi‐cyclic (QC) LDPC codes and then construct NBLDPC cycle codes of large girth based on circulants and finite fields by randomly choosing the nonzero field elements in their parity‐check matrices. For enlarging the girth values, our approach is twofold. First, we give an exhaustive search of circulants with column/row weight ρ and design a masking matrix with good cycle distribution based on the edge‐node relation in undirected graphs. Second, according to the designed masking matrix, we construct the exponent matrix based on finite fields. The iterative decoding performances of the constructed codes on the additive white Gaussian noise (AWGN) channel are finally provided.