With the development of deep learning, trackers based on convolutional neural networks (CNNs) have made significant achievements in visual tracking over the years. The fully connected Siamese network (SiamFC) is a typical representation of those trackers. SiamFC designs a two-branch architecture of a CNN and models visual tracking as a general similarity-learning problem. However, the feature maps it uses for visual tracking are only from the last layer of the CNN. Those features contain high-level semantic information but lack sufficiently detailed texture information. This means that the SiamFC tracker tends to drift when there are other same-category objects or when the contrast between the target and the background is very low. Focusing on addressing this problem, we design a novel tracking algorithm that combines a correlation filter tracker and the SiamFC tracker into one framework. In this framework, the correlation filter tracker can use the Histograms of Oriented Gradients (HOG) and color name (CN) features to guide the SiamFC tracker. This framework also contains an evaluation criterion which we design to evaluate the tracking result of the two trackers. If this criterion finds the SiamFC tracker fails in some cases, our framework will use the tracking result from the correlation filter tracker to correct the SiamFC. In this way, the defects of SiamFCs high-level semantic features are remedied by the HOG and CN features. So, our algorithm provides a framework which combines two trackers together and makes them complement each other in visual tracking. And to the best of our knowledge, our algorithm is also the first one which designs an evaluation criterion using correlation filter and zero padding to evaluate the tracking result. Comprehensive experiments are conducted on the Online Tracking Benchmark (OTB), Temple Color (TC128), Benchmark for UAV Tracking (UAV-123), and Visual Object Tracking (VOT) Benchmark. The results show that our algorithm achieves quite a competitive performance when compared with the baseline tracker and several other state-of-the-art trackers.
With the increasing use of wireless devices, available spectrum resources are becoming increasingly precious. Cooperative spectrum sensing (CSS) is a key technology in cognitive radio (CR) networks. It enhances spectrum utilization efficiency by enabling multiple secondary users (SUs) to jointly detect the spectrum occupancy of the primary user (PU). However, its inherent data fusion mechanism makes it vulnerable to spectrum sensing data falsification (SSDF) attack. To enhance the system’s SSDF fault tolerance, this paper proposes a novel two-stage hybrid malicious user (MU) detection model integrating random forest (RF) and particle swarm optimization-support vector machine (PSO-SVM), denoted as RF-SVM. This model first employs an RF for preliminary screening of SUs, followed by support vector machine (SVM) for fine-grained classification, thereby combining the strengths of both algorithms. Experimental results demonstrate that the proposed RF-SVM model outperforms both the PSO-SVM and standalone RF models, as well as advanced methods such as the enhanced support vector machine (ESVM) and isolation forest-spectral clustering (IFSC), in terms of both detection accuracy and false alarm control. This model maintains stable performance even under scenarios with high-intensity attacks and limited training samples, while also achieving efficient computation. These advantages fully verify its effectiveness, robustness and practical application value.
1 April 2026 in IEEE Transactions on Affective Computing
Multi-source domain adaptation (MSDA) aims to transfer knowledge from multi-source domains to an unlabeled target domain, making it highly suitable for cross-session and cross-subject emotion recognition based on electroencephalography (EEG). Nevertheless, current MSDA models primarily focus on the relationship between the global feature distribution of the source and the target domain while neglecting the relationship between local feature pairs, resulting in poor robustness. In this study, we propose an EEG-based Multi-Source Domain Knowledge Transfer Framework (MS-KTF) for cross-session and cross-subject emotion recognition. The MS-KTF uses a triplet loss combined with a maximum mean discrepancy (MMD) to achieve unbiased alignment at both the domain and data-pair levels of each pair data in the source and target domains. Moreover, the introduction of knowledge distillation enables MS-KTF to achieve better robustness without requiring additional training time. We evaluated the MS-KTF's performance on the DEAP, SEED and SEED-IV datasets for two-, three- and four-class emotion recognition, respectively. The experimental results demonstrated that MS-KTF achieves competitive performance in both cross-session and cross-subject scenarios.
The extensive utilization of the Snappy compression algorithm in digital devices such as smartphones, IoT, and digital cameras has played a crucial role in alleviating demands on network bandwidth and storage space. This paper presents an improved Snappy compression algorithm optimized for implementation on field programmable gate arrays (FPGAs). The proposed algorithm enhances the compression ratio by refining the encoding format of Snappy and introduces an innovative approach utilizing fingerprints within the dictionary to minimize storage space requirements. Additionally, the algorithm incorporates a pipeline structure to optimize performance. Experimental results demonstrate that the proposed algorithm achieves a throughput of 1.6 GB/s for eight hardware kernels. The average compression ratio is 2.27, representing a 6.1% improvement over the state-of-the-art Snappy FPGA implementation. Notably, the proposed algorithm architecture consumes fewer on-chip storage resources compared to other advanced algorithms, striking a balance between logic and storage resource utilization. This optimization leads to higher FPGA resource utilization efficiency. Our design addresses the growing demand for efficient lossless data compression solutions in consumer electronics, ultimately contributing to advancements in modern digital ecosystems.
1 January 2025 in IEICE Transactions on Fundamentals of Electronics, Communications and Computer Sciences
Large-scale interactive systems require efficient performance and scalability to accommodate user demands. The response time of individual requests can be a limiting factor for overall performance. One effective solution is to employ a consistent hashing algorithm for request distribution, which alleviates the load on each server and enhances response speed. However, existing consistent hash algorithms often fail to account for the heterogeneity of resources, which may degrade performance and result in underutilized resources. To solve this problem, this paper proposes a solution based on the study of cluster load balancing using a consistent hash algorithm to increase the resource utilization. Firstly, The scheme adopts a two-dimensional spatial coordinate positioning. Partitions are formed by dividing the ring created by the value range of the CRC32 or CRC64 hash function. The partition and relative position for a key are determined by its vertical and horizontal coordinate, respectively. Secondly, nodes with different performance are assigned with different weights, and each partition also has a corresponding weight which determined by the number and weight of nodes within it. Experiments have shown that this scheme can achieve a search rate of more than 30 million keys per second, with a resource utilization rate of over 80% within the cluster, and maintain stability when changing the number of nodes, thereby improving the overall performance of the system.