Efficient data compression is crucial for the storage and transmission of visual data. However, in facial expression recognition (FER) tasks, lossy compression often leads to feature degradation and reduced accuracy. To address these challenges, this study proposes an end-to-end model designed to preserve critical features and enhance both compression and recognition performance. A custom loss function is introduced to optimize the model, tailored to balance compression and recognition performance effectively. This study also examines the influence of varying loss term weights on this balance. Experimental results indicate that fine-tuning the compression model alone improves classification accuracy by 0.71 % and compression efficiency by 49.32 %, while joint optimization achieves significant gains of 4.04 % in accuracy and 89.12 % in efficiency. Moreover, the findings demonstrate that the jointly optimized classification model maintains high accuracy on both compressed and uncompressed data, while the compression model reliably preserves image details, even at high compression rates.

Object detectors based on convolutional neural networks (CNNs) have been widely applied for autonomous driving and industrial defect detection. Recent studies have shown that they are vulnerable to adversarial examples. Existing attack methods rely on the output of the object detector’s detection head to generate adversarial examples, which limits their applicability. Furthermore, optimizing the loss function of the object detector to generate adversarial perturbations can cause the perturbations to overfit to the source object detector, reducing the transferability of adversarial examples. To address these issues, we propose a novel adversarial example generation framework consisting of a spatial-frequency information masking (SFIM) method and two independent attack branches. The SFIM method masks partial spatial and frequency domain information of the augmented examples to generate diverse adversarial examples. The two independent attack branches, one targeting the backbone network and the other targeting the region proposal network (RPN), are fused with the SFIM method. The two attack branches do not rely on the output of the object detector’s detection head and are applicable to different attack scenarios. Extensive experiments on the PASCAL VOC and MS COCO datasets show that the adversarial examples generated by the proposed framework are highly transferable and can effectively attack black-box detectors of different architectures.

Internet of Medical Things-based healthcare system (IoMTHS) is a kind of industrial information system that integrates life monitoring, pathological inference and drug therapy. However, the sensitive nature and high value of its data make it a prime target for cyberattacks. Although many multiserver authentication protocols have been studied in recent years to ensure that only authorized users can access medical services, new vulnerabilities are always identified and covertly utilized by the smarter adversary due to lack of continuous monitoring and dynamic authentication, reducing the trustworthiness of IoMTHS. To address above challenges, in this article, we propose a multiserver authentication scheme with integrated monitoring (MAIM) for IoMTHS, which achieves user locked access control by strictly and continuously binding system access permissions and user behavior. MAIM consists of a three-factor-based static authentication (TFSA) and a deep learning-based continuous authentication (DLCA). TFSA utilizes double-anonymity strategy to protect users’ privacy and track their malicious behaviors, and uses physical unclonable function (PUF) to protect the security of privacy information in users’ devices and servers, which achieves lightweight and three-factor secrecy. The DLCA trains a deep neural network to recognize the legitimacy of users based on the user behavior transmitted by their sensing devices. TFSA is provably secure under the random oracle model, whereas DLCA exhibits high feasibility with experimental accuracy reaching 100%.

Vehicle visual localization uses vision sensors to capture environmental information, enabling precise localization of autonomous vehicles within their surroundings. However, current visual localization methods generally have some shortcomings: on one hand, they are limited by the camera’s field of view, on the other hand, their robustness is often inadequate under challenging conditions such as lighting changes, long-term scene changes, or occlusions. To address these issues, we formulate a general spherical camera model for both fisheye and panoramic cameras and propose a minimal solution for pose estimation using this model based on vehicle motion characteristic. The minimal solution cannot filter outliers, so a robust estimation framework is necessary. For outlier-rejection, we introduce two frameworks: a probabilistic optimal RANSAC and a globally optimal graph-based framework. We conduct a probabilistic analysis of the RANSAC to demonstrate its enhanced robustness given by the proposed minimal solution. To achieve robustness to extreme outliers (higher than 90%), we decouple the rotation and translation space through the minimal solution to construct maximum consensus graph for the two sub-problems. We then employs a maximum clique search algorithm to find the optimal solutions, achieving deterministic convergence while maintaining real-time performance. Extensive experiments with synthetic data, real-world fisheye images, and 360∘panoramic images validate the robustness and efficiency of our proposed algorithms.

Signal sparse representation (SR) is an evolving research topic. The sparse coding performance is highly dependent on the properties of the system matrix that are usually hard to be guaranteed. Preconditioning is a technique that transforms a linear system into another one with more favorable properties for sparse solution. In this paper, we investigate the problem of constructing suitable preconditioner to improve the performance of the SR system. We formulate the model for designing the preconditioner by making the product of preconditioner and dictionary strictly approximate to a unit-norm tight frame (UNTF) which has been proved valid in facilitating sparse coding. A parametrization and gradient based approach is presented to solve for the UNTF-based preconditioner. Experiments on speech signals are carried out to demonstrate the performance of the proposed method.

This paper studies the uncertainty principle of the second-order local polynomial Fourier transform (LPFT). It shows that the uncertainty product of the LPFT is time-independent when the Gaussian window is used to segment the signal. Meanwhile when the extra parameter is estimated correctly, the uncertainty product of the LPFT becomes a constant. Compared to the short-time Fourier transform and the Wigner-Ville distribution, it shows that the LPFT provides a better resolution of signal presentation in the time-frequency domain. Simulation for a speech signal is also given to confirm that the LPFT is capable of revealing more spectrum details when the frequency contents change dramatically.

This paper studies the effects of cross-terms in the local polynomial periodograms (LPP) by defining and examining the cross local polynomial periodogram (CLPP) and the time-frequency correlative coefficients (TFCC). Our observations show that the cross-terms of signal components can be ignored if they do not intersect and the window length used in the LPP is sufficient. In addition, the TFCC is useful for the analysis of time-frequency coherency between LFM signals and gives us a parametric measure on the amount of the cross-terms in the LPP.