Frequency Modulated Continuous Wave (FMCW)-based mmWave radar has attracted widespread attention due to its non-contact and high spatial resolution for multi-target vital signs monitoring. However, current works mostly focus on how to improve detection performance under the steady scenarios with a single target, while unsteady scenarios and physical mutual interference from multiple targets are rarely considered. In this work, we propose an innovative method for multi-target vital signs for unsteady scenarios, such as aerobic exercise. The method automatically distinguishes between steady state and motion state, and completes a best-effort vital signs detection during unsteady state. To differentiate multiple targets, we design a weight vector enhancement method combined with target space localization, and then, apply the variational mode decomposition (VMD) algorithm to extract the respiratory and heart rates of a single target. Moreover, for evaluating the efforts of exercise, we propose Multi-target Motion Recognition (MMR) based on MobileNet-V2 network to recognize the motion states of multiple targets. The experimental results showed that the mean absolute error of respiratory rate and heart rate decreased to 1.37 bpm and 2.56 bpm, respectively. Meanwhile, the MMR algorithm achieves close to 98.1% accuracy in recognizing motion states.

In this paper, we focus on weakly supervised referring expression comprehension (REC), and identify that the lack of fine-grained visual capability greatly limits the upper performance bound of existing methods. To address this issue, we propose a novel framework for weakly supervised REC, namely Dynamic Visual routing Network (DViN), which overcomes the visual shortcomings from the perspective of feature combination and alignment. In particular, DViN is equipped with a novel sparse routing mechanism to efficiently combine features of multiple visual encoders in a dynamic manner, thus improving the visual descriptive power. Besides, we further propose an innovative weakly supervised objective, namely Routing-based Feature Alignment (RFA), which facilitates the visual understanding of routed features through the intra-modal and inter-modal alignment. To validate DViN, we conduct extensive experiments on four REC benchmark datasets. Experiments demonstrate that DViN achieves state-of-the-art results on four benchmarks while maintaining competitive inference efficiency. Besides, the strong generalization ability of DViN is also validated on weakly supervised referring expression segmentation. Source codes are anonymously released at: https://github.com/XxFChen/DViN.

As interest grows in generating long, detailed image captions, standard evaluation metrics become increasingly unreliable.N-gram-based metrics though efficient, fail to capture semantic correctness.Representational Similarity (RS) metrics, designed to address this, initially saw limited use due to high computational costs, while today, despite advances in hardware, they remain unpopular due to low correlation to human judgments.Meanwhile, metrics based on large language models (LLMs) show strong correlation with human judgments, but remain too expensive for iterative use during model development.We introduce SPECS (Specificity-Enhanced CLIP-Score), a reference-free RS metric tailored to long image captioning.SPECS modifies CLIP with a new objective that emphasizes specificity: rewarding correct details and penalizing incorrect ones.We show that SPECS matches the performance of open-source LLMbased metrics in correlation to human judgments, while being far more efficient.This makes it a practical alternative for iterative checkpoint evaluation during image captioning model development.

In the evolving landscape of vehicular networks, it is crucial to ensure robust security and efficient data handling. In this work, we introduce a novel federated learning (FL) algorithm integrated within a distributed edge intelligence (DEI) framework, enhanced by a blockchain consensus mechanism, specifically designed for Internet of Vehicles (IoV) to enhance data privacy, efficiency, and system resilience. Motivated by the pressing need for improved data privacy and security in the IoV, our approach can not only prioritize these aspects but also enhance the efficiency and accuracy of distributed machine learning. The proposed consensus mechanism, by integrating Proof-of-Knowledge (PoK) with practical Byzantine fault tolerance (PBFT), is crafted to be lightweight, making it suitable for the dynamic and resource-constrained vehicular environments. Our evaluation findings demonstrate the algorithm’s superior performance and scalability, suggesting its applicability in diverse IoV scenarios and its potential to facilitate secure, robust, and efficient collaborative learning.

Watermark is essential for protecting the intellectual property of private images. However, a wide range of watermark removal attacks, especially many AI-powered ones, can automatically predict and remove watermarks, posing serious concerns. In this paper, we present the design ofImprints, a defensive watermarking framework that fortifies watermarks against watermark removal attacks. By formulating an optimization problem that deters watermark removal attacks, we design image-independent/dependent defensive watermark models for effective batch/customized protection. We further enhance the watermark to be transferable to unseen watermark removal attacks and robust to editing distortions. Extensive experiments verify thatImprintsoutperforms existing baselines in terms of its immunity to 8 state-of-the-art watermark removal attacks and 3 commercial black-box watermark removal software. The source code is available athttps://github.com/Imprints-wm/Imprints.

Benefiting from smartphones’ powerful computing and sensing capabilities, biometric authentication is widely used on them to conveniently verify user’s identity. However, most biometric features can be easily acquired or reproduced, making them vulnerable to replay and impersonation attacks. To address these issues, we propose DBreathLock, a non-contact identity authentication system that leverages a smartphone to capture unique biometric features from users’ deep breaths. Specifically, the smartphone emits inaudible frequency-modulated continuous waves (FMCW)-based sonar signals to capture chest-abdominal-joint (C-A-joint) movements, increasing the difficulty of impersonating legitimate users. By analyzing the energy features of CA-joint movements, we develop a Deep Breath Activity Detection (DBAD) method to detect deep breath fragments alongside the capability of resisting advanced replay attacks. To further fortify the resistance to advanced replay attacks, we employ the morphological features of C-A-joint movements and SVC model to recognize these attacks. Following this, we construct mutual information (MI) sequences via calculating the correlation between C-A-joint movements and breath sounds to reduce identity authentication failure caused by physiological changes. Then, a multi-stream identity authentication model is designed to verify legitimate users by fusing the features from C-A-joint movements, deep breath sounds, and MI sequences. Extensive real-world experiments involving 20 users demonstrate that DBreathLock achieves an authentication accuracy of 98.33%. Additionally, it successfully defends against both advanced replay and impersonation attacks.