Driver emotion recognition (DER) is critical for driver monitoring in intelligent cockpits. However, existing methods face the dual challenges of label noise and feature degradation under real-world driving conditions, such as illumination variations and facial occlusions. To address these issues, this article proposes DER-NR, a novel approach that simultaneously mitigates feature degradation and enhances label reliability in driving scenarios. For robust feature extraction, we design the driver-environment feature-enhanced noise suppression (DFNS) module, which integrates an efficient multiscale attention (EMA) module, contextual transformer attention (CoTAttention), and a feature weighting mechanism to dynamically reconstruct facial expression features affected by environmental constraints. During training, the pseudolabel generation (PLG) module generates pseudolabels based on feature confidence ranking and iteratively refines them to rectify noisy labels. Ablation studies and comparison experiment further confirm the method’s generalizability and state-of-the-art (SOTA) performance in addressing feature degradation and label reliability issues caused by environmental constraints. Experimental results demonstrate that DER-NR achieves consistent improvements of accuracy across different datasets and noise levels. On RAF-DB, it leads by 2.92% at 20% noise and 3.44% at 50% noise. On DEFE, the margins are 9.84% at 20% noise and 1.20% at 50% noise. On KMU-FED, it outperforms the second-best method by 3.63% at 0% noise and by 3.50% at 5% noise. On AffectNet, it outperforms the second-best method by 2.76% at 20% noise. Additionally, real-world edge-device deployment shows that our method improves prediction accuracy without excessively increasing computational overhead.

The technological landscape is rapidly evolving toward large-scale systems. Networks supporting massive connectivity through numerous Internet of Things (IoT) devices are at the forefront of this advancement. In this paper, we examine Wireless Power Transfer (WPT)-enabled networks, where a server requires to collect data from these IoT devices to compute a task with massive Ultra-Reliable and Low-Latency Communication (mURLLC) services. We focus on information freshness, using Age-of-Information (AoI) as the key performance metric. Specifically, we aim to minimize the maximum AoI among IoT devices by optimizing the scheduling policy. Our analytical findings demonstrate the convexity of the problem, enabling efficient solutions. We introduce the concept of AoI-oriented cluster capacity and analyze the relationship between the number of supported devices and network AoI performance. Numerical simulations validate our proposed approach's effectiveness in enhancing AoI performance, highlighting its potential for guiding the design of future IoT systems requiring mURLLC services.

WiFi sensing represents an emerging technology that enables human sensing in a low-cost, non-invasive manner. Respiration, as a vital sign, can be employed not only for the assessment of physiological conditions but also for human identification. This paper proposes a highly accurate real-time vital sign monitoring algorithm and a high-precision person identification algorithm. The real-time monitoring algorithm utilizes WiFi device antenna diversity and Maximum Ratio Combining Discrete Wavelet Transform (MRC-DWT) to recover respiration signals, combining with the Autocorrelation Function (ACF) for accurate instantaneous respiration estimation. The identification algorithm leverages biometric information from respiration waveforms and the Multi-dimensional Dynamic Time Warping (MD-DTW) algorithm with modern classifiers for high-precision recognition. This paper validates the efficacy of the proposed algorithms by leveraging the WiFi Respiration Track (WiRes-Track) System. Experimental results demonstrate that the system is capable of discriminating among multiple users, attaining an average recognition accuracy of 91% for six users and a respiration rate accuracy of 96.8%