REV Journal on Electronics and Communications

Intelligent AutoEncoder-Based Modulation: Optimizing Transmission Performance over Non-Ideal Wireless Channels

2026-07-12T04:13:04+00:00

This paper presents an End-to-End wireless transceiver architecture based on deep AutoEncoder (AE) networks that jointly optimizes the transmitter and receiver as a single differentiable system, replacing the conventional cascade of independently designed signal processing blocks. The channel model incorporates three concurrent non-ideal impairments: nonlinear distortion from the power amplifier (PA) characterized by the Rapp model, progressive carrier frequency offset (CFO), and flat Rayleigh fading. Through the training process and testing scenarios across three progressively evolving architectures, namely single-symbol constellation shaping, multi-symbol blind CFO compensation, and implicit neural forward error correction (Neural FEC), the obtained results confirm that the AE is capable of autonomously learning PAresilient signal constellations, performing blind CFO estimation without pilot signals, and unifying the modulation and channel coding processes into a single optimal system representation. Monte Carlo BER simulations show that the proposed architecture achieves 3–5 dB SNR gain over conventional 16-QAM with ZF equalization, provides 2–3 dB gain relative to ideally CFO-compensated 16-QAM, and the Neural FEC configuration successfully performs the channel coding function, exhibiting effective error correction performance.

Enhancing FMCW Radar-Based Human Activity Recognition using DI-ResNet

2026-07-12T04:14:38+00:00

This paper proposes a method to enhance the quality of Human Activity Recognition (HAR) based on Frequency Modulated Continuous Wave (FMCW) radar. The method utilizes a Dual Input-ResNet (DI-ResNet) model to address the limitations associated with relying solely on isolated range or micro-Doppler (m-D) features for activity classification. Specifically, two parallel ResNet-18 backbones are employed to extract deep semantic features from two distinct data domains: the Range-Time (RT) spectrogram and the Doppler-Time (DT) spectrogram. These features are subsequently fused via a concatenation layer to synthesize global context, thereby generating a more comprehensive representation of the performed activities. Experimental results demonstrate that the proposed method achieves superior recognition performance compared to single-stream baseline networks. Notably, the proposed model effectively mitigates confusion among activities exhibiting kinematic similarity, improving the Recall metric for the “SitDown” activity by 42.1% compared to traditional single-input methods. Furthermore, the accuracy for complex hand gestures such as "Drink" significantly increased by 19.8%, substantially minimizing the high misclassification rate associated with "Grab." Finally, the model achieves near-ideal reliability for safety-critical activities, attaining a recognition accuracy of 99.5% in fall detection, thereby confirming its potential for practical deployment.

Efficient Multimodal Feature Refinement via Adaptive RGB-IR Interaction for Robust Drone Detection and Classification

2026-07-12T04:17:40+00:00

The rapid proliferation of unmanned aerial vehicles (UAVs) has intensified the need for robust surveillance systems capable of distinguishing drones from biological entities like birds in unpredictable environments. While multispectral vision provides a resilient alternative to uni-modal sensors under adverse weather and lighting, existing architectures often struggle with cross-modal feature alignment and noise-induced spatial distortions. This paper proposes Multispectral Attention Context and Receptive-field Network (MACR-Net), an ultra-lightweight multimodal framework designed for high-precision drone detection. MACR-Net introduces a Global-Local Cross-Scale Interaction (GLCI) module to capture multi-scale semantic context and a Multimodal Spatial Cross-Perception (MSCP) mechanism to adaptively fuse RGB-IR streams while preserving target-specific thermal and structural signatures. Furthermore, we design an improved hybrid neck integrating Coordinate-Aware Attention (CAA) and Receptive Field Deformable (RFD) modules to anchor precise spatial coordinates and mitigate geometric distortions. Experimental results on the benchmark Multimodal Drone Detection Dataset demonstrate that MACR-Net outperforms state-of-the-art models, achieving a peak mAP_50 of 91.13% and a significant mAP_50-95 of 65.77%. Remarkably, the architecture maintains an extremely compact footprint with only 2.77M parameters and 0.77 GFLOPs, establishing an optimal balance between superior detection robustness and real-time feasibility for resource-constrained edge deployment.

Fast Automatic Shoot Score Approach Combining Optical Flow And Mobilenetv2 Classification

2026-07-12T04:18:50+00:00

This paper presents a fast automatic shooting scoring system designed for fixed-camera setup in real shooting range, where the paper target may undergo slight physical movement caused by wind, recoil-induced stand vibration. The proposed method combines an initial fixed homography with continuously updated dense optical flow for drift-free global alignment, residual optical flow for sub-pixel frame-to-frame registration, adaptive background modeling via motion-compensated subtraction for bullet hole candidate generation, and a highly efficient MobileNetV2 classifier for robust false-positive rejection. By maintaining precise registration to a static high-resolution Score Image template and leveraging the lightweight yet powerful quantized MobileNetV2 network, the system achieves over 98.5\% scoring accuracy at more than 10 fps entirely on CPU even when the target exhibits light movement and under changing natural lighting. Extensive experiments on indoor and outdoor live-fire datasets confirm robustness and real-time performance compared to traditional per-frame feature-matching methods and heavyweight deep learning approaches.

Performance and Memory Trade-offs in SM4 Implementation on Embedded ARM Cortex-M Microcontrollers

2026-07-12T07:22:39+00:00

The SM4 block cipher has been widely adopted in security applications across embedded systems, particularly as part of China’s national cryptographic standards. However, its practical deployment on resource-constrained microcontrollers remains challenging due to limited processing power and memory. This paper investigates the trade-offs between performance and memory usage in various SM4 software implementations on ARM Cortex-M microcontrollers. We evaluate and compare three implementation strategies: (1) S-box lookup tables stored in Flash and SRAM, (2) T-table optimization that combines substitution and transformation operations, and (3) direct computation of the S-box using Galois Field (GF) logic. Each implementation is benchmarked on a 32-bit STM32 microcontroller to measure encryption latency, SRAM and Flash memory usage, and code complexity. The results reveal that the T-table implementation in SRAM provides the best performance with the lowest encryption latency, albeit at the cost of high SRAM consumption. Conversely, the GF logic implementation minimizes memory usage but suffers from the slowest execution time. This study provides important insights for selecting the most suitable SM4 implementation strategy for embedded systems with varying resource constraints and real-time requirements.

Factors Limiting the Frequency of RTD Oscillators

2026-07-12T07:24:29+00:00

The availability of compact, coherent, room-temperature sources with high output power is critical for numerous applications in the terahertz (THz) frequency range. Resonant-tunneling-diode (RTD) oscillators represent one of the most promising technologies for these requirements. This paper analyzes the factors limiting the maximum oscillation frequencies of RTD oscillators, specifically focusing on contact parasitics, spreading resistance, and antenna losses. By mitigating the effects of these parasitic elements, we optimize the antenna parameters—such as the slot antenna length and width—and the RTD mesa area to maximize oscillation frequencies for a given RTD wafer.