Abstract：Technology roadmaps serve as a core tool for forecasting and delineating the evolutionary path of technological innovation. Currently, China’s integrated circuit industry faces multiple overlapping challenges, including domestically made demands, emerging technologies in the post-Moore era, and explosive market growth driven by artificial intelligence. It is no longer feasible to fully follow mainstream global technological trajectories. Under supply chain constraints, it is imperative to explore independent next-generation technological paths, consolidate consensus on cutting-edge directions, and formulate localized IC technology roadmaps adapted to domestic industrial conditions. Based on the analysis and research on the key directions, organizational methods, participating entities, and main characteristics of the mainstream technology roadmaps in the global integrated circuit field, this paper puts forward thoughts and suggestions for China to formulate an independent integrated circuit technology roadmap in the post-Moore Era.

Abstract：To address the challenges of large-scale variation, low contrast, and diversified overlapping defects in weld images under complex backgrounds, an improved YOLOv11-based algorithm for surface weld defect detection is proposed. A SEAMHead structure is introduced to adaptively enhance defect-relevant salient regions via spatial and channel attention mechanisms while suppressing redundant background interference, thereby improving the model′s capability to perceive defect boundaries and fine-grained details. In addition, a WaveletPool module is employed to effectively extract texture and structural information across multiple scales, enhancing the capture of key weld features under complex conditions. The proposed model achieves precision, recall, F1-score, and mAP50 of 88.1%, 87.8%, 87.9% and 88.7%, respectively, outperforming other YOLO-series models in overall performance. The results demonstrate that the proposed method significantly improves defect detection capability while maintaining high detection accuracy, achieving an effective balance between precision and recall, and is particularly suitable for efficient weld defect inspection in industrial scenarios where missed detections are highly undesirable.

Abstract：Aiming at the problem of insufficient accuracy in automatic inventory counting caused by dense stacking and mutual occlusion of materials in warehouse scenarios, an end-to-end high-precision instance counting method is studied. The method innovatively constructs a dual-branch regression counting network named WareCountNet++, the core of which lies in integrating branch-level transfer learning and multi-layer feature fusion mechanisms. The network extracts features in parallel through an attention-enhanced branch and a customized detection backbone branch, and utilizes the proposed mechanisms to promote the interaction of heterogeneous features and the deep integration of multi-scale information at multiple levels, thereby enhancing the model′s ability to characterize complex scenes. Experimental results on a self-built warehouse materials dataset show that the proposed WareCountNet++ model achieves a Mean Absolute Error of 1.8, a Root Mean Square Error of 3.2, a relative MAE and a relative RMSE of 4.2% and 7.5% respectively, and a Coefficient of Determination (R2) of 0.993. Its comprehensive performance significantly outperforms mainstream detection-based counting methods such as the YOLO series. Ablation experiments confirmed the key contribution of branch-level transfer learning in reducing relative error and the key contribution of multi-layer feature fusion in improving counting accuracy. The research shows that this method can effectively enhance the robustness and accuracy of counting in dense and occluded scenes, providing a feasible technical solution for intelligent warehouse inventory.

Abstract：A multi-protocol and multi-channel audio interface is designed to address the limitations of traditional audio interfaces, such as limited protocol support and restricted channel capacity. The design supports full-duplex communication and employs eight independent serializers as the core. By integrating a clock and frame synchronization generation module, it produces flexibly configurable clock and frame sync signals. Combined with a formatting unit, the interface supports three transmission modes: TDM, DIT, and Burst. This enables compatibility with multiple audio protocols and formats through a single interface, including I²S and TDM (supporting up to 32 channels based on TDM mode), S/PDIF (based on DIT mode), and burst control frames (based on Burst mode), with a maximum capacity of 256 audio data streams. Functional verification via NC-Verilog simulation demonstrates that all design functions perform as expected.

Abstract：To address the challenges faced by modern high-performance signal processing in terms of real-time performance and integration, this paper designs a heterogeneous signal processing board based on FPGA and DSP microsystems. The core innovation of this board lies in the adoption of system-level packaging technology, which constructs two heterogeneous microsystems centered around FPGA and DSP respectively, and integrates DDR3 and Flash storage, achieving tight coupling between computation and storage. This paper elaborates on the overall architecture, microsystem composition, hardware implementation, and system testing and verification scheme of the board. Test results show that the board achieves high-speed data exchange between microsystems through the SRIO bus, completes low-latency control communication using the EMIF interface, and ensures reliable external data links through Gigabit Ethernet. This design combines heterogeneous computing with advanced integration technology, providing a high-performance, miniaturized, and highly reliable hardware solution for signal processing platforms in fields such as radar and communication.

Abstract：As internet information technology continues to evolve and advance, information security has become increasingly critical. As a next-generation power information communication method, HPLC+HRF dual-mode communication technology has prompted State Grid Corporation of China to impose higher requirements on dual-mode carrier communication modules. The modules must undergo identity authentication through security technologies, with communication information encrypted accordingly. This solution employs a security algorithm coprocessor to handle authentication and encryption/decryption during carrier communication, utilizing high-speed SPI for data transmission. It features coprocessor upgrade capabilities, expanding the security application scenarios for power communication. The solution not only meets State Grid Corporation's security requirements for authentication and encryption/decryption but also satisfies overall dynamic power consumption demands in low-power mode. It has been applied in Smart Chips' self-developed concentrators, Type I and Type II collectors' communication units, as well as single-phase and three-phase meter carrier communication units, distributed across local communication systems in transmission, distribution, and power consumption information collection systems across various provincial grids.

Abstract：With the advancement of electronic information and packaging technologies, traditional board-level solutions composed of numerous discrete components in defense-related fields such as missile and UAV systems can no longer meet the requirements for miniaturization. In response to the current trend of lightweighting in weapon systems, this paper proposes an indigenous System-in-Package (SiP) design based on ARM+FPGA. The SiP integrates four circuit types—ARM Cortex-M4 MCU, FPGA, SPI Flash, and LDO circuits—onto the packaging substrate, achieving a leap from board-level to chip-level miniaturization through advanced packaging techniques. While meeting the size and weight requirements for current equipment miniaturization, this SiP design maintains compatibility with existing discrete component solutions in both hardware and software development. Testing results confirm that the circuit functionality and performance fully meet requirements, effectively addressing users' needs for miniaturization and lightweighting.

Abstract：Ultrasonic microscopy technology has been widely adopted in non-destructive testing of wafers, through-silicon vias, and insulated gate bipolar transistor (IGBT) packaging devices owing to its advantages of high resolution and real-time imaging. To achieve synchronized high-speed acquisition and transmission of ultrasonic signals at the hundred-megahertz level, a high-speed ultrasonic data acquisition board with a sampling rate of 2 GSa/s and a resolution of 8 bits was developed. The system employs a field-programmable gate array as the control core, integrating circuit designs for analog-to-digital conversion data acquisition, data buffering via DDR3 memory, and high-speed data transmission via peripheral component interconnect express. A single master clock-based synchronization control method is proposed to address the clock synchronization challenge inherent in high-speed acquisition systems. On this basis, host computer software was developed, and the system was integrated into an ultrasonic microscopy platform for C-scan imaging experiments. Experimental results demonstrate that all performance metrics of the acquisition board meet the design specifications, with satisfactory imaging quality achieved, confirming the board's reliability and practical value for engineering applications in ultrasonic microscopy systems.

Abstract：To address the challenges of noise interference and feature extraction difficulty for rotating machinery acoustic signals under complex working conditions, variable speeds, and multiple fault types, this paper proposes a bearing fault diagnosis method based on multi-domain feature fusion and the XGBoost classifier. Utilizing multi-condition bearing audio signals from the BJTU-RAO Bogie dataset and the University of Ottawa motor fault dataset, time-domain, frequency-domain, and Mel-frequency cepstral coefficients (MFCC) features are extracted from the raw signals. A core feature set characterized by high contribution and low redundancy is selected by integrating XGBoost feature importance ranking, Shapley additive explanations value analysis, and feature correlation heatmaps, thereby reducing model complexity and enhancing interpretability. During the classification phase, the strong generalization capability of XGBoost is leveraged for fault identification under multiple working conditions. Experimental results demonstrate that the proposed method achieves weighted average accuracy and recall rates exceeding 96% across all conditions in the BJTU-RAO dataset. In the University of Ottawa dataset, it achieves accuracy over 98% under constant-speed conditions and maintains above 83% under variable-speed conditions. These findings validate the high accuracy, robustness, and generalization performance of the method in multi-condition and multi-fault scenarios. The innovative integration of multi-domain acoustic feature fusion with feature importance analysis enables efficient diagnosis of rotating machinery acoustic signals, offering significant potential for online monitoring in noisy industrial environments.

Abstract：In response to the issue of phase noise effects in scatter communication systems employing high-order modulation and high carrier frequencies, this paper proposes a phase noise simulation method suitable for scatter communication systems. It investigates the impact of phase noise of varying intensities on the demodulation performance of scatter communication systems with different modulation orders and with or without channel coding. Additionally, a phase noise suppression algorithm based on average noise suppression and segmented iterative estimation and compensation is proposed. This algorithm can achieve phase noise tracking and suppression in low signal-to-noise ratio environments with minimal pilot overhead. Simulation results demonstrate that the algorithm exhibits strong robustness against phase noise of varying intensities, different modulation orders, and systems with or without channel coding, significantly mitigating the impact of phase noise on the demodulation performance of scatter communication systems.

Abstract：D-PMU data in distributed grids often goes missing due to equipment faults and communication issues, impacting grid reliability and constraining fault prediction and self-healing accuracy. This paper proposes the SAITS model, which uses a self-attention mechanism to fill missing D-PMU data. SAITS combines masked prediction and reconstruction tasks, leveraging diagonally masked self-attention to capture long-term dependencies and correlations and effectively impute missing data. It achieves high precision even with up to 90% missing data, outperforming LOCF, BRITS, and TF models in RMSE and MAE, with an R² score above 97%. SAITS enhances data integrity and reliability, and improves the reliability of power grid protection equipment, providing robust support for fault prediction and self-healing in smart power IoT systems to further strengthen the system′s self-healing capability and operational reliability.

Abstract：To address the core pain points in traditional park construction, such as information silos， data fragmentation and insufficient business scalability, this paper proposes a solution for smart parks integrated management platform with microservice architecture as the core. Firstly, a five-layer open and scalable "Cloud-Network-Terminal" system architecture is constructed, clarifying the collaborative logic of the front-end sensing layer, information transmission layer, data resource layer, platform service layer, and business presentation layer. Secondly, the focus is placed on designing the software architecture of the integrated management platform, modular implementation views, process interaction mechanisms, and redundant deployment solutions. Leveraging the independent deployment and dynamic expansion characteristics of microservices, the barriers between business systems are broken, enabling data fusion analysis and intelligent decision support. Practical application shows that the platform can increase the response efficiency of park business by 40% and achieve a data sharing rate of 92%. It effectively solves the problems of business chimneys and information silos, providing key technical support for the intelligent transformation of park management.

Abstract：Medical image segmentation plays a pivotal role in intelligent diagnosis and clinical decision-making, serving as the foundation for disease detection, organ delineation, and intraoperative navigation. In recent years, the rapid development of deep learning has greatly advanced segmentation algorithms, yet challenges remain due to modality heterogeneity, blurred boundaries, and limited annotated data. This paper provides a comprehensive review of deep learning–based medical image segmentation methods, covering discriminative models such as Convolutional Neural Networks, Recurrent Neural Networks, and Transformers, as well as generative approaches including Generative Adversarial Networks and diffusion models. Furthermore, this paper discusses the adaptation strategies of large foundation models, represented by the Segment Anything Model, in medical imaging tasks. In addition, this work summarizes the key challenges faced by current segmentation methods and presents experimental comparisons of representative models across multiple multimodal datasets. Finally, the paper outlines the current research trends and future directions, offering insights for improving segmentation model design and facilitating their clinical applications.

Abstract：During the process of abnormal sampling value determination in the equipment status monitoring unit, if detection is only carried out by mining a single feature target, it is impossible to synchronously associate the detection objects within the interval, resulting in a decrease in the accuracy of the output results. For this purpose, an analysis and research on the abnormal detection technology of sampling values for the equipment condition monitoring unit of intelligent substations is proposed. The sampling values of the monitoring unit are quantified by using autoregressive fitting. After eliminating the non-uniform error values, the regression processing results are output. An abnormal response mechanism is established, using the autoregressive results as guidance, to mine the abnormal features of associated sampling within the normal response distribution interval, synchronously sample the detection objects, and determine the specific abnormal targets. Based on the abnormal targets output, alarm identification and location are carried out. Combined with the detection deviation, backtracking compensation and correction are conducted to ensure that the substation equipment is restored to the expected operating state. The experimental results show that the false detection rate of the proposed method is between 0.8% and 1.4%. It is the highest in the initial period, reaching 1.4%, and gradually stabilizes in the subsequent periods, controlled between 1.0% and 1.1%, without significant fluctuations. This indicates that the detection accuracy and effectiveness have improved, and the performance is superior and reliable.

Abstract：To meet the requirements of wide-spectrum monitoring for receivers with large bandwidth, high dynamic range, and strong anti-interference capability, this paper designs an ultra-wideband RF receiver front-end system based on a superheterodyne dual-conversion architecture, covering the operating frequency band from 0.1 to 10 GHz. By employing a dual-path parallel low-noise amplifier (LNA) structure, the receiver′s ability to capture weak signals across its wide operating frequency range is effectively enhanced. Through rational frequency planning and the design of an image-rejection filter bank, combined with the introduction of multiple high-roll-off bandpass filters at the intermediate frequency (IF) stage, the receiver′s suppression capability against image interference and adjacent-channel interference is significantly improved. The measured results show that within the operating frequency band, the noise figure is less than 6.13 dB, the gain is greater than 47 dB, the gain dynamic range exceeds 83 dB, the output third-order intercept point (OIP3) is greater than 25.7 dBm, the adjacent-channel rejection is greater than 36.9 dBc, the image rejection is greater than 64 dBc, and the Error Vector Magnitude (EVM) of a 64QAM modulated signal at a symbol rate of 10 Msym/s is less than 2.89%. All performance metrics meet the design requirements.

Abstract：VHF/UHF band blade antennas, favored for their simple structure and omnidirectional radiation characteristics, are widely employed in aviation communications. However, conventional blade antennas suffer from large physical dimensions, which challenge the requirement for miniaturization in aviation communication equipment. This paper proposes a miniaturized blade antenna achieved through an integrated design of a fan-shaped radiating patch and matching network. The antenna measures 160 mm×80 mm×100 mm (corresponding to 0.05λlow×0.03λlow×0.03λlow, where λlow is the wavelength at the lowest operating frequency). The fabricated antenna exhibits a voltage standing wave ratio below 3.5 across the 100~800 MHz frequency range, demonstrating ultra-wideband impedance performance. Additionally, the antenna maintains a gain greater than -5 dBi from 400 to 800 MHz, with H-plane radiation pattern ripple under 3 dB, enabling broadband omnidirectional radiation. The measured performance of the fabricated blade antenna agrees well with simulation, validating the effectiveness of this design.

Abstract：To address the nonlinear distortion issues in high-efficiency transmission for high-power phased array shortwave transmission systems, this study proposes a digital predistortion (DPD) scheme for phased array shortwave transmission systems based on a memory polynomial algorithm. The proposed scheme employs Fast Fourier Transform (FFT) technology to perform phase calibration on the 100% amplitude test signal from the coupled output of multiple solid-state transmitters in the system. It then uses a memory polynomial model to identify the nonlinear characteristics and implements real-time predistortion correction of the excitation baseband signal in the FPGA of the excitation device. Experimental results demonstrate that the proposed scheme effectively reduces the harmonic distortion across the entire frequency band to within 2%, achieves an amplitude asymmetry of better than 2%, and ensures carrier suppression superior to ±2%. This satisfies the system specifications outlined in the standard "Technical Requirements and Measurement Methods for Medium and Shortwave Amplitude Modulation Broadcast Transmitters" (GY/T 225-2007) and effectively addresses the issues of spectral regeneration and waveform distortion caused by the saturation characteristics of solid-state transmitters, significantly improving the system's transmission quality.

Abstract：The insulation aging of DC unarmored cables in urban rail traction power supply systems often leads to non-metallic short circuits, resulting in cable burnout and seriously affecting power supply safety. This article is based on cable current and real-time estimation of cable temperature, providing technical support for cable over temperature protection. Firstly, based on the principle of thermoelectric analogy, a transient thermal circuit model was constructed, which includes cable conductors, insulation layers, comprehensive waterproof layers, and outer sheaths. The equivalent thermal resistance of air convection heat dissipation was considered to achieve real-time calculation of cable temperature. Then an experimental platform was built to conduct temperature rise experiments under constant load, impact load, and actual typical load. The experimental results show that the error between the calculated temperature and the measured temperature values of the model under the three operating conditions is less than 2%, which can meet the practical application requirements.

Abstract：The A2B(Automotive Audio Bus) developed by ADI has become a mainstream solution for automotive digital audio, and its bit error rate has a significant impact on the sound quality of automotive audio signals. This paper proposes a bit error rate test scheme for the A2B bus based on FPGA(Filed Programed Gate Array). Three AD2428W chips are used to build an A2B bus architecture with one master and two slaves. The FPGA completes the configures of the A2B bus, sends test data to the A2B bus, and simultaneously receives the data output by the bus to check the number of bit errors, and then uploads the number of bit errors to the upper computer through the serial port. Through the design of the hardware and software of the test system, the continuous 288 hours test shows that the number of bit errors is zero and the bit error rate is lower than 1×10-14. The test verifies that the system can effectively detect bit errors, and the program is easy to modify and transplant.

Abstract：To meet the application requirements of autonomous safety, high reliability and miniaturization for the high-voltage power distribution system of special vehicles, a high-voltage intelligent power distribution control module based on the domestic 32-bit MCU HKG32F407 is designed. With the domestic core processor as the control center, the module integrates multi-channel contactor drive, high-precision current/voltage acquisition, temperature detection, CAN bus communication, fault storage and other functions. Adopting designs such as high-low voltage isolation, wide-voltage input, and dual hardware & software protection, it realizes core capabilities including 12-channel contactor control, over-current/over-voltage protection, adhesion detection, and real-time status reporting. Tests and verifications show that the system features stable operation, high acquisition accuracy, and strong anti-interference capability, adapting to a wide temperature range of -40℃~80℃. It can realize the localization of core devices, and is applicable to scenarios such as new energy special vehicles and industrial high-voltage power distribution, possessing favorable engineering application value.

Abstract：To address the issue of high peak sidelobe levels in sparse array optimization, an optimization method based on the improved whale migration algorithm is proposed. Based on the original whale migration algorithm, an adaptive regulation mechanism is introduced in the follower update stage to dynamically balance global search and local exploitation, reducing the risk of falling into local optima. In the leader update stage, a team collaboration strategy is incorporated to improve convergence efficiency through multi-level information sharing. Meanwhile, the directional search capability for the global optimal solution is enhanced by a leadership strategy. Simulation results demonstrate that the proposed method achieves better peak sidelobe level suppression and higher optimization stability than existing algorithms, validating its effectiveness.

Abstract：To address the stringent requirements for high-precision, multi-channel, high-speed synchronous data acquisition in applications such as phased-array radar, distributed measurement, and high-frequency communications, this paper presents a multi-device synchronized acquisition system based on the AD9680 analog-to-digital converter and the HMC7043 clock synchronization chip. The system employs the LMK04828 as the master clock reference and utilizes three HMC7043 devices to achieve precise clock synchronization and signal distribution across twelve AD9680 ADCs. A dual-FPGA architecture is adopted, with each FPGA controlling 12 ADC channels, and high-speed data transmission is implemented via the JESD204B interface protocol. Experimental results under a 1 000 MHz sampling rate with various typical input signals demonstrate that the inter-channel phase skew remains within 3°, fully satisfying high-precision synchronization requirements. Furthermore, all dynamic performance metrics meet the design specifications, offering a reliable and practically valuable engineering solution for multi-channel, high-frequency synchronous signal acquisition.

Abstract：To overcome the limitations of single-feature extraction methods in rolling bearing fault diagnosis, this paper proposes a dual-branch fusion model based on Mamba and Swin Transformer. The model first converts vibration signals into time–frequency images using the Gramian Angular Field (GAF) technique. The Swin Transformer branch is then employed to extract image features, while the improved eMamba-1D and 2D Mamba modules are used to process time-series signals, leveraging their complementary strengths. A cross-channel attention mechanism is introduced to achieve efficient fusion between temporal and image features, thereby enhancing diagnostic accuracy and robustness. Experimental results demonstrate that the proposed model achieves over 99% diagnostic accuracy under both constant and variable operating conditions, effectively identifying various bearing fault types. Validation on the CWRU bearing dataset and the University of Ottawa variable-speed bearing dataset further shows that the model outperforms other comparative methods in terms of classification accuracy, precision, recall, and F1-score, confirming its strong generalization capability and high efficiency under complex operating conditions.

Abstract：This study systematically implements four mainstream spatial interpolation algorithms based on a self-developed grid modeling and hybrid computing components, including Inverse Distance Weighting (IDW), Kriging, Discrete Smooth Interpolation (DSI), and Radial Basis Function (RBF) interpolation. A complete grid interpolation module was constructed, and multiple comparative experiments were designed to evaluate the algorithms from multiple dimensions, including accuracy, computational efficiency, robustness, and smoothness. Experimental results show that grid spacing significantly affects both interpolation accuracy and computational efficiency; appropriate grid spacing can substantially reduce error metrics for all algorithms. IDW demonstrates superior computational efficiency and accuracy. Considering spatial structure and smoothness, the RBF algorithm achieves higher precision. The implementation and comparative analysis of these spatial interpolation algorithms provide key technical support for the self-developed grid modeling and hybrid computing components, enhance functional completeness and engineering applicability, and represent a critical step toward the domestication and functional upgrading of core components.

Abstract：Aiming at the dual challenges of limited local feature representation and inaccurate semantic alignment in point cloud scene understanding, this paper proposes a 3D understanding framework integrating a hierarchical meta-architecture and dual-stream context-awareness. Firstly, PointMeta is adopted as the visual backbone, and a multi-scale adaptive feature extraction network is constructed through meta-function decomposition to achieve adaptive fusion of multi-scale features. Then, a context-aware spatial-semantic alignment strategy is introduced to align 3D object features and linguistic features from the perspectives of semantics and context, thereby realizing cross-modal information fusion. Additionally, extra spatial information is incorporated into text labels to reduce ambiguities caused by the relationships between contextual features and language. Experiments on scene understanding using public datasets ScanRefer and Nr3D show that the proposed network model achieves significant improvements compared to PointNet++. Compared with the results of other network models on public datasets, the proposed model exhibits varying degrees of performance enhancement, demonstrating stronger generalization ability and favorable practical application value.