In order to support diverse service demands of communication scenarios, 5G adopts the mixed-numerology system based on orthogonal frequency division multiplexing (OFDM).The mixed-numerology systems still have the problem of high peak-to-average power ratio (PAPR).Furthermore, its design structure makes it difficult to apply traditional PAPR suppression methods straight forwardly, in which mixed signals are generated by superimposing multiple sub-signals with different parameters.An optimization model for minimizing PAPR in mixed-numerology systems was established based on tone reservation method.Utilizing the characteristic of single-peak of time-domain kernel, the multiple copies were generated in different numerologies based on the peak position of the mixed signal.The mixed time-domain kernel was then superimposed to reduce the peaks of the mixed signal.Since the proposed algorithm didn’t require inverse fast Fourier transform (IFFT)/ fast Fourier transform (FFT) operations during iteration, it had relatively low computational complexity.Simulation results show that the proposed algorithm effectively reduces the PAPR of mixed signals.

Aiming at the problem that more and more kinds of radar compound jamming signals and too few training samples were difficult to make the deep learning model reach the optimal state, a semi-supervised transfer learning recognition method for radar compound jamming under small samples was proposed, which solved the problem of low network training accuracy caused by the difficulty in obtaining labeled samples through unlabeled samples.The feature extractor and classifier obtained after pre-training of single jamming data set were transferred to small-scale compound jamming data set, and the model was fine-tuning by using weight imprinting and semi-supervised learning.The model parameters were optimized by the proposed nearest neighbor correlation loss nearest neighbor correlation loss (NNCL).The experimental results show that the recognition accuracy of the model can reach 93.20% when the jamming-to-noise ratio is 10 dB and there are only 5 labeled samples of the new class of compound jamming signals.

In Internet of vehicles (IoV), vehicle-oriented applications are generally computation-intensive and latency-sensitive.Introducing idle computing resources from mobile vehicles as a supplement to network computing power can effectively alleviate the load pressure on edge servers.The problem of task allocation for edge computation offloading in the context of IoV environment were researched.By fully leveraging the combined computing resources of roadside units (RSU), user vehicles, and mobile vehicles within the RSU service range, a computation offloading strategy based on the sparrow search algorithm was proposed and referred to as sparrow search based computation offloading scheme (S²COS), aiming to optimize the overall system energy consumption.In addition, this strategy fully taked into account practical network issues such as service time constraints caused by vehicle mobility and the potential occurrence of computation node failures.The simulation results demonstrate that S²COS can meet the latency requirements for computation-intensive and latency-sensitive tasks, while significantly reducing system energy consumption.

In the control of network terminal power, due to the overall low analysis of network state, the adaptive regulation ability of communication network terminal power is poor.Therefore, the power line carrier communication network terminal power control method based on adaptive hash function was proposed.Considering the dynamic fluctuation of the actual operating state of the power line carrier communication network, the adaptive mechanism was set for the hash function, and the adaptive hash function was constructed for the original running state data, and the final hash value was obtained in the variable chaotic mapping system.In the power control stage of the power line carrier communication network terminal, the communication network terminal was affected by the interference of other communication network terminals, and it was used as the input value of the adaptive hash function to realize the power control of each communication network terminal.In the test results, the proposed control method not only effectively reduces the increase range of network terminal power under constant communication conditions, but also effectively reduces the network terminal power output under dynamic communication conditions, which has a high adaptive adjustment ability.

Cooperative spectrum sensing is a key technology in cognitive radio.A data-driven intelligent cooperative spectrum sensing system was proposed with spatial diversity running on a smart contract to address issues of security, privacy, incentive and hidden terminals in cooperative spectrum sensing.Specifically, a motivated spectrum sensing system was designed by taking advantage of the decentralization of blockchain technology and the immutability of data.Secondly, a deep learning-based approach was proposed to identify malicious users in the system.In addition, to achieve higher accuracy in recruiting sensing nodes more efficiently in the system, a hard decision cooperative spectrum sensing fusion algorithm based on performance weights and spatial diversity was designed.The experimental results indicate that the proposed solution outperforms traditional cooperative spectrum sensing algorithms in terms of security, privacy, motivation, and sensing accuracy.

Multi-polarization range high resolution radar is an important mean for ground target detection.In the echo formed by it, the target occupies multiple range cells and becomes an extended target.The traditional spread target detection method relies on energy, and the detection performance decreases when the signal-to-clutter ratio decreases.A spread target detection algorithm based on polarization decomposition features was proposed, which improved the detection performance under low signal-to-clutter ratio by using the difference of polarization scattering characteristics between target and clutter.Specifically, 16 kinds of polarization decomposition features were extracted to form feature vectors as detection statistics, and then support vector data description (SVDD) was used to obtain the detection threshold.When training the detection threshold, the polarization decomposition features of clutter were extracted as training data.In order to ensure the false alarm probability, two penalty parameters were introduced into the objective function of SVDD.The experimental results show that the proposed method requires a signal-to-clutter ratio of about 12.6 dB in the case of Gobi background, false alarm probability of 10^-4 and detection probability of 90%, which is about 1.7 dB lower than the energy-based methods.

In response to the classification and identification problems of 5 kHz channels, 25 kHz channels, broadband interference channels, narrowband interference channels, and single tone interference channels in the ultrashort wave frequency band, a classification and identification method for ultrashort wave channels based on mirror filled spectrum and LA-ResNet50 (LBP attention ResNet50) was proposed.The problem of difficulty in distinguishing between satellite channels and background noise under low signal-to-noise ratio, as well as the identification of signal channels and interference channels with similar characteristics, has been effectively solved.Firstly, the proposed method performs mirror symmetry on the ultrashort wave spectrum and fills it in, while blackening the edges of the spectrum to construct a mirror-filled spectrum, which improves the discrimination of different types of channel spectra.Then, channel attention was introduced into ResNet50 to focus the attention of the network model on the channel.Finally, a loss function based on cross entropy and local binary pattern (LBP) was proposed to improve the extraction effect of subtle texture features on signal channels and interference channels images.The proposed method based on mirror-filled spectrum and LA-ResNet50 has shown an improvement of 19.8%, 8.2%, 1.8%, and 0.8% in classification accuracy for ultrashort wave channels compared to the traditional method utilizing fast Fourier transform (FFT) spectrum thresholding, the YOLOv5s target detection and classification method based on mirror-filled spectrum, the Attention-ResNet50 method with attention mechanism based on mirror-filled spectrum, and the Transformer network method under a signal-to-noise ratio (SNR) of 10 dB.

Existing intrusion detection techniques often require numerous malicious samples for model training.However, in real-world scenarios, only a small number of intrusion traffic samples can be obtained, which belong to few-shot scenarios.To address this challenge, a network intrusion detection method designed for few-shot scenarios was proposed.The method comprised two main parts: a packet sampling module and a meta-learning module.The packet sampling module was used for filtering, segmenting, and recombining raw network data, while the meta-learning module was used for feature extraction and result classification.Experimental results based on three few-shot datasets constructed from real network traffic data sources show that the method exhibits good applicability and fast convergence and effectively reduces the occurrence of outliers.In the case of 10 training samples, the maximum achievable detection rate is 99.29%, while the accuracy rate can reach a maximum of 97.93%.These findings demonstrate a noticeable improvement of 0.12% and 0.37% respectively, in comparison to existing algorithms.

To address the issue of weak identification and low accuracy in recognizing features of target vehicles at different distances in road scene images, a vehicle feature recognition algorithm based on optimized convolutional neural network (CNN) was proposed.Firstly, a multi-scale input based on the PAN model was employed to capture target vehicle features at varying distances.Subsequently, improvements were made to the network model by incorporating multi-pool, batch normalization (BN) layers, and Leaky ReLU activation functions within the CNN architecture.Furthermore, the generalization ability of the network model was enhanced by introducing a hybrid attention mechanism that focuses on important features and regions in the vehicle image.Lastly, a multi-level CNN structure was constructed to achieve feature recognition for vehicles.Simulation experiment results conducted on the BIT-Vehicle database within a single scene show the proposed algorithm’s significant enhancements in single-object and multi-object recognition rates compared to CNN, R-CNN, ABC-CNN, Faster R-CNN, AlexNet, VGG16, and YOLOV8.Specifically, improvements of 16.75%, 10.9%, 4%, 3.7%, 2.46%, 1.3%, and 1% in single-object recognition, as well as 17.8%, 10.5%, 2.5%, 3.8%, 2.7%, 1.1%, and 1.3% in multi-object recognition, have been demonstrated by the proposed algorithm, respectively.Over the more complex UA-DETRAC datasets, more precise results have been also achieved by the proposed algorithm in recognizing target vehicles at various distances compared to other algorithms.

The 5G flow residence ratio is an important indicator used by communication operators to evaluate the coverage capacity of 5G networks, and it is also an important starting point for the networks’ planning, optimization, maintenance and operation.To ensure user perception, the actual wireless transmission influencing factors, existing network structure and network coordination strategy, etc., need to be considered.The resident capacity of a 5G network in the 3.5 GHz frequency band was evaluated, analyzed and calculated, and the theoretical ultimate value of the 5G traffic resident ratio based on the analysis of network and users’ big data was put forward, which explained the positive relationship between user perception and residence ratio, and provided a reference for the reasonable theoretical ultimate value of 5G traffic resident ratio.

Faced with the fact that the spatial distribution of buildings is not considered in the location of 5G network base stations, which makes the base station site selection effect unsatisfactory and leads to the problem of small coverage of base stations, a research method for 5G network base station site selection based on fruit fly optimization algorithm was proposed.The distance from the front line base station to the circular coverage center was calculated, and the satellite communication equipment was used to locate the return signal to obtain the optional position area of the rear base station.The energy consumption of data reception, transmission and aggregation were calculated, and the load balancing of base station data was ensured under the premise of balanced distribution of cellular units.A two-layer location model for base station site selection was built, constraints were set, and the optimal location of base station was determined.The fruit fly optimization algorithm was used to update the position, and after individual update iterative processing, it was not to fall into local optimization and obtain the fine base station site selection position.It can be seen from the experimental results that the minimum losses of the front line and rear base stations in this method are 50 dB and 70 dB, respectively, and the coverage radius of the base station is 6 km, which is consistent with the coverage of the ideal base station.

New computing applications present new challenges to communication networks, for example, traditional network cannot meet the demands of flexible deployment, customized services, security and trustworthiness of new services.As a development paradigm that separated network technology and supported environment, polymorphic network provided new solutions to support the differentiated demands of various new applications.A type of computation-oriented network modality namely computing power network modality was discussed, for the computation-oriented business requirements based on the development trend of computing power network.Firstly, the design of computing power network modality was investigated, then the key technologies of computing power network modality were analyzed.Finally, the application scenarios of computing power network were prospected.It is hoped that the discussion can provide a reference for the future research and development on application of computation-oriented polymorphic network.

Recently, artificial intelligence generated content (AIGC) technology has achieved various disruptive results and has become a new trend in AI research and application, driving AI into a new era.Firstly, the development status of AIGC technology was analyzed, focusing on generative models such as generative adversarial networks and diffusion models, as well as multimodal technologies, and surveying and elaborating on the existing technological capabilities for text, speech, image and video generation.Then, the risks brought by AIGC technology in the field of content security governance were focused and analyzed, including fake information, content infringement, network and software supply chain security, data leakage and other aspects.Finally, in view of the above security risks, counter strategies were proposed from the technical, application and regulatory levels, respectively.

With the development of information and cloud technology, the dynamic demand for OTN service is gradually increasing.As a next generation OTN switching technology, OSU needs to provide network and cloud services for high-quality enterprise leased lines and boutique home broadband services.The OSU-based OTN needs to be able to provide integrated services and automated service provisioning through end-to-end management and control capabilities to support one-hop service into the cloud.Firstly, the background, technical characteristics, and standardization progress of OSU technology were introduced.Then, a service-aware control framework for OSU-OTN was proposed to meet the demand for efficient bearer of government and enterprise leased lines.Finally, an experimental validation of the end-to-end management and control capability of the access OSU-OTN controller system was carried out.

Pan-video refers to the business that takes video as the form of information presentation and interaction.It is mainly divided into three categories: traditional video business, emerging video business and derivative video business.However, how to accurately evaluate the quality of video playback has always been a problem.A pan-video playback quality evaluation model and parameters were proposed, which mainly included: video switching delay, video start delay, video playback pause times and duration, etc.It has been applied in the IPTV quality monitoring system of the current network, and the practice shows that the application effect is good.The proposed research results can be further applied to all pan-video application scenarios in the field of video network products.

Firstly, several software defined network (SDN) architectures and their key technologies were introduced.Secondly, technical research on the applicability of these SDN architectures in optical transport network (OTN) on several key aspects such as network layering and network abstraction, southbound interface protocol selection, and path computation element (PCE) function integration was conductd and analyzed.Finally, guiding opinions and suggestions on the migration of the existing OTN network to these SDN technical architectures were given.