Browse
Recent Submissions
Now showing 1 - 5 of 11
- ItemReconstruction of Mixed Boundary Objects and Classification Using Deep Learning and Linear Sampling Method(Společnost pro radioelektronické inženýrství, 2024-06) Harisha, S. B.; Mallikarjun, E.; Amit, M.The linear sampling method is a simple and reliable linear inversion technique for determining the morphological features of unknown objects under investigation. Nevertheless, there are many challenges that this method depends on the frequency of operation and it is unable to produce satisfactory results for objects with complex shapes. This paper proposes a hybrid model, which combines conventional linear sampling method and deep learning for the reconstruction of mixed boundary objects. In this approach, the initial approximation of mixed boundary objects derived from linear sampling method serves as the training data for the U-Net based convolutional neural network. The network then learns to correlate this approximation with the corresponding ground truth profiles. Along with the reconstruction of mixed boundary objects, they are also classified as dielectric or conductor, and count of each object type are measured. Furthermore, the low-frequency and high-frequency characteristics of the linear sampling method are analyzed, and its limitations are overcome by combining it with a deep learning approach. The effectiveness of the proposed model is validated using several examples of synthetic and experimental data. The results demonstrate that the proposed method outperforms the conventional Linear sampling method in terms of accuracy.
- ItemFull-automatic Segmentation Algorithm of Brain Tumor Based on RFE-UNet and Hybrid Focal Loss Function(Společnost pro radioelektronické inženýrství, 2024-06) Wang, Y.; Tian, H.; Ji, Y.; Liu, M.Semantic segmentation of glioma and its subregions plays a critical role in the entirely clinical workflow of brain cancer diagnosis, monitoring, and treatment planning. Recently, automatic tumor segmentation has attracted a lot of attention, especially supervised learning methods based on neural networks, and the popular “U-shaped” network architecture has achieved state-of-the-art performance in many fields of medical image segmentation. Despite the success of these models, the commonly used small convolution kernel can only extract local features, and more global contextual features cannot be learned, resulting in the disappointed performance of modeling long-range information. At the same time, due to the difficulty of obtaining medical image data, and the imbalance of tumor data in which tumor usually occupies a relatively small volume compared with the background, the adverse influence on the training of the model occurs. In this paper, a novel segmentation framework including TensorMixup data augmentation, improved Receptive Field Expansion UNet (RFE-UNet) and hybrid loss function is designed. Specifically, the TensorMixup algorithm in the data preprocessing phase is used to provide more high-quality training data. In the training phase, both a RFE-UNet network and a hybrid loss function are proposed respectively. RFE-UNet network adds Receptive field expansion module based on Dilated convolution in the first three stages of skip connection, which is used to learn more local and global features. In addition, hybrid loss function is mainly composed of focal loss and focal Tversky loss，focal loss increasing the weight of fewer samples and focal Tversky loss focusing on learning the characteristics of samples with incorrect predictions，which is adopted to alleviate data imbalance. The experimental results on the BraTs2019 dataset show that the average Dice value of the proposed algorithm in the intact tumor, tumor core, and enhanced tumor region can reach 91.55%, 89.23%, and 84.16% respectively, which proves the feasibility and effectiveness of using the proposed architecture.
- ItemBalanced Linear-Phase Bandpass Filter Equalized with Negative Group Delay Circuit(Společnost pro radioelektronické inženýrství, 2024-06) Wang, Z. C.; Wang, Z. B.; Gao, M.; Liu, H.; Fang, S.A novel balanced linear-phase bandpass filter is proposed to achieve differential-mode linear-phase filtering and common-mode suppression characteristics. The balanced linear-phase bandpass filter consists of a proposed compact balanced bandpass filter and negative group delay circuits, in which the circuits are loaded on the ports of the filter as branches. The linear-phase performance is achieved through negative compensation of group delay fluctuations using negative group delay circuit equalization. In order to verify the design method, a 3-order balanced linear-phase bandpass filter is designed, simulated, manufactured, and measured. The results show that the group delay fluctuation of the balanced bandpass filter has been reduced by 89.6 % from 1.110 to 0.115 ns. The minimum common-mode suppression within the passband is 41.4 dB. The proposed balanced bandpass filter has an excellent differential-mode linear-phase transmission and common-mode suppression performances.
- ItemA Novel IoT Intrusion Detection Model Using 2dCNN-BiLSTM(Společnost pro radioelektronické inženýrství, 2024-06) Xiang, R. H.; Li, S. S.; Pan, J. L.With the continuous advancement of Internet of Things (IoT) intelligence, IoT security issues have become more and more prominent in recent years. The research on IoT security has become a hot spot. A lightweight IoT intrusion detection model fusing a convolutional neural network, bidirectional long short-term memory network is proposed. It aims to improve processed data security and attack detection accuracy. First, sampling is performed by a hybrid sampling algorithm fusing SMOTE and ENN. Its aim is to minimize the impact of imbalanced-data and ensure data quantity in the process. Then, the data features are extracted by 2-dimensional convolutional neural network (2dCNN), and the effect of useless information is reduced by mean pooling and maximum pooling, so it can be adapted to the demanding resource environment of the IoT. On this basis, long-range dependent temporal features are extracted using bidirectional long short-term memory (BiLSTM), which aims to fully extract data features to improve detection accuracy in the limited resource environment. Finally, the algorithm is validated on the UNSW_NB15 dataset, and the results of the experiments reaches 93.5% at Accuracy, 86.4% at Precision, 85.3% at Recall and 85.8% at F1-Score. According to the results, the proposed algorithm can generate higher-quality samples, achieve higher detection rate with faster inference time and spend lower memory costs.
- ItemInterrupted Sampling Repeater Jamming Suppression with Pulse Doppler Radar Using Linear Interpulse Frequency Coding(Společnost pro radioelektronické inženýrství, 2024-06) Guo, Y.; Zhou, D.; Ding, Y.Interrupted sampling repeater jamming (ISRJ) is an advanced coherent jamming, and the suppression of this jamming has become a critical problem for modern radar electronic countermeasure. In this paper, we propose a countermeasure based on linear interpulse frequency-coding linear frequency modulation (LIFC-LFM) signal. The LIFC refers to the linear encoding of the frequency of each pulse transmitted by the radar system, which can change the distribution of the false targets formed by ISRJ in the range-Doppler (RD) spectrum. In this context, we design the frequency coding value to effectively separate the true and false targets in the RD spectrum. Furthermore, we propose a fast-time phase compensation method to separate the true and false targets in the Doppler dimension. Finally, ISRJ can be suppressed by oblique projection processing. Simulation examples demonstrate that the proposed method has an excellent and robust ISRJ suppression effect for direct forwarding ISRJ, repeated forwarding ISRJ, and frequency shifting ISRJ. Meanwhile, the signal-to-noise ratio loss caused by the jamming suppression is small.
- «
- 1 (current)
- 2
- 3
- »