2023/2

Browse

Recent Submissions

Now showing 1 - 5 of 11
  • Item
    Optimization of the Crest Factor for Complex-Valued Multisine Signals
    (Společnost pro radioelektronické inženýrství, 2023-06) Cseppento, B.; Retzler, A.; Kollar, Z.
    Multisine signals are commonly used in the measurement of dynamic systems and wireless channels. For optimal measurements with a high dynamic range, a low Crest Factor (CF) excitation signal is required. In this paper, a modified approach to optimize the crest factor for complex-valued multisine signals is presented. The approach uses a nonlinear optimization method where the real and imaginary parts can also be optimized for low CF. Furthermore, extensions of the real-valued multisine CF optimization methods are presented for complex-valued cases. The proposed methods are validated and compared using simulations. Based on the results it is shown that the novel approach can lead to more optimal signal design and lower CF compared to other techniques for complex-valued multisine signals.
  • Item
    Machine Learning Based Classification of IoT Traffic
    (Společnost pro radioelektronické inženýrství, 2023-06) Velichkovska, B.; Cholakoska, A.; Atanasovski, V.
    With the rapid expansion and widespread adoption of the Internet of Things (IoT), maintaining secure connections among active devices can be challenging. Since IoT devices are limited in power and storage, they cannot perform complex tasks, which makes them vulnerable to different types of attacks. Given the volume of data generated daily, detecting anomalous behavior can be demanding. However, machine learning (ML) algorithms have proven successful in extracting complex patterns from big data, which has led to active applications in IoT. In this paper, we perform a comprehensive analysis, including 4 ML algorithms and 3 neural networks (NNs), and propose a pipeline which analyzes the influence data reduction (loss) has on the performance of these algorithms. We use random undersampling as a data reduction technique, which simulates reduced network traffic data. The pipeline investigates several degrees of data loss. The results show that models trained on the original data distribution obtain accuracy that verges on 100%. XGBoost performs best from the classic ML algorithms. From the deep learning models, the 2-layered NN provides excellent results and has sufficient depth for practical application. On the other hand, when the models are trained on the undersampled data, there is a decrease in performance, most notably in the case of NNs. The most prominent change is seen in the 4-layered NN, where the model trained on the original dataset detects attacks with a success of 93.53%, whereas the model trained on the maximally reduced data has a success of only 39.39%.
  • Item
    Radar-Based Human Motion Recognition by Using Vital Signs with ECA-CNN
    (Společnost pro radioelektronické inženýrství, 2023-06) Chen, K.; Gu, M.; Chen, Z.
    Radar technologies reserve a large latent capacity in dealing with human motion recognition (HMR). For the problem that it is challenging to quickly and accurately classify various complex motions, an HMR algorithm combing the attention mechanism and convolution neural network (ECA-CNN) using vital signs is proposed. Firstly, the original radar signal is obtained from human chest wall displacement. Chirp-Z Transform (CZT) algorithm is adopted to refine and amplify the narrow band spectrum region of interest in the global spectrum of the signal, and accurate information on the specific band is extracted. Secondly, six time-domain features were extracted for the neural network. Finally, an ECA-CNN is designed to improve classification accuracy, with a small size, fast speed, and high accuracy of 98%. This method can improve the classification accuracy and efficiency of the network to a large extent. Besides, the size of this network is 100 kb, which is convenient to integrate into the embedded devices.
  • Item
    An Improved Latch for SerDes Interface: Design and Analysis under PVT and AC Noise
    (Společnost pro radioelektronické inženýrství, 2023-06) Kumar, M.; Mondal, A. J.
    Digital subsystem prefers CMOS process, but it is difficult to manage speed and average power (Pavg) trade-off in each era with power supply voltage (Vdd) scaling. Current mode logic (CML) has emerged as an alternative to design the fundamental block of a SerDes, namely, the latch. However, available CML circuits consume significant Pavg and suffer from rapid input slewing. Typically, fast switching inputs enable current flow to effective supply voltage VP and overcharges output. In fact, VP is different than externally applied Vdd and oscillates with time as and when an abrupt current is drawn. This affects delay td and introduces jitter. The topic presents a new latch for SerDes interface using a new current steering circuit and coupled to a power delivery network (PDN). The significant point is to attain an almost constant td in comparison to conventional designs while the Vdd changes. The post-layout results at 0.09-μm CMOS and 1.1 V Vdd indicate that the Pavg and td are 339.5 µW and 61.9 ps, respectively, at 27OC. Surprisingly, the td variation is noted to be minimum and the power supply noise induced jitter is around 1.5 ns when VP close to the circuit varies due to sudden current.
  • Item
    Joint PHD Filter and Hungarian Assignment Algorithm for Multitarget Tracking in Low Signal-to-Noise Ratio
    (Společnost pro radioelektronické inženýrství, 2023-06) Xiao, S.; Tao,H; Shen, X.; Zhang, L.; Hu, M.
    Multitarget tracking (MTT) for image processing in low signal-to-noise ratio (SNR) is difficult and computationally expensive because the distinction between the target and the background is small. Among the current MTT algorithms, Random Finite Set (RFS) based filters are computationally tractable. However, the probability hypothesis density (PHD) filter, despite its low computational complexity, is not suitable for MTT in low SNR. The generalized labeled multi-Bernoulli (GLMB) filter and its fast implementation are unsuitable for realtime MTT due to their high computational complexity. To achieve realtime MTT in low SNR, a joint PHD filter and Hungarian assignment algorithm is first proposed in this work. The PHD filter is used for preliminary tracking of targets while the Hungarian assignment algorithm is employed to complete the association process. To improve the tracking performance in low SNR, a new track must undergo a trial period and a valid track will be terminated only if it is not detected for several frames. The simulation results show that the proposed MTT algorithm can achieve stable tracking performance in low SNR with small computational complexity. The proposed filter can be applied to MTT in low SNR that require realtime implementation.