http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
Efficient Deep Learning Model for Data-Limited Modulation Recognition
Judith Nkechinyere Njoku(주디스),Angela Caliwag(안젤라),Wansu Lim(임완수) 한국통신학회 2021 한국통신학회 학술대회논문집 Vol.2021 No.6
Deep learning (DL) has been successfully applied for modulation recognition tasks. Most of the existing applications pay little attention to the volume of data used. In reality, real historic data for modulation recognition could be limited. Thus it is better to train recognition models on small amounts of data. This is however a huge challenge, since the performance of DL models depend on sufficient data. In this paper, we introduce an efficient system model based on CNN and different data augmentation methods, for the purpose of modulation recognition. Our system employs random rotation, flip, zoom, random shift and resize methods for data augmentation. The CNN model employs small filter sizes, pooling layers, and dropout layers to improve the network. We apply a small dataset consisting of 40 constellation images per modulation type, to the system. We further analyze the performance based on three data augmentation intervals. From our experiments, the model achieved an accuracy of 32% without data augmentation and 76.07%, 35.71% and 96.42% on the three data-augmentation intervals.
Automatic Radar Waveform Recognition using the Wigner-Ville distribution and AlexNet-SVM
Njoku Judith Nkechinyere(주디스),Manuel Eugenio Morocho-Cayamce(유제니오),Wansu Lim(임완수) 한국통신학회 2020 한국통신학회 학술대회논문집 Vol.2020 No.8
In this paper, we propose a radar signal modulation algorithm to recognize three different radar signals amidst other wireless communication waveforms, including Barker, linear frequency modulation, and rectangular codes. First we extract the features of the original signal by computing its smoothed pseudo Wigner-Ville distribution. Second, we construct a transfer learning-based convolutional neural network over AlexNet to further extract features from the time-frequency images. Finally, a support vector machine classifier is applied for the signal classification. We also perform a similar analysis with models which use AlexNet for both feature extraction, and classification tasks. Results show that the proposed model which incorporates the linear classifier, achieves the highest recognition accuracy of 97.8%.
Hunger marketing and Blockchain Technology: Applications in Wireless Spectrum Management
Njoku Judith Nkechinyere(운저구 주디스 인게친녜레),Manuel Eugenio Morocho-Cayamce(마누엘 에우제니오 모로초 카얌셀라),Wansu Lim(임완수) 한국통신학회 2019 한국통신학회 학술대회논문집 Vol.2019 No.11
Due to the ever-increasing demand by bandwidth-hungry mobile applications and the prevalent growth in wireless communication, effective spectrum management continues to constitute an important issue. So many spectrum management techniques have been employed in different areas including broadband satellite systems, cognitive acoustic networks, railway cognitive radio networks, and smart grid network environments. Spectrum management mechanisms have evolved to meet the different requirements of increasing spectrum use efficiency. In this paper, we discuss two state of the art approaches for spectrum management: Hunger marketing and Blockchain technology. We summarize the pros and cons of these technologies and their application in spectrum management.