Hosei University Repository
Not a member yet
22008 research outputs found
Sort by
A Multi-bit ΔΣ Down-converting ADC with Even-Harmonic Mixer and Mismatch Shaper
Get PDFThis paper proposes a multi-bit ΔΣ down-converting analog-to-digital converter (DC-ADC) using an even harmonic mixer (EHMIX) and mismatch shaper to realize software radios and digital radio frequencies. This converter achieves high signal-to-noise-and-distortion ratio (SNDR) and contributes to the realization of high-speed, low-power wireless communication systems. For conversion errors caused by element variation (mismatch) in multi-bit digital-to-analog converters (DAC) , mismatch shapers called data weighted averaging (DWA) and noise shaping dynamic element matching (NSDEM) are used. In this paper, we propose multi-bit ΔΣ DC-ADC with EHMIX and mismatch shaper. Simulink estimated SNDR improved by -26dB in 4-bit 1st-order ΔΣ DC-ADC with EHMIX and DWA -38dB in 4-bit 2nd-order ΔΣ DC-ADC with EHMIX and NSDEM
BASIC RESEARCH ON DEVELOPMENT OF THE GENE NETWORK SIMULATOR WITH STOCHASTIC NONLINEAR SEQUENTIAL CIRCUIT
Get PDFIn this paper, for the development of a hardware-based gene network simulator, nonlinear stochastic circuit gene network (p53 – Mdm2 network, and Hes1 protein – Hes1 mRNA network) models are proposed. The proposed models can reproduce typical nonlinear phenomena of the target gene network models. Also, theoretical analysis methods of the proposed model are presented. Furthermore, it is then revealed that the proposed models can be implemented with fewer circuit elements and operated with lower energy than the straightforward numerical integration models
Scale Estimation of Monocular Camera SLAM using a 2D Lidar
Get PDFSLAM using a monocular camera has the problem that scale cannot be uniquely estimated, and scale drift occurs. This study uses 2D Lidar scan data to improve scale drift and enable scale estimation. Specifically, we improve scale estimation and scale drift by averaging the ratio of distances between the 3D point cloud reconstructed by monocular SLAM and the 2D Lidar point cloud by mapping the data based on the viewing direction. The accuracy is further improved by following the concept of M-estimation
Preparation of colloidal Si nanocrystal embedded LiCl crystal powder
Get PDFColloidal silicon nanocrystals (Si-nc) were embedded into LiCl powder by re-crystallization method in solution. The maximum value of Si-nc concentration is ~3% with respect to LiCl weight. The luminescence properties of the Si-nc embedded LiCl crystal powder were investigated. We found that the recombination rates of Si-nc in LiCl decreased compared to that in soluiton, indicating the suppression of non-radiative rate
PHASOR DIAGRAM ANALYSIS OF TWO-LAYER ELECTRODE RECEIVER IN INTRA-BODY COMMUNICATION
Get PDFThis paper describes the phasor diagram analysis of two-layer electrode receiver in Intra-body communication (IBC). The two-layer electrode of a receiver can function as a differential detection that allows reduction of the radiated noise. The noise reduction mechanism can be analyzed using phasor diagram. As a experimental result, the maximum signal-to-noise ratio could be obtained by adjusting the noise to a minimum. we confirmed that phasor diagram analysis of two-layer electrode receiver were effective tool for researching stable communication
IMAGE SYHTHESIS FOR MANGA BASED ON LANDMARKS
Get PDFWith the goal of creating a Manga drawing assistance system using automatic generation techniques based on Deep Learning, this paper describes the construction of a system that generates Manga characters. I propose a method to generate a cartoon image from random noise and landmark information and a method to generate a cartoon image from the original image, its landmark information and target landmark information. And, I created Manga image datasets for each method
FEATURE ANALYSIS OF SOCIAL MEDIA AND PROPOSAL OF PLATFORMS WITH RELIABLE INFORMATION
Get PDFIn recent years, with the increase in Internet usage, people have more opportunities to obtain information using social media. However, problems such as hoaxes, fake news, and flames occur on social media, and the reliability of the information posted is not guaranteed. In this research, we focus on the loss of credibility of social media and propose a new social media platform using blockchain
Object Detection by Ontology-Based Hierarchy
Get PDFConventional object detection is designed to have a single output. This is based on the assumption that all classes are treated exclusively. However, each class may be divided into hierarchical structures based on ontology. In this case, some classes may be more difficult to identify than others. Also in recent years, AI applications have been introduced in various fields. However, as the performance of AI is improved, the complexity of the model increased, making it difficult to explain the basis of the model’s decisions. In this paper, based on SSD, which is an object detection model, we divide the network into a hierarchical structure based on ontology, and detect both abstract and more detailed concepts to improve the detection rate and explainability. In this experiment, 11540 trainval data of Pascal VOC2007 and Pascal VOC2012 are used for learning, and the model is evaluated with test data of Pascal VOC2007
Application of Tomosynthesis Technology in Non-Destructive Inspection Imaging
Get PDFIn non-destructive inspection of substrates, a technique called X-ray tomosynthesis method has made it possible to inspect sliced layers. But inspection of precision machine substrate by a non-destructive inspection may miss small defective parts by the effect of pins for cooling process. The purpose of this study is to eliminate the effect of pins in substrate inspection and to improve image quality. The paper reviews the results of a rolling ball method, simple calculation method, band pass filter method, and deep image prior for removing pins. In addition, this paper also proposed that small defectives areas be made higher resolution by a super-resolution technique. Comparing the resultant images, the band-pass filtered reconstructed images showed the best results. In addition, super-resolution by machine learning was also found to be effective in improving image quality
Classification of Datasets Based on Combination Algorithm of Clustering and Neural Network
Get PDFWith the development of the data era, data’s importance is increasing daily. For data analysis, the accuracy of data classification is again the basis to ensure the data analysis is carried out smoothly. Therefore, in order to improve the accuracy of the classification of unknown data, this paper proposes a new approach based on combinatorial clustering and neural network algorithm. The results show that the algorithm combining clustering algorithm and neural network will help to improve the accuracy of data classification and identification effectively