| FU XINGJIAN | M, 1回目発表 | 光メディアインタフェース | 向川 康博, 清川 清, 船冨 卓哉, 田中 賢一郎, 久保 尋之 |
|
title: Camera distortion rectification with convolutional nerual networks
abstract: Camera distortions cause images to deviate from the rectilinear pinhole camera projections. Without the access to camera parameters, these distortions are usually corrected by calibration procedures prior to the shooting. We propose a novel convolutional neural network which estimates distortion parameters on single images. In contrast to existing researches on fisheye lens rectification, our study focuses on general lens distortions and anticipate a diversity of optic distortions. The network was trained with synthetic dataset, including pictures collected from the Internet and CG generated scenes. Our network showed satisfactory responses to both synthetic and real distortions. language of the presentation: English | |||
| IVAN HALIM PARMONANGAN | M, 1回目発表 | 知能コミュニケーション | 中村 哲, 佐藤 嘉伸, 国田 勝行(BS), 田中 宏季, Sakriani Sakti |
|
title: Speech Quality Evaluation of Synthesized Japanese Speech using EEG
abstract: As synthesized speech technology becomes more widely used, the synthesized speech quality must be assessed to ensure that it is acceptable. Subjective evaluation metrics, such as mean opinion score (MOS), can only provide an overall impression without any further detailed information about the speech. Therefore, this study proposes predicting speech quality using electroencephalographs (EEG), which are more objective and have high temporal resolution. In this paper, we use one natural speech and four types of synthesized speech lasting two to six seconds. First, to obtain ground truth of MOS, we gathered ten subjects to give opinion score on a scale of one to five for each recording. Second, another nine subjects were asked to measure how close to natural speech each synthesized speech sounded. The subjects' EEGs were recorded while they were listening to and evaluating the listened speech. The best accuracy achieved for classification was 96.61% using support vector machine, 80.36% using linear discriminant analysis, and 59.9% using logistic regression. For regression, we achieved root mean squared error as low as 1.133 using SVR and 1.353 using linear regression. This study demonstrates that EEG could be used to evaluate the perceived speech quality objectively. language of the presentation: English | |||
| JIA HAOHUI | M, 1回目発表 | ネットワークシステム学 | 岡田 実, 笠原 正治, 東野 武史, Dong Duong Thang, Chen Na |
|
title: A study on Sparse Channel Estimation Based on Compressed Sensing in MIMO-SCFMA
abstract: In the Massive MIMO antennas systems, the process of channel estmation becomes complexity and tedious as the the numbers of received antennas growing . To reduce the the overhead of pilotsymbols in massive MIMO systems, we can use the inherent sparsity of wireless channel,and it is capable of improving the performance of channel estimation. The sparse channel estimation is based on the mutual coherence of MIMO channel's coefficients and recieved signal,which shows the most effect elements in the systems by Compressed Sensing. Furthermore, the channel estimation based on the Compreseed Sensing can improve the accuracy comparing the conventional proposal. language of the presentation: English | |||
| KUO CHENG-YU | M, 1回目発表 | ロボットラーニング | 杉本 謙二, 松原 崇充, 小笠原 司 |
|
title: Fast Gaussian Process Approximation in Moment-Matching for Robotics
abstract: Moment Matching is a long term prediction method with Gaussian Process. As known, Gaussian Process is a powerful regression method but suffer from large scale training size, which will result in massive computing time. By using approximating method to speed up Gaussian Process enable Moment-Matching to apply on systems with bigger scale(or dimensions), for example, high DoF robotics. With fast Moment-Matching predictions, realtime predictions can also be realized. language of the presentation: *** English or Japanese (choose one) *** | |||
| LI WAI HEI | M, 1回目発表 | 自然言語処理学 | 松本 裕治, 中村 哲, 新保 仁, 進藤 裕之 |
|
title:Automatic Generation of Medical Report
abstract: Analysing X-ray images is a very important task in clinic. Doctors and radiologist may spend a lot of their working hour on analying those images and writing a report according to the findings in those images. However it was never a plesant task for them. As learning to analyse those images takes a very long time and it is hard for inexperienced radiologist to anaylse X-ray image while being a repetitve task for doctors. Even worse, there may not enough radiologists to serve all patients in some area. So the possibility of automatic generation of medical report came to interest. language of the presentation: English | |||
| NEPAL SUBODH | D, 中間発表 | ネットワークシステム学 | 岡田 実, 杉本 謙二, 東野 武史, Duong Quang Thang |
|
title: Radio coverage prediction and optimization technique for digital terrestrial television broadcasting system.
abstract: An approach of predicting and optimizing the radio coverage of Multi-Frequency Network (MFN) with Digital Video Broadcasting-Second Generation Terrestrial (DVB-T2) system is proposed in this work. The coverage of a nationwide Multi-Frequency Network (MFN) with Digital Video Broadcasting-Second Generation Terrestrial (DVB-T2) system implemented in Nepal is maximized using the optimum values of transmit frequencies, power levels and antenna parameters. The propagation prediction method recommended in the ITU-R P.1812 with Digital Elevation Model (DEM) has been used to predict signal strength at each of the receiving location. Based on the received signal strength and standard Quality of Service (QoS) values as key factors for determining radio coverage, Genetic Algorithm (GA) is used to optimize the network. The results show that the optimized allocation of power levels, frequencies and antenna patterns to the group of transmitters significantly increases both the geographical coverage and population coverage. Hence, the overall results imply the effectiveness of the method to be used for optimizing the broadcasting network for a smooth transition from analogue to digital. Further, the co channel interference among the radio transmitter sharing the same transmit frequency is also evaluated and analyzed for different cases of transmit frequency and number of spot frequencies allocated. language of the presentation: English | |||