Time filter

Source Type

Li X.,Fujitsu Research and Development Center Co. | Liu W.,Fujitsu Research and Development Center Co. | Fan W.,Fujitsu Research and Development Center Co. | Sun J.,Fujitsu Research and Development Center Co. | Satoshi N.,Fujitsu Research and Development Center Co.
International Conference on Signal Processing Proceedings, ICSP | Year: 2017

This paper presents a method for image perspective correction using camera intrinsic parameters. This method is based on two assumptions: a) the taken picture have a rectangle area, but didn't know the rectangle area's aspect ratio; b) the camera's intrinsic parameters should obtain by picture (Intrinsic parameters can easily obtain in iPhone or Android phone). The rectangle's perspective distortion is constrained by camera's intrinsic parameters. So this paper's method find this constraint to get the distorted rectangle's aspect ratio. Then using this ratio can get the homography matrix. This method also can be used to detect card or document rectangle area under complex background. We took some pictures by iPhones and android phones, the results shows that our method can correct document pictures' perspective distortion. © 2016 IEEE.


Xu Z.,Fujitsu Research and Development Center Co. | Hou C.,Fujitsu Research and Development Center Co. | Xia Y.,Fujitsu Research and Development Center Co. | Sun J.,Fujitsu Research and Development Center Co. | And 2 more authors.
IEEE Region 10 Annual International Conference, Proceedings/TENCON | Year: 2017

Data stream mining has gained growing attentions recently. Concept drift is a particular problem in data stream mining, which is defined as the distribution of data may change over time. Most of current methods try to estimate the current distribution or reconstruct the current distribution from a mixture of old distributions. They suffer problems of estimation and reconstruction error respectively. In this paper, we found that a classifier that fits the current distribution can be obtained more directly than the current methods by ensembling classifiers trained with increasing number of recent data. This strategy guarantees that no matter when and how concept drift happens, there is always a classifier that suits the current data distribution. So our method only needs to select the current distribution classifier out of all classifiers we hold. This is much easier than estimation and reconstruction. We test our method on four real world data sets. Comparing with other methods, our method is the best algorithm in terms of average accuracy. © 2016 IEEE.


Kimura Y.,Fujitsu Limited | Naol S.,Fujitsu Limited | Naol S.,Fujitsu Research and Development Center Co. | Nakata T.,Fujitsu Limited
Fujitsu Scientific and Technical Journal | Year: 2016

As the core research and development (R&D) organization driving innovation in the Fujitsu Group, Fujitsu Laboratories (Japan) collaborates globally with three overseas research laboratories: Fujitsu Laboratories of America, Fujitsu Research and Development Center (China), and Fujitsu Laboratories of Europe. These overseas laboratories research and develop key technologies in close collaboration with Fujitsu Laboratories (Japan) while also pursuing standardization activities and original R&D projects reflecting the distinctive features of their regions. They also actively pursue open innovation through tie-ups with leading universities and research institutions in their respective areas. Furthermore, to support Fujitsu's business activities in their regions, they are committed to business incubation as exemplified by their collaboration with startup companies and introduction of advanced technologies through forums and exhibits. This paper introduces the research and development and associated activities at each of these overseas research laboratories, which play an important role in supporting the Fujitsu Group.


Yi W.,Fujitsu Research and Development Center Co. | Hua Z.,Fujitsu Research and Development Center Co. | Jianming W.,Fujitsu Research and Development Center Co.
IWCMC 2012 - 8th International Wireless Communications and Mobile Computing Conference | Year: 2012

Enhanced Physical Downlink Control Channel (EPDCCH) is one of the most important technical improvements in 3GPP LTE-Advanced standards with respect to the new nonuniform network deployments. Frequency diversity as well as frequency selective scheduling gain together with beamforming gain is expected. In order to achieve frequency scheduling gain, the search space of EPDCCH should contain subcarriers associated with good channel quality. In this paper, we propose a modified search space mapping scheme that maximizes the frequency scheduling gain with given frequency resource. Simulation results shows proposed search space function is effective in improving the EPDCCH performance under frequency selective fading channels. Up to 5 dB signal-to-noise ratio (SNR) gain is observed compared with existing search space function. © 2012 IEEE.


Wu C.,Fujitsu Research and Development Center Co. | Fan W.,Fujitsu Research and Development Center Co. | He Y.,Fujitsu Research and Development Center Co. | Sun J.,Fujitsu Research and Development Center Co. | Naoi S.,Fujitsu Research and Development Center Co.
Proceedings - International Conference on Pattern Recognition | Year: 2012

This paper presents a handwritten digit recognition method based on cascaded heterogeneous convolutional neural networks (CNNs). The reliability and complementation of heterogeneous CNNs are investigated in our method. Each CNN recognizes a proportion of input samples with high-confidence, and feeds the rejected samples into the next CNN. The samples rejected by the last CNN are recognized by a voting committee of all CNNs. Experiments on MNIST dataset show that our method achieves an error rate 0.23% using only 5 C-NNs, on par with human vision system. Using heterogeneous networks can reduce the number of CNNs needed to reach certain performance compared with networks built from the same type. Further improvements include fine-tuning the rejection threshold of each CNN and adding CNNs of more types. © 2012 ICPR Org Committee.


Song L.,Fujitsu Research and Development Center Co. | Wang X.,Fujitsu Research and Development Center Co.
2015 IEEE 82nd Vehicular Technology Conference, VTC Fall 2015 - Proceedings | Year: 2015

With massive MIMO and/or millimeter wave technologies, the system has large numbers of antennas and/or frequency bands to increase the capacity and the number of the accessed users. In this case, conventional precoding method performed at the baseband becomes impractical, since the processing complexity is large and the cost of the hardware is high. Therefore, hybrid radio frequency (RF) and baseband (BB) precoding is an alternative to achieve the tradeoff between complexity and performance. In this paper, beam selection methods based on the thought of spatial correlation, variable beamwidth and beam combination are proposed for RF beamforming used in singleuser and multiple-user scenarios, respectively. Simulation results show that the proposed beam selection methods perform well with much lower complexity than brute-force search. © 2015 IEEE.


Wu C.,Fujitsu Research and Development Center Co. | Fan W.,Fujitsu Research and Development Center Co. | He Y.,Fujitsu Research and Development Center Co. | Sun J.,Fujitsu Research and Development Center Co. | Naoi S.,Fujitsu Research and Development Center Co.
Proceedings of International Conference on Frontiers in Handwriting Recognition, ICFHR | Year: 2014

Deep learning methods have recently achieved impressive performance in the area of visual recognition and speech recognition. In this paper, we propose a handwriting recognition method based on relaxation convolutional neural network (R-CNN) and alternately trained relaxation convolutional neural network (ATR-CNN). Previous methods regularize CNN at full-connected layer or spatial-pooling layer, however, we focus on convolutional layer. The relaxation convolution layer adopted in our R-CNN, unlike traditional convolutional layer, does not require neurons within a feature map to share the same convolutional kernel, endowing the neural network with more expressive power. As relaxation convolution sharply increase the total number of parameters, we adopt alternate training in ATR-CNN to regularize the neural network during training procedure. Our previous CNN took the 1st place in ICDAR'13 Chinese Handwriting Character Recognition Competition, while our latest ATR-CNN outperforms our previous one and achieves the state-of-the-art accuracy with an error rate of 3.94%, further narrowing the gap between machine and human observers (3.87%). © 2014 IEEE.


Duan L.,Beijing University of Technology | Qiao H.,Beijing University of Technology | Wu C.,Fujitsu Research and Development Center Co. | Yang Z.,Beijing University of Technology | Ma W.,Beijing University of Technology
Advances in Intelligent Systems and Computing | Year: 2014

We propose a method to predict human saccadic scanpaths on natural images based on a bio-inspired visual attention model. The method integrates three related factors as driven forces to guide eye movements, sequentially visual saliency, winner-takes-all and visual memory, respectively. When predicting a current fixation of saccadic scanpaths, we follow physiological visual memory characteristics to eliminate the effects of the previous selected fixation. Then, we use winner-takes-all to select the fixation on the current saliency map. Experimental results demonstrate that the proposed model outperform other methods on both static fixation locations and dynamic scanpaths. © Springer International Publishing Switzerland 2014.


Wang Y.,Fujitsu Research and Development Center Co. | Zhou H.,Fujitsu Research and Development Center Co. | Wu J.,Fujitsu Research and Development Center Co.
Proceedings - 5th International Conference on Systems and Networks Communications, ICSNC 2010 | Year: 2010

In this paper, an efficient power allocation scheme of reference signals (RS) for uplink MU-MIMO in LTE system is introduced. In LTE system, some virtual subcarriers at the borders of the allocated bandwidth are used which invalidates the conventional power allocation schemes for uniformly placed reference signal. Further, RS for multiuser are multiplexed by cyclic shift. Constant power is required within one step of cyclic shift which adds more constraint on power allocation. The proposed RS power allocation scheme minimizes the mean square error (MSE) of the least square (LS) estimates considering the effect of virtual subcarriers and the constraint due to the cyclic shift. The effectiveness of the proposed scheme is verified by simulation results, showing that 2 dB SNR gain is achieved as opposed to conventional scheme with equal power reference signal. © 2010 IEEE.


Fan W.,Fujitsu Research and Development Center Co. | Sun J.,Fujitsu Research and Development Center Co. | Naoi S.,Fujitsu Research and Development Center Co.
Proceedings of SPIE - The International Society for Optical Engineering | Year: 2015

We describe a document image segmentation algorithm to classify a scanned document into different regions such as text/line drawings, pictures, and smooth background. The proposed scheme is relatively independent of variations in text font style, size, intensity polarity and of string orientation. It is intended for use in an adaptive system for document image compression. The principal parts of the algorithm are the generation of the foreground and background layers and the application of hierarchical singular value decomposition (SVD) in order to smoothly fill the blank regions of both layers so that the high compression ratio can be achieved. The performance of the algorithm, both in terms of its effectiveness and computational efficiency, was evaluated using several test images and showed superior performance compared to other techniques. © 2015 SPIE.

Loading Fujitsu Research and Development Center Co. collaborators
Loading Fujitsu Research and Development Center Co. collaborators