Tomohiro WARASHINA Kazuo AOYAMA Hiroshi SAWADA Takashi HATTORI
This paper presents an efficient method using Hadoop MapReduce for constructing a K-nearest neighbor graph (K-NNG) from a large-scale data set. K-NNG has been utilized as a data structure for data analysis techniques in various applications. If we are to apply the techniques to a large-scale data set, it is desirable that we develop an efficient K-NNG construction method. We focus on NN-Descent, which is a recently proposed method that efficiently constructs an approximate K-NNG. NN-Descent is implemented on a shared-memory system with OpenMP-based parallelization, and its extension for the Hadoop MapReduce framework is implied for a larger data set such that the shared-memory system is difficult to deal with. However, a simple extension for the Hadoop MapReduce framework is impractical since it requires extremely high system performance because of the high memory consumption and the low data transmission efficiency of MapReduce jobs. The proposed method relaxes the requirement by improving the MapReduce jobs, which employs an appropriate key-value pair format and an efficient sampling strategy. Experiments on large-scale data sets demonstrate that the proposed method both works efficiently and is scalable in terms of a data size, the number of machine nodes, and the graph structural parameter K.
Takuma WATANABE Hiroyoshi YAMADA Motofumi ARII Ryoichi SATO Sang-Eun PARK Yoshio YAMAGUCHI
Soil moisture retrieval from polarimetric synthetic aperture radar (SAR) imagery over forested terrain is quite a challenging problem, because the radar backscatter is affected by not only the moisture content, but also by large vegetation structures such as the trunks and branches. Although a large number of algorithms which exploit radar backscatter to infer soil moisture have been developed, most of them are limited to the case of bare soil or little vegetation cover that an incident wave can easily reach the soil surface without serious disturbance. However, natural land surfaces are rarely free from vegetation, and the disturbance in radar backscatter must be properly compensated to achieve accurate soil moisture measurement in a diversity of terrain surfaces. In this paper, a simple polarimetric parameter, co-polarized backscattering ratio, is shown to be a criterion to infer moisture content of forested terrain, from both a theoretical forest scattering simulation and an appropriate experimental validation under well-controlled condition. Though modeling of forested terrain requires a number of scattering mechanisms to be taken into account, it is essential to isolate them one by one to better understand how soil moisture affects a specific and principal scattering component. For this purpose, we consider a simplified microwave scattering model for forested terrain, which consists of a cloud of dielectric cylinders as a representative of trunks, vertically stood on a flat dielectric soil surface. This simplified model can be considered a simple boreal forest model, and it is revealed that the co-polarization ratio in the ground-trunk double-bounce backscattering can be an useful index to monitor the relative variation in the moisture content of the boreal forest.
Masahiro NISHI Koichi SHIN Teruaki YOSHIDA
In the digital terrestrial TV broadcasting system, it is important to evaluate both quantitative levels and sources of overreach interference, because it can degrade the TV service quality. This paper newly proposes an overreach measurement method that simultaneously monitors RSSI (Received Signal Strength Indicator) and CNR (Carrier to Noise power Ratio) of the TV waves and RSSI of FM waves. The results of measurements conducted in Hiroshima prefecture show that our proposed method can evaluate the level of overreach interference in the TV waves and also identify the source of the interference. Total 43 overreach interference events were found in the proposed method from one-year measurement in 2012. Based on M profile data, this paper also shows that the main factor of the overreach interference in this measurement is duct propagation due to meteorological condition.
Jarich VANSTEENBERGE Masayuki MUKUNOKI Michihiko MINOH
The Hough voting framework is a popular approach to parts based pedestrian detection. It works by allowing image features to vote for the positions and scales of pedestrians within a test image. Each vote is cast independently from other votes, which allows for strong occlusion robustness. However this approach can produce false pedestrian detections by accumulating votes inconsistent with each other, especially in cluttered scenes such as typical street scenes. This work aims to reduce the sensibility to clutter in the Hough voting framework. Our idea is to use object segmentation and object pose parameters to enforce votes' consistency both at training and testing time. Specifically, we use segmentation and pose parameters to guide the learning of a pedestrian model able to cast mutually consistent votes. At test time, each candidate detection's support votes are looked upon from a segmentation and pose viewpoints to measure their level of agreement. We show that this measure provides an efficient way to discriminate between true and false detections. We tested our method on four challenging pedestrian datasets. Our method shows clear improvements over the original Hough based detectors and performs on par with recent enhanced Hough based detectors. In addition, our method can perform segmentation and pose estimation as byproducts of the detection process.
Takeshi USUI Yoshinori KITATSUJI Hidetoshi YOKOTA Kiyohide NAKAUCHI Yozo SHOJI Nozomu NISHINAGA
It is known that the IP Multimedia Subsystem (IMS) provides various telecom services e.g., VoIP, instant messaging, and video conferencing. In the IMS, these services are provided with session initiation protocol (SIP) handled by call/session control function servers (CSCFs). Completing the SIP signaling call flow without delay is vital to provide satisfactory services to the users. For service continuity, previous studies redundantly synchronized session states of CSCFs with backup servers. This paper proposes an IMS restoration system that selectively stores the session states. This is achieved by utilizing the retransmission mechanism of SIP. Time-based simulation emulating the process of backup servers shows that the proposed system reduces the number of backup servers to less than 38% compared to the previous studies, without degrading the service quality.
Bo HAO Jun WANG Zhaocheng WANG
This paper presents an efficient multi-service allocation scheme for the digital television terrestrial broadcasting systems in which the fixed service is modulated by orthogonal frequency division multiplexing and quadrature amplitude modulation (OFDM/QAM) with larger FFT size and the added mobile service is modulated by OFDM and offset quadrature amplitude modulation (OQAM) with smaller FFT size. The two different types of services share one 8MHz broadcasting channel. The isotropic orthogonal transform algorithm (IOTA) is chosen as the shaping filter for OQAM because of its isotropic convergence in time and frequency domain and the proper FFT size is selected to maximum the transmission capacity under mobile environment. The corresponding transceiver architecture is also proposed and analyzed. Simulations show that the newly added mobile service generates much less out-of-band interference to the fixed service and has a better performance under fast fading wireless channels.
Kriangkrai LIMTHONG Kensuke FUKUDA Yusheng JI Shigeki YAMADA
Detecting a variety of anomalies caused by attacks or accidents in computer networks has been one of the real challenges for both researchers and network operators. An effective technique that could quickly and accurately detect a wide range of anomalies would be able to prevent serious consequences for system security or reliability. In this article, we characterize detection techniques on the basis of learning models and propose an unsupervised learning model for real-time anomaly detection in computer networks. We also conducted a series of experiments to examine capabilities of the proposed model by employing three well-known machine learning algorithms, namely multivariate normal distribution, k-nearest neighbor, and one-class support vector machine. The results of these experiments on real network traffic suggest that the proposed model is a promising solution and has a number of flexible capabilities to detect several types of anomalies in real time.
Byoung-Kwang KIM Meiguang JIN Woo-Jin SONG
In this paper, we propose a new matting algorithm using local and nonlocal neighbors. We assume that K nearest neighbors satisfy the color line model that RGB distribution of the neighbors is roughly linear and combine this assumption with the local color line model that RGB distribution of local neighbors is roughly linear. Our assumptions are appropriate for various regions such as those that are smooth, contain holes or have complex color. Experimental results show that the proposed method outperforms previous propagation-based matting methods. Further, it is competitive with sampling-based matting methods that require complex sampling or learning methods.
Toru NAKASHIKA Tetsuya TAKIGUCHI Yasuo ARIKI
This paper presents a voice conversion technique using speaker-dependent Restricted Boltzmann Machines (RBM) to build high-order eigen spaces of source/target speakers, where it is easier to convert the source speech to the target speech than in the traditional cepstrum space. We build a deep conversion architecture that concatenates the two speaker-dependent RBMs with neural networks, expecting that they automatically discover abstractions to express the original input features. Under this concept, if we train the RBMs using only the speech of an individual speaker that includes various phonemes while keeping the speaker individuality unchanged, it can be considered that there are fewer phonemes and relatively more speaker individuality in the output features of the hidden layer than original acoustic features. Training the RBMs for a source speaker and a target speaker, we can then connect and convert the speaker individuality abstractions using Neural Networks (NN). The converted abstraction of the source speaker is then back-propagated into the acoustic space (e.g., MFCC) using the RBM of the target speaker. We conducted speaker-voice conversion experiments and confirmed the efficacy of our method with respect to subjective and objective criteria, comparing it with the conventional Gaussian Mixture Model-based method and an ordinary NN.
Daisuke OCHI Hideaki KIMATA Yoshinori KUSACHI Kosuke TAKAHASHI Akira KOJIMA
Due to the recent progress made in camera and network environments, on-line video services enable people around the world to watch or share high-quality HD videos that can record a wider angle without losing objects' details in each image. As a result, users of these services can watch videos in different ways with different ROIs (Regions of Interest), especially when there are multiple objects in a scene, and thus there are few common ways for them to transfer their impressions for each scene directly. Posting messages is currently the usual way but it does not sufficiently enable all users to transfer their impressions. To transfer a user's impressions directly and provide users with a richer video watching experience, we propose a system that enables them to extract their favorite parts of videos as ROI trajectories through simple and intuitive manipulation of their tablet device. It also enables them to share a recorded trajectory with others after stabilizing it in a manner that should be satisfactory to every user. Using statistical analysis of user manipulations, we have demonstrated an approach to trajectory stabilization that can eliminate undesirable or uncomfortable elements due to tablet-specific manipulations. The system's validity has been confirmed by subjective evaluations.
Youwen ZHU Tsuyoshi TAKAGI Rong HU
Recently, Yuan et al. (IEEE Infocom'13, pp.2652-2660) proposed an efficient secure nearest neighbor (SNN) search scheme on encrypted cloud database. Their scheme is claimed to be secure against the collusion attack of query clients and cloud server, because the colluding attackers cannot infer the encryption/decryption key. In this letter, we observe that the encrypted dataset in Yuan's scheme can be broken by the collusion attack without deducing the key, and present a simple but powerful attack to their scheme. Experiment results validate the high efficiency of our attacking approach. Additionally, we also indicate an upper bound of collusion-resistant ability of any accurate SNN query scheme.
Baeksop KIM Jiseong KIM Jungmin SO
This letter presents a scheme to improve the running time of exemplar-based image inpainting, first proposed by Criminisi et al. In the exemplar-based image inpainting, a patch that contains unknown pixels is compared to all the patches in the known region in order to find the best match. This is very time-consuming and hinders the practicality of Criminisi's method to be used in real time. We show that a simple bounding algorithm can significantly reduce number of distance calculations, and thus the running time. Performance of the bounding algorithm is affected by the order of patches that are compared, as well as the order of pixels in a patch. We present pixel and patch ordering schemes that improve the performance of bounding algorithms. Experiments with well-known images used in inpainting literature show that the proposed reordering scheme can reduce running time of the bounding algorithm up to 50%.
Jin-Ping HE Kun GAO Guo-Qiang NI Guang-Da SU Jian-Sheng CHEN
Considering the real existent fact of the ideal edge and the learning style of image analogy without reference parameters, a blind image recovery algorithm using a self-adaptive learning method is proposed in this paper. We show that a specific local image patch with degradation characteristic can be utilized for restoring the whole image. In the training process, a clear counterpart of the local image patch is constructed based on the ideal edge assumption so that identification of the Point Spread Function is no longer needed. Experiments demonstrate the effectiveness of the proposed method on remote sensing images.
Kenshi SAHO Takuya SAKAMOTO Toru SATO Kenichi INOUE Takeshi FUKUDA
The classification of human motion is an important aspect of monitoring pedestrian traffic. This requires the development of advanced surveillance and monitoring systems. Methods to achieve this have been proposed using micro-Doppler radars. However, reliable long-term data and/or complicated procedures are needed to classify motion accurately with these conventional methods because their accuracy and real-time capabilities are invariably inadequate. This paper proposes an accurate and real-time method for classifying the movements of pedestrians using ultra wide-band (UWB) Doppler radar to overcome these problems. The classification of various movements is achieved by extracting feature parameters based on UWB Doppler radar images and their radial velocity distributions. Experiments were carried out assuming six types of pedestrian movements (pedestrians swinging both arms, swinging only one arm, swinging no arms, on crutches, pushing wheelchairs, and seated in wheelchairs). We found they could be classified using the proposed feature parameters and a k-nearest neighbor algorithm. A classification accuracy of 96% was achieved with a mean calculation time of 0.55s. Moreover, the classification accuracy was 99% using our proposed method for classifying three groups of pedestrian movements (normal walkers, those on crutches, and those in wheelchairs).
To develop an envelope-tracking (ET) amplifier for next-generation cellular base stations, we propose a wideband envelope modulator, consisting of a linear-mode class-B amplifier and a switch-mode class-D amplifier. The function of the modulator is to track the envelope signal and supply voltage to an RF amplifier. To meet the requirements of a large-current and high-voltage supply that can handle a wideband signal, an “Alexander current-feedback amplifier topology” is applied to the linear-mode class-B amplifier. The Alexander topology not only boosts the voltage but also enhances the current capacity of a commercial high-speed operational amplifier (op-amp) by means of a push-pull stage with current mirrors and a buffer amplifier at the output of the op-amp. With this topology, a linear-mode amplifier can provide several-ampere-level current to a 11-Ω load. A prototype of the wideband envelope modulator is shown to achieve the efficiency of 71% with a 20-MHz WiMAX envelope signal at output power of 72W.
Tran Lan Anh NGUYEN Gueesang LEE
Segmenting indicated objects from natural color images remains a challenging problem for researches of image processing. In this paper, a novel level set approach is presented, to address this issue. In this segmentation algorithm, a contour that lies inside a particular region of the concerned object is first initialized by a user. The level set model is then applied, to extract the object of arbitrary shape and size containing this initial region. Constrained on the position of the initial contour, our proposed framework combines two particular energy terms, namely local and global energy, in its energy functional, to control movement of the contour toward object boundaries. These energy terms are mainly based on graph partitioning active contour models and Bhattacharyya flow, respectively. Its flow describes dissimilarities, measuring correlative relationships between the region of interest and surroundings. The experimental results obtained from our image collection show that the suggested method yields accurate and good performance, or better than a number of segmentation algorithms, when applied to various natural images.
Dubok PARK David K. HAN Changwon JEON Hanseok KO
Images captured under foggy conditions often exhibit poor contrast and color. This is primarily due to the air-light which degrades image quality exponentially with fog depth between the scene and the camera. In this paper, we restore fog-degraded images by first estimating depth using the physical model characterizing the RGB channels in a single monocular image. The fog effects are then removed by subtracting the estimated irradiance, which is empirically related to the scene depth information obtained, from the total irradiance received by the sensor. Effective restoration of color and contrast of images taken under foggy conditions are demonstrated. In the experiments, we validate the effectiveness of our method compared with conventional method.
Kazuya TAKAHASHI Tatsuya MORI Yusuke HIROTA Hideki TODE Koso MURAKAMI
In recent years, real-time streaming has become widespread as a major service on the Internet. However, real-time streaming has a strict playback deadline. Application level multicasts using multiple distribution trees, which are known as forests, are an effective approach for reducing delay and jitter. However, the failure or departure of nodes during forest-based multicast transfer can severely affect the performance of other nodes. Thus, the multimedia data quality is degraded until the distribution trees are repaired. This means that increasing the speed of recovery from isolation is very important, especially in real-time streaming services. In this paper, we propose three methods for resolving this problem. The first method is a random-based proactive method that achieves rapid recovery from isolation and gives efficient “Randomized Forwarding” via cooperation among distribution trees. Each node forwards the data it receives to child nodes in its tree, and then, the node randomly transferring it to other trees with a predetermined probability. The second method is a reactive method, which provides a reliable isolation recovery method with low overheads. In this method, an isolated node requests “Continuous Forwarding” from other nodes if it detects a problem with a parent node. Forwarding to the nearest nodes in the IP network ensures that this method is efficient. The third method is a hybrid method that combines these two methods to achieve further performance improvements. We evaluated the performances of these proposed methods using computer simulations. The simulation results demonstrated that our proposed methods delivered isolation recovery and that the hybrid method was the most suitable for real-time streaming.
Scheduling restriction is attracting much attention in LTE-Advanced as a technique to reduce the power consumption and network overheads in interference coordinated heterogeneous networks (HetNets). Such a network with inter-cell interference coordination (ICIC) provides two radio resources with different channel quality statistics. One of the resources is protected (unprotected) from inter-cell interference (hence, called protected (non-protected) resource) and has higher (lower) average channel quality. Without scheduling restriction, the channel quality feedback would be doubled to reflect the quality difference of the two resources. We present a simple scheduling restriction scheme that addresses the problem without significant performance degradation. Users with relatively larger (smaller) average channel quality difference between the two resources are scheduled in the protected (non-protected) resource only, and a boundary user, determined by a proportional fair resource allocation (PFRA) under simplified static channels, is scheduled on one of the two resources or both depending on PFRA. Having most users scheduled in only one of the resources, the power consumption and network overheads that would otherwise be required for the channel quality feedback on the other resource can be avoided. System level simulation of LTE-Advanced downlink shows that the performance degradation due to our scheduling restriction scheme is less than 2%, with the average feedback reduction of 40%.
Shizue NAGAHARA Takenori OIDA Tetsuo KOBAYASHI
Diffusion-weighted (DW)-functional magnetic resonance imaging (fMRI) is a recently reported technique for measuring neural activities by using diffusion-weighted imaging (DWI). DW-fMRI is based on the property that cortical cells swell when the brain is activated. This approach can be used to observe changes in water diffusion around cortical cells. The spatial and temporal resolutions of DW-fMRI are superior to those of blood-oxygenation-level-dependent (BOLD)-fMRI. To investigate how the DWI signal intensities change in DW-fMRI measurement, we carried out Monte Carlo simulations to evaluate the intensities before and after cell swelling. In the simulations, we modeled cortical cells as two compartments by considering differences between the intracellular and the extracellular regions. Simulation results suggested that DWI signal intensities increase after cell swelling because of an increase in the intracellular volume ratio. The simulation model with two compartments, which respectively represent the intracellular and the extracellular regions, shows that the differences in the DWI signal intensities depend on the ratio of the intracellular and the extracellular volumes. We also investigated the MPG parameters, b-value, and separation time dependences on the percent signal changes in DW-fMRI and obtained useful results for DW-fMRI measurements.