Linear Discriminant Analysis (LDA) is a well-known feature extraction method for supervised subspace learning in statistical pattern recognition. In this paper, a novel method of LDA based on a new L1-norm optimization technique and its variances are proposed. The conventional LDA, which is based on L2-norm, is sensitivity to the presence of outliers, since it used the L2-norm to measure the between-class and within-class distances. In addition, the conventional LDA often suffers from the so-called small sample size (3S) problem since the number of samples is always smaller than the dimension of the feature space in many applications, such as face recognition. Based on L1-norm, the proposed methods have several advantages, first they are robust to outliers because they utilize the L1-norm, which is less sensitive to outliers. Second, they have no 3S problem. Third, they are invariant to rotations as well. The proposed methods are capable of reducing the influence of outliers substantially, resulting in a robust classification. Performance assessment in face application shows that the proposed approaches are more effectiveness to address outliers issue than traditional ones.
Yaohua WANG Shuming CHEN Hu CHEN Jianghua WAN Kai ZHANG Sheng LIU
The efficiency of ubiquitous SIMD (Single Instruction Multiple Data) media processors is seriously limited by the bottleneck effect of the scalar kernels in media applications. To solve this problem, a dual-core framework, composed of a micro control unit and an instruction buffer, is proposed. This framework can dynamically decouple the scalar and vector pipelines of the original single-core SIMD architecture into two free-running cores. Thus, the bottleneck effect can be eliminated by effectively exploiting the parallelism between scalar and vector kernels. The dual-core framework achieves the best attributes of both single-core and dual-core SIMD architectures. Experimental results exhibit an average performance improvement of 33%, at an area overhead of 4.26%. What's more, with the increase of the SIMD width, higher performance gain and lower cost can be expected.
Kang Yong LEE Jinsul KIM Kee Seong CHO Won RYU Ho-Jin LEE
In this letter, we propose a new QoS scheduling algorithm, referred to as the explicit traffic aware scheduling algorithm with explicit queue length notification (ETA-EQN), which aims at supporting multimedia services in HCCA of IEEE 802.11e WLAN. The results from intensive simulations with NS-2 verify that ETA-EQN provides much better network performance than the reference scheduler in terms of throughput, delay, and packet loss.
Xianglei XING Sidan DU Hua JIANG
We extend the Nonparametric Discriminant Analysis (NDA) algorithm to a semi-supervised dimensionality reduction technique, called Semi-supervised Nonparametric Discriminant Analysis (SNDA). SNDA preserves the inherent advantages of NDA, that is, relaxing the Gaussian assumption required for the traditional LDA-based methods. SNDA takes advantage of both the discriminating power provided by the NDA method and the locality-preserving power provided by the manifold learning. Specifically, the labeled data points are used to maximize the separability between different classes and both the labeled and unlabeled data points are used to build a graph incorporating neighborhood information of the data set. Experiments on synthetic as well as real datasets demonstrate the effectiveness of the proposed approach.
Koji KAJIWARA Tatsushi YAMASAKI
In this paper, we propose an optimal supervisory control method for discrete event systems (DESs) that have different preferences. In our previous work, we proposed an optimal supervisory control method based on reinforcement learning. In this paper, we extend it and consider a system that consists of several local systems. This system is modeled by a decentralized DES (DDES) that consists of local DESs, and is supervised by a central supervisor. In addition, we consider that the supervisor and each local DES have their own preferences. Each preference is represented by a preference function. We introduce the new value function based on the preference functions. Then, we propose the learning method of the optimal supervisor based on reinforcement learning for the DDESs. The supervisor learns how to assign the control pattern so as to maximize the value function for the DDES. The proposed method shows the general framework of optimal supervisory control for the DDES that consists of several local systems with different preferences. We show the efficiency of the proposed method through a computer simulation.
Qiang SUN Chen SUN Shi JIN Yuan ZHANG Xiqi GAO
In this paper, we investigate the user scheduling algorithms with statistical eigen-mode transmission (SET) for downlink multiuser multiple-input multiple-output (MU-MIMO) system by utilizing the statistical channel state information (SCSI). Given the objective of maximizing the ergodic achievable sum rate per group (EASRPG), our first proposal, the Munkres user assignment algorithm (MUAA), solves the optimal user grouping problem. Different from the conventional user grouping algorithm (e.g. max-min method), MUAA can efficiently solve the user assignment problem and acquire an optimal solution. However, some user groups of the optimal solution called “unfriendly” groups severely degrade the EASRPG by performing the multiuser SET (MU-SET) due to excessive inter-user interference. To overcome this obstacle, the MUAA with sequential iterative separation (MUAA-SIS) is proposed to find the “unfriendly” groups and switch from the MU-SET to the single-user SET. Finally, our numerical results show that MUAA-SIS offers a higher EASRPG.
Huiyun JING Xin HE Qi HAN Xiamu NIU
BRISK (Binary Robust Invariant Scalable Keypoints) works dramatically faster than well-established algorithms (SIFT and SURF) while maintaining matching performance. However BRISK relies on intensity, color information in the image is ignored. In view of the importance of color information in vision applications, we propose CBRISK, a novel method for taking into account color information during keypoint detection and description. Instead of grayscale intensity image, the proposed approach detects keypoints in the photometric invariant color space. On the basis of binary intensity BRISK (original BRISK) descriptor, the proposed approach embeds binary invariant color presentation in the CBRISK descriptors. Experimental results show that CBRISK is more discriminative and robust than BRISK with respect to photometric variation.
Lin SHAN Sonia AISSA Hidekazu MURATA Susumu YOSHIDA Liang ZHAO
The important issue of an adaptive scheduling scheme is to maximize throughput while providing fair services to all users, especially under strict quality of service requirements. To achieve this goal, we consider the problem of multiuser scheduling under a given fairness constraint. A novel Adaptive Fairness and Throughput Control (AFTC) approach is proposed to maximize the network throughput while attaining a given min-max fairness index. Simulation results reveal that comparing to straightforward methods, the proposed AFTC approach can achieve the desired fairness while maximizing the throughput with short convergence time, and is stable in dynamic scenarios. The trade-off between fairness and throughput can be accurately controlled by adjusting the scheduler's parameters.
Youchan JEON Haesoo KIM Jinwoo PARK
IEEE 802.11ac uses DL MU-MIMO and channelization for very high throughput. We propose a new algorithm for adaptive channelization in DL MU-MIMO WLANs. The proposed scheme can enhance the throughput efficiency for DL MU-MIMO transmission by adaptively applying the channelization scheme according to data size and MCS. Moreover, our proposal can reduce the overhead due to the exchange of control frames for DL MU-MIMO transmission in WLANs. A performance evaluation shows that the proposed scheme is superior to IEEE 802.11ac WLAN.
Retdian NICODIMUS Shigetaka TAKAGI
A design methodology for implementation of low-noise switched-capacitor low-pass filter (SC LPF) with small capacitance spread is proposed. The proposed method is focused on the reduction of operational amplifier noise transfer gain at low frequencies and the reduction of total capacitance. A new SC LPF topology is proposed in order to adapt the correlated double sampling and charge scaling technique at the same time. Design examples show that proposed filter reduces the total capacitance by 65% or more compared to the conventional one without having significant increase in noise transfer gain.
Toshihiro KONISHI Keisuke OKUNO Shintaro IZUMI Masahiko YOSHIMOTO Hiroshi KAWAGUCHI
This paper presents a second-order ΔΣ analog-to-digital converter (ADC) operating in a time domain. In the proposed ADC architecture, a voltage-controlled delay unit (VCDU) converts an input analog voltage to a delay time. Then, the clocks outputs from a gated ring oscillator (GRO) are counted during the delay time. No switched capacitor or opamp is used. Therefore, the proposed ADC can be implemented in a small area and with low power. For that reason, it has process scalability: it can keep pace with Moore's law. A time error is propagated to the second GRO by a multi-stage noise-shaping (MASH) topology, which provides second-order noise-shaping. In a standard 40-nm CMOS process, a SNDR of 45 dB is achievable at input bandwidth of 16 kHz and a sampling rate of 8 MHz, where the power is 408.5 µW. Its area is 608 µm2.
Masayoshi SHIMAMURA Takeshi IKENAGA Masato TSURU
The explosive growth of the usage along with a greater diversification of communication technologies and applications imposes the Internet to manage further scalability and diversity, requiring more adaptive and flexible sharing schemes of network resources. Especially when a number of large-scale distributed applications concurrently share the resource, efficacy of comprehensive usage of network, computation, and storage resources is needed from the viewpoint of information processing performance. Therefore, a reconsideration of the coordination and partitioning of functions between networks (providers) and applications (users) has become a recent research topic. In this paper, we first address the need and discuss the feasibility of adaptive network services by introducing special processing nodes inside the network. Then, a design and an implementation of an advanced relay node platform are presented, by which we can easily prototype and test a variety of advanced in-network processing on Linux and off-the-shelf PCs. A key feature of the proposed platform is that integration between kernel and userland spaces enables to easily and quickly develop various advanced relay processing. Finally, on the top of the advanced relay node platform, we implement and test an adaptive packet compression scheme that we previously proposed. The experimental results show the feasibility of both the developed platform and the proposed adaptive packet compression.
In this letter, we prove that for fading multiuser orthogonal frequency division multiplexing networks, a simple fixed rate scheduling scheme with only 1 bit channel state information feedback is capable of achieving the optimal performance in the wideband limit. This result indicates that the complexities of both the feedback and channel coding schemes can be reduced with nearly no system performance penalty in wideband wireless communication environments.
Ramesh K. POKHAREL Prapto NUGROHO Awinash ANAND Abhishek TOMAR Haruichi KANAYA Keiji YOSHIDA
High phase noise is a common problem in ring oscillators. Continuous conduction of the transistor in an analog tuning method degrades the phase noise of ring oscillators. In this paper, a digital control tuning which completely switches the transistors on and off, and a 1/f noise reduction technique are employed to reduce the phase noise. A 14-bit control signal is employed to obtain a small frequency step and a wide tuning range. Furthermore, multiphase ring oscillator with a sub-feedback loop topology is used to obtain a stable quadrature outputs with even number of stages and to increase the output frequency. The measured DCO has a frequency tuning range from 554 MHz to 2.405 GHz. The power dissipation is 112 mW from 1.8 V power supply. The phase noise at 4 MHz offset and 2.4 GHz center frequency is -134.82 dBc/Hz. The FoM is -169.9 dBc/Hz which is a 6.3 dB improvement over the previous oscillator design.
This paper presents analysis and design of passive RC polyphase filters (RCPFs) in tutorial style. Single-phase model of a single-stage RCPF is derived, and then, multi-stage RCPFs are analyzed and obtained some restrictions for realizable poles and zeros locations of RCPFs. Exact design methods of RCPFs with equal ripple type, and Butterworth type responses are explained for transfer function design and element value design along with some design examples.
Guangchun LUO Hao CHEN Caihui QU Yuhai LIU Ke QIN
Tree partitioning arises in many parallel and distributed computing applications and storage systems. Some operator scheduling problems need to partition a tree into a number of vertex-disjoint subtrees such that some constraints are satisfied and some criteria are optimized. Given a tree T with each vertex or node assigned a nonnegative integer weight, two nonnegative integers l and u (l < u), and a positive integer p, we consider the following tree partitioning problems: partitioning T into minimum number of subtrees or p subtrees, with the condition that the sum of node weights in each subtree is at most u and at least l. To solve the two problems, we provide a fast polynomial-time algorithm, including a pre-processing method and another bottom-up scheme with dynamic programming. With experimental studies, we show that our algorithm outperforms another prior algorithm presented by Ito et al. greatly.
Takashi KUROKI Toshihiko SHIBAZAKI Teruhiro KINOSHITA
Current distributions induced on a circular disk of conductor are analyzed rigorously for an electric dipole incidence, when the source dipole is polarized parallel to the disk and located at an arbitrary position, and they are evaluated numerically. As the height of the dipole increases, the current distribution of the dipole approaches that of the plane wave incidence. Using a multiple precision arithmetic, numerical data for the current distribution are obtained for larger radii of a disk than the former approach.
Ken KIHARA Marina SEKI Sakuichi OHTSUKA
We investigate the visibility of scrolling text presented nearby a dynamically moving image. In two experiments, we evaluate the subjective speed and readability of scrolled fake addresses presented immediately above a moving grating pattern that covers a large part of the visual field. The drift speed and direction of the grating were controlled to reveal the visibility of the text. The results show that the scrolling addresses exhibited slower subjective speed and better readability when the grating moved in the same direction as the scrolling addresses. On the contrary, faster subjective speed and worse readability of the scrolling addresses were raised by the grating moving in the opposite direction. The strength of these effects was dependent on the speed difference between the scrolling addresses and the grating. These results suggest that the visibility of the scrolling text, assessed in terms of subjective speed and readability, strongly depends on nearby moving images.
Due to universal frequency reuse, cell edge users in HSDPA suffer from serious inter-cell interference (ICI). In this letter we present a coordinated scheme for HSDPA which can mitigate ICI by interference avoidance in spatial domain. A system level simulation shows that our scheme can effectively improve the performance of the cell edge users.
The broadcast scheduling problem (BSP) in wireless ad-hoc networks is a well-known NP-complete combinatorial optimization problem. The BSP aims at finding a transmission schedule whose time slots are collision free in a wireless ad-hoc network with time-division multiple access (TDMA). The transmission schedule is optimized for minimizing the frame length of the node transmissions and maximizing the utilization of the shared channel. Recently, many metaheuristics can optimally solve smaller problem instances of the BSP. However, for complex problem instances, the computation of metaheuristics can be quite time and memory consuming. In this work, we propose a greedy genetic algorithm for solving the BSP with a large number of nodes. We present three heuristic genetic operators, including a greedy crossover and two greedy mutation operators, to optimize both objectives of the BSP. These heuristic genetic operators can generate good solutions. Our experiments use both benchmark data sets and randomly generated problem instances. The experimental results show that our genetic algorithm is effective in solving the BSP problem instances of large-scale networks with 2,500 nodes.