图像分类中基于分类矢量量化的视觉词袋模型

特征表示是图像识别和分类的基础，视觉词袋是一种图像的特征表示方法。分析现有视觉词典构建方法的不足，提出一种新的视觉词典构建方法。首先利用梯度方差把特征矢量分为光滑类和边缘类，然后分别针对不同类别的特征矢量进行视觉词典的构建，最后根据两类视觉词典生成视觉词袋。图像分类实验表明，提出的新方法能提高分类准确率。     

Observer‐based fuzzy adaptive quantized control for uncertain nonlinear multiagent systems

This paper focuses on consensus quantized control design problem for uncertain nonlinear multiagent systems with unmeasured states. Every follower can be denoted through a system with unmeasurable states, hysteretic quantized input, and unknown nonlinearities. Fuzzy state observer and Fuzzy logic systems are employed to estimate unmeasured states and approximate unknown nonlinear functions, respectively. The hysteretic quantized input can be split into two bounded nonlinear functions to avoid chattering problem. By combining adaptive backstepping and first‐order filter signals, an observer‐based fuzzy adaptive quantized control scheme is designed for each follower. All signals exist in closed‐loop systems are semiglobally uniformly ultimately bounded, and all followers can accomplish a desired consensus results. Finally, a numerical example is employed to elaborate the effectiveness of proposed control strategy.     

基于视觉显著模型的音视频交叉水印算法

提出一种基于Itti视觉显著模型的音视频交叉水印算法.首先利用音频码流的低通幅值特性生成水印,并嵌入到视频码流中;其次解码视频码流的关键帧Ⅰ帧,获取YUV序列上的亮度(Y)分量,并对其进行显著性分析;最后分块对非显著区域进行DCT变换,并通过量化索引调制嵌入水印.实验结果表明,该算法将嵌入水印后的关键帧平均PSRN值提高了1.678 5 dB,缓解了水印不可见性和鲁棒性的矛盾,实现了音视频码流之间的交叉保护.     

融合矢量量化与LBP的图像检索算法

基于内容的图像检索算法一直是图像领域研究的热门课题,因此提出一种新的融合矢量量化与LBP的图像检索算法。首先,将彩色图像转化到HSI颜色空间,进行矢量量化编码,统计图像码字出现的频数,形成颜色直方图,完成颜色特征的提取;然后,再将彩色图像转化成灰度图像,利用局部二进制模式(LBP)算法提取纹理特征;最后,相似度计算采用颜色特征和纹理特征相似度加权平均,并且改变颜色特征和纹理特征的权值,多次实验,得到使查准率最高的权值。实验结果表明,算法能有效地提升图像检索性能。     

Exploring Two‐Dimensional Transport Phenomena in Metal Oxide Heterointerfaces for Next‐Generation,High‐Performance,Thin‐Film Transistor Technologies

In the last decade, metal oxides have emerged as a fascinating class of electronic material, exhibiting a wide range of unique and technologically relevant characteristics. For example, thin‐film transistors formed from amorphous or polycrystalline metal oxide semiconductors offer the promise of low‐cost, large‐area, and flexible electronics, exhibiting performances comparable to or in excess of incumbent silicon‐based technologies. Atomically flat interfaces between otherwise insulating or semiconducting complex oxides, are also found to be highly conducting, displaying 2‐dimensional (2D) charge transport properties, strong correlations, and even superconductivity. Field‐effect devices employing such carefully engineered interfaces are hoped to one day compete with traditional group IV or III–V semiconductors for use in the next‐generation of high‐performance electronics. In this Concept article we provide an overview of the different metal oxide transistor technologies and potential future research directions. In particular, we look at the recent reports of multilayer oxide thin‐film transistors and the possibility of 2D electron transport in these disordered/polycrystalline systems and discuss the potential of the technology for applications in large‐area electronics.     

Optimal control of uncertain quantized linear discrete‐time systems

In this paper, the adaptive optimal regulator design for unknown quantized linear discrete‐time control systems over fixed finite time is introduced. First, to mitigate the quantization error from input and state quantization, dynamic quantizer with time‐varying step‐size is utilized wherein it is shown that the quantization error will decrease overtime thus overcoming the drawback of the traditional uniform quantizer. Next, to relax the knowledge of system dynamics and achieve optimality, the adaptive dynamic programming methodology is adopted under Bellman's principle by using quantized state and input vector. Because of the time‐dependency nature of finite horizon, an adaptive online estimator, which learns a newly defined time‐varying action‐dependent value function, is updated at each time step so that policy and/or value iterations are not needed. Further, an additional error term corresponding to the terminal constraint is defined and minimized along the system trajectory. The proposed design scheme yields a forward‐in‐time and online scheme, which enjoys great practical merits. Lyapunov analysis is used to show the boundedness of the closed‐loop system; whereas when the time horizon is stretched to infinity as in the case of infinite horizon, asymptotic stability of the closed‐loop system is demonstrated. Simulation results on a benchmarking batch reactor system are included to verify the theoretical claims. The net result is the design of the optimal adaptive controller for uncertain quantized linear discrete‐time systems in a forward‐in‐time manner. Copyright © 2014 John Wiley & Sons, Ltd.     

Space Efficient Quantization for Deep Convolutional Neural Networks

Deep convolutional neural networks (DCNNs) have shown outstanding performance in the fields of computer vision, natural language processing, and complex system analysis. With the improvement of performance with deeper layers, DCNNs incur higher computational complexity and larger storage requirement, making it extremely difficult to deploy DCNNs on resource-limited embedded systems (such as mobile devices or Internet of Things devices). Network quantization efficiently reduces storage space required by DCNNs. However, the performance of DCNNs often drops rapidly as the quantization bit reduces. In this article, we propose a space efficient quantization scheme which uses eight or less bits to represent the original 32-bit weights. We adopt singular value decomposition (SVD) method to decrease the parameter size of fully-connected layers for further compression. Additionally, we propose a weight clipping method based on dynamic boundary to improve the performance when using lower precision. Experimental results demonstrate that our approach can achieve up to approximately 14x compression while preserving almost the same accuracy compared with the full-precision models. The proposed weight clipping method can also significantly improve the performance of DCNNs when lower precision is required.

     

Lattice Vector Quantization Applied to Speech and Audio Coding

Lattice vector quantization(LVQ) has been used for real-time speech and audio coding systems.Compared with conventional vector quantization,LVQ has two main advantages:It has a simple and fast encoding process,and it significantly reduces the amount of memory required.Therefore,LVQ is suitable for use in low-complexity speech and audio coding.In this paper,we describe the basic concepts of LVQ and its advantages over conventional vector quantization.We also describe some LVQ techniques that have been used in speech and audio coding standards of international standards developing organizations(SDOs).     

Analysis of switched quantizer based on the quadratic spline functions

In this paper, an approximation of the optimal compressor function using the quadratic spline functions with 2L?=?8 segments is described. Since the quadratic spline with 2L?=?8 segments provides better approximation of the optimal compression function than quadratic spline with 2L?=?4 segments, capitalizing on the benefits of the obtained spline approximation, quantizer designing process is firstly performed for the so assumed number of segments and the Laplacian source of a unit variance. Then, to enhance the usability of the proposed model, the switched quantization technique is applied and a beneficial analysis is derived, providing insight in the robustness of the proposed quantizer performances with respect to the mismatch in designed for and applied to variances. Reached quality has been compared to another model from the literature, and it has been shown that the proposed model outperforms the previous model by almost 1.3?dB.