基于FPGA的高效可伸缩的MobileNet加速器实现

MobileNet网络是一种广泛应用于嵌入式领域的深度神经网络,为了解决其硬件实现效率低的问题,同时达到在不同硬件资源下具有一定可伸缩性,提出了基于FPGA的一款MobileNet网络加速器结构,针对网络的堆叠结构特性设计了三级流水的加速阵列,并实现了在0～4000乘法器开销下都达到70％ 以上的计算效率.最终在XIL...     

基于3D可扩展PE阵列CNN加速器的设计

卷积神经网络具有参数大、运算量大的特点,当将其具体应用在移动端设备时,需要在满足帧率(速度)的前提下,尽量减少功耗与芯片面积.考虑满足现有移动端网络的兼容性、性能和面积等因素,设计一个基于3D可扩展PE阵列的CNN加速器.该加速器兼容3×3卷积、3×3深度可分离卷积、1×1卷积和全连接层,其PE阵列能根据具体应用的网络...     

Optimized scalable cache management for video streaming system

An important research issue in video streaming is how to efficiently utilize the network resources to provide clients instant access to multiple video objects. Caching strategy and transmission scheme are the two essential points inside the video streaming framework. Recent research efforts on them are not sufficient due to their inflexible support for scalable encoded video streams and heterogeneous requests from clients. In this paper, we propose an optimized caching strategy (OCS) and a scalable transmission scheme (STS) for scalable coded video streaming. By exploring the characteristics of video streaming workload and system design objectives, OCS and STS work efficiently to minimize both network bandwidth cost and user access latency. Firstly, we analyze the caching problem for the proxy-assisted video streaming system and derive a maneuverable caching scenario. Secondly, we develop an efficient transmission scheme for scalable coded videos. Thirdly, we formulate a multi-objective optimization model with closed-form expressions to obtain the optimized caching strategy. Finally, with designed algorithms, an excellent compromise between two competing objectives (minimizing the bandwidth cost and the access latency) is achieved. We start our evaluation by studying the optimized caching strategy for a single video object. Then we apply the strategy to multiple video objects and illustrate the tradeoff between the optimization objectives. Our evaluation results show that compared with other caching strategies, the proposed optimized scalable caching strategy can achieve a significant reduction in bandwidth cost with even a small proxy cache size. Meanwhile, the best performance (in terms of bandwidth cost) is obtained together with the proposed scalable batch-patching transmission scheme.

A geometric reasoning based algorithm for point pattern matching

Point pattern matching (PPM) is an important topic in computer vision and pattern recog-nition . It can be widely used in many areas such as image registration, object recognition, motion de-tection, target tracking, autonomous navigation, and pose estimation. This paper discusses the in-complete matching problem of two point sets under Euclidean transformation. According to geometric reasoning, some definitions for matching clique, support point pair, support index set, and support in-dex matrix, etc. are given. Based on the properties and theorems of them, a novel reasoning algo-rithm is presented, which searches for the optimal solution from top to bottom and could find out as many consistent corresponding point pairs as possible. Theoretical analysis and experimental results show that the new algorithm is very effective, and could be, under some conditions, applied to the PPM problem under other kind of transformations.     

Algorithms for pattern matching

This paper describes four algorithms of varying complexity used for pattern matching, and investigates their behaviour. The algorithms are tested using patterns of varying length from several alphabets. It is concluded that although there is no overall ‘best’ algorithm, the more complex algorithms are worth considering as they are generally more efficient in terms of number of comparisons made and execution time.     

基于WM算法的多模式匹配改进算法WMN

入侵检测系统中有部分时间用来进行模式匹配,因此提高模式匹配算法的处理速度具有重要的意义。从两个方面改进了目前公认效率较高的多模式匹配算法WM,设计出WMN算法,并对该算法进行了性能测试和分析。实验证明该文提出的WMN算法能够有效提高模式匹配的处理速度。     

Secure pattern matching based on bit parallelism

International Journal of Information Security - In this paper, we consider the problem of secure pattern matching that uses evaluation of non-deterministic string matching automata (NSMA). Our...     

Reconfigurable memory architecture for scalable IP forwarding engines

In this paper, a reconfigurable memory architecture and lookup technique for IP packet forwarding engines is presented. This is achieved by partitioning the forwarding table into smaller partial lookup tables for each output port and allowing a forwarding engine to process them in parallel. This effectively reduces the complexity of finding the ‘longest prefix match’ problem to the ‘first prefix match’ problem. Our method is a flexible technique that significantly elevates the scalability of the next generation network processors and other packet processing devices. Such scalability facilitates migration to IPv6 and benefits network equipments especially in terms of growing routing table size, traffic, frequency of route updates and bandwidth requirement.     

基于GPU的AC模式匹配改进算法

字符串匹配算法的应用非常广泛,在信息检索、信息安全等领域都起着关键的作用。近年来,由于GPU通用计算的高速发展,且GPU具有很强的并行计算能力和很高的存储器访问带宽,利用GPU来加速字符串匹配算法吸引了越来越多的关注。提出的改进的AC模式匹配算法,在对前人工作的基础上,进一步消除了output表的存储,将纹理存储器中的查表操作转换为数值比较操作,与改进前算法相比,速度提高了80%以上;进一步的,引入了多个可变参数,提高AC算法的有效数据匹配率,并优化线程块的大小,优化后的算法与采用一种特殊匹配方式的高效的PFAC算法相比,速度提高了9%以上。     

基于EMD方法的时间序列分层相似性匹配算法

时间序列数据挖掘是时态数据挖掘的一个重要方面,针对金融时间序列非稳定、非线性的特点,使用EMD方法进行序列趋势的提取,得到了原始时间序列的长期趋势。在此基础上提出了子序列分层匹配算法,首先进行时间序列趋势的粗匹配,在结果集中进一步进行细节匹配,与传统方法相比,提高了相似性匹配的效率,减少了结果集的冗余。     

一种基于极坐标变换的点模式匹配算法

点模式匹配是目标识别、图像配准与匹配、姿态估计等计算机视觉与模式识别应用方向的基础问题之一。提出了一种新的利用点特征进行匹配的算法,该算法根据点集的分布与点位置信息,构建了点的特征属性图,通过极坐标变换得到对数极坐标的特征图,并利用几何不变矩方法对特征图进行描述。由特征描述向量的比较,获得粗匹配结果,然后通过几何约束迭代的方法获取最终的点集匹配结果。本文贡献如下:一,构建了一种点的极坐标变换特征,并运用不变矩进行描述,使所提特征具有旋转与平移的不变性;二,提出了利用点特征与整体点集几何约束结合的匹配算法,能有效克服出格点与噪声带来的不利影响。最终实验说明了算法的有效性和鲁棒性。     

基于HASH-TCAM算法的高速模式匹配研究*

针对基于硬件的模式匹配算法处理长模式串时吞吐率不高的问题,提出了基于将哈希压缩与TCAM查表相结合的算法——HASH-TCAM算法。通过哈希算法将待匹配的关键字预处理,减少其长度,解决了40 Gbps线速下的长模式串匹配问题,并通过40 Gbps测试仪验证了该算法的可行性。分析表明,该算法在查询的固定关键字长度为72 Byte,模式集数目为5000,哈希压缩后地址的编码宽度为46 bit时,模式识别模块以牺牲冲突的代价实现了51.2 Gbps的吞吐率,可以满足40 Gbps链路中DPI算法的逐包线速的需求     

基于线模式的指纹细节点匹配算法

提出了一种新的基于线模式的指纹图像细节点匹配算法,这种算法先分别找出最佳匹配点对,确定相对旋转角度,较快地确定两个指纹的相关性,之后再搜索漏的匹配点对,去掉错误的匹配点对。算法受指纹增强、后处理的影响较小,细节点相似程度受其邻域细节点影响较小。介绍了算法的具体实现方法,并用实际指纹数据进行了测试。初步测试的结果表明,此算法是指纹图像细节点匹配的一种有效算法。     

基于模式串匹配的Ethernet协议识别算法研究

协议识别是网络安全与对抗领域中的一项关键技术。简要介绍了网络协议识别的重要性,分析了Ethernet类协议的特征,构建了协议识别的系统模型,依据Ethernet类协议的特征,提出了一种新的基于模式串匹配的协议识别算法NEWMATCH,并对该算法的性能进行了分析,并与传统模式串匹配算法（如：BF、KMP、BM等）进行了比较。该算法在Ethernet类协议识别方面具有一定的适用性。     

基于隔离子采样的鲁棒图模型点匹配算法*

为解决概率图模型匹配算法对模板点集中的外点敏感的问题,对隔离子进行了采样,并推导出采样次数和模板点集中的外点比例关系,以保证隔离子中的点能最大概率地为内点,使得推理算法中每个点的信息能得到传递。另外,算法中的互匹配解决了传统图模型匹配算法中多对一问题。实验证明,算法相对于原先的JT算法具有很高的鲁棒性和正确率。     

基于模糊模式识别的车辆定位地图匹配算法

针对传感器给出的车辆定位信息的不确定性,提出了基于模糊模式识别技术获得匹配道路的算法。给出的算法不需要车辆行驶的速度、方向参数,利用车辆定位轨迹与电子地图道路网之间的相似性,结合车辆实际行驶情况,得出定位点到候选道路之间的距离、相邻采样时刻定位点连线与候选道路之间的夹角、候选道路与历史匹配道路连通性的隶属函数,按照最大隶属原则选择匹配道路。通过实验验证,表明该算法具有可靠性,适用面广,能有效地提高车辆定位的精度。     

A pattern based PSO approach for block matching in motion estimation

Block matching motion estimation is a popular method in developing video coding applications. A new algorithm has been proposed for reducing the number of search points using a pattern based particle swarm optimization (PSO) for motion estimation. The conventional particle swarm optimization has been modified to provide accurate solutions in motion estimation problems. This leads to very low computational cost and good estimation accuracy. Due to the center biased nature of the videos, the proposed approach uses an initial pattern to speed up the convergence of the algorithm. Simulation results show that improvements over other fast block matching motion estimation algorithms could be achieved with 31%～63% of search point reduction, without degradation of image quality.