基于MEMS谐振器硬件储备池计算的类脑信号处理方法

近年来兴起的人工神经网络由于具有较强的自学习适应性和并行信息处理能力,从而在信号处理领域显示出巨大潜力。储备池计算是一种由递归神经网络衍生而来的类脑神经形态计算范式,对随时间变化的连续信号具有非常好的分类和时序预测能力。本论文提出利用MEMS（Micro-Electro-Mechanical System）梁谐振器的非线性响应特征,设计并搭建了两种储备池计算的拓扑架构。此外,面向雷达信号处理中信号预测、图像识别、雷达信号特征分类和提取等应用需求,针对性地选择了NARMA（Nonlinear Auto Regressive Moving Average Equation of Order）预测任务、MNIST（Mixed National Institute of Standards and Technology）-手写数字图像识别、LFM（Linear frequency modulated）脉冲波形识别与特征提取等测试任务对论文所提两种不同储备池计算架构进行试验验证。同时,实验结果也充分展示了基于非线性MEMS谐振器的储备池计算硬件系统在雷达信号预测、分类与特征提取等应用领域中的应用潜力。为复杂电磁环境下,雷达信号处理提供新的有力工具,也为MEMS传感技术与雷达信号处理技术的交叉融合进行积极探索。      

光电储备池计算系统研究进展

储备池计算是一种高效处理时间信号的仿生学研究方法。储备池计算的框架是由非线性周期性动力系统加上输入层和输出层构成。储备池计算有很多种可能的实现方法。主要介绍的是基于单个非线性节点加延迟反馈线体系结构的光电储备池。介绍了储备池工作原理和光电储备池模型,以及模拟输入端和模拟输出端的最新研究进展,论述了其未来的研究趋势。     

基于回声状态网络的时间序列预测方法研究

针对回声状态网络(Echo State Networks,ESNs)输入序列延迟时间(和嵌入维数D的选择以及储备池的适应性问题,利用自相关性分析法从被预测样本序列构建ESNs网络输入,并通过移动通信话务量的预测问题,采用实验分析的方法讨论了储备池参数选择对于时间序列预测性能的影响.与采用ARMA和BP神经网络的预测方法相比,新方法在保证预测精度和效率的情况下,具有更好的泛化能力.     

基于多小波双变换的非线性卫星信道盲均衡算法

针对非线性卫星信道Volterra盲均衡系统收敛缓慢、计算复杂高等不足,提出了基于多小波双变换的非线性卫星信道盲均衡算法.该算法用Wiener均衡器代替Volterra均衡器,减小了均衡器结构的复杂性;用平衡正交多小波对Wiener均衡器的输入信号进行变换,降低了输入信号的自相关性;在Wiener均衡器输出端增加一级判决反馈滤波器,同时对其输入信号作平衡多小波变换,又降低了判决反馈滤波器输出信号的自相关性.仿真结果验证了该算法的有效性.     

单电感多路输出DC-DC转换器的建模与设计

建立了单电感多路输出(SIMO) DC-DC转换器的数学模型,通过Matlab验证了SIMO模型的可行性.为解决SIMO控制电路复杂以及静态工作电流大的问题,提出在电流域设计SIMO控制器的方法.该方法将SIMO多路输出电压反馈信号转换为电流信号,与基准电流信号和电感电流信号一并输入多环路PWM控制器,PWM控制器在电流域完成需要的运算及环路频率补偿.仿真结果表明,该算法可行,电路设计方法简洁,在0.13 μm CMOS工艺下,三通道SIMO的面积为0.76 mm2,静态电流为110 μA,驱动能力为200 mA.     

具有人脑功能的半导体

<正> 日本最近研制成功与人脑神经细胞有同样结构和功能的新型半导体。该新型半导体的构造和功能与旧型单导体完全不同,它的一个元件有数个信号的输入电极,能处理多个输入信号,与人脑细胞能接受数个信号一样。用它制作与过去同样的逻辑电路,半导体管的数量可减少到原来的十分之一以下。它还能根据需要迅速转变电路的机能和效力。     

选择用于平台罗经故障检测的神经网络结构的方法

在基于神经网络的平台罗经故障检测中,为了提高故障检测的灵敏度,要求神经网络拟合平台罗经动态系统时均方预测误差的均值及其标准差均小。影响神经网络对系统拟合精度的因素既有隐层节点数也有输入延迟数。本文以均方预测误差的均值和标准差共同作为评价神经网络逼近平台罗经动态系统性能的指标,并借用系统化交叉证实法的结构,且增加一个外循环用以同时选择输入延迟数,构建选择用于平台罗经故障检测的神经网络结构的方法。     

多采样率控制器设计在光电跟踪系统中的应用

采用保持器、采样器和提升技术对输入、反馈和输出多采样率多环路系统进行分析,并将多采样率的周期离散时变系统化为时不变系统,且在输入和反馈端依据信号特性引入高频低通滤波器进行信号频谱延拓周期的扩展.在三环结构光电跟踪控制系统中的应用表明:离散控制器有和连续控制器一致的阶跃响应,在引入高频低通滤波器扩展频域延拓周期后,由低频位置输入引入的系统速度震荡基本上完全被抑制.     

一种新型动力学神经网络的理论算法研究

静态神经网络缺乏动力学行为,动态神经网络由微分方程描述,神经元具有反馈环,更适合描述动态系统.以前的研究中所提出的一些改进的递归网络中引入的都是输入层到关联层的反馈或是输出层到输入层的反馈,但是这些反馈系数是常数,是不可调的,限制了网络反映动态性能的能力.为此,提出了一种由带有积分器和可调反馈系数的神经元构成的动力学神经网络,并利用梯度下降法得到了网络的学习算法.     

优化回声状态网络混沌跳频码预测方法研究

以混沌跳频码预测为背景,针对现有预测方法中存在的缺乏记忆能力导致识别准确率不高以及运算量大等问题,论文提出了基于优化回声状态网络的混沌跳频码预测方法。该方法在继承回声状态网络优良性能的同时,利用改进遗传算法优化网络储备池参数,较好地解决了参数选择问题,使其具有更强的针对性和更好的预测效果。论文以logistic-kent映射、Lorenz系统和Mackey-Glass系统跳频码为样本数据,通过改进遗传算法确定最优储备池参数并进行仿真实验,将仿真结果与其他文献结果作了比较,证明了该预测方法的优越性。      

An adaptive phase alignment algorithm for cartesian feedback loops [Applications corner]

An adaptive algorithm to correct phase misalignments in Cartesian feedback linearization loops for power amplifiers has been presented. It yields an error smaller than 0.035 rad between forward and feedback loop signals once convergence is reached. Because this algorithm enables a feedback system to process forward and feedback samples belonging to almost the same algorithm iteration, it is suitable to improve the performance not only of power amplifiers but also any other digital feedback system for communications systems and circuits such as all digital phase locked loops. Synchronizing forward and feedback paths of Cartesian feedback loops takes a small period of time after the system starts up. The phase alignment algorithm needs to converge before the feedback Cartesian loop can start its ideal behavior. However, once the steady state is reached, both paths can be considered synchronized, and the Cartesian feedback loop will only depend on the loop parameters (open-loop gain, loop bandwidth, etc.). It means that the linearization process will also depend only on these parameters since the misalignment effect disappears. Therefore, this algorithm relieves the power amplifier linearizer circuit design of any task required for solving phase misalignment effects inherent to Cartesian feedback systems. Furthermore, when a feedback Cartesian loop has to be designed, the designer can consider that forward and feedback paths are synchronized, since the phase alignment algorithm will do this task. This will reduce the simulation complexity. Then, all efforts are applied to determining the suitable loop parameters that will make the linearization process more efficient.     

Effects of Optical Buffering on the Performance of Manhattan Street Networks

Multihop networks are strong candidates for the implementation of high-speed networks, ranging from back-plane networks for cluster computing to metropolitan-area networks. Besides using deflection routing for contention resolution, optical buffering may be used to enhance the performance of optical multihop networks and to reduce or even eliminate the need for optical-electrical conversions. The enhancements obtained by augmenting traditional MSN (Manhattan Street Networks) with optical buffering are evaluated. The MSN considered here is based on a novel 6 × 6 optical switching node with up to two fiber delay loops. Impact of the parameters associated with optical buttering, such as the number of delay loops and the loop length, are discussed in detail.     

基于高双折射PCF的双Sagnae环高分辨率FBG解调系统

提出了一种基于高双折射光子晶体光纤（HB-PCF）的双Sagnac环的级联结构。理论计算表明,由光环形器构成的该级联结构,如双环中HB-PCF长度相等,与单环相比,可使干涉消光比增加1倍;实验中,双环结构输出干涉消光比达到43dB。基于边缘滤波器原理,实现了光纤Bragg光栅（FBG）高分辨率解调,静、动态应变分辨率分...     

Efficient automatic parallelization of a single GPU program for a multiple GPU system

Single GPU scaling is unable to keep pace with the soaring demand for high throughput computing. As such executing an application on multiple GPUs connected through an off-chip interconnect will become an attractive option to explore. However, much of the current code is written for a single GPU system. Porting such a code for execution on multiple GPUs is difficulty task. In particular, it requires programmer effort to determine how data is partitioned across multiple GPU cards and then launch the appropriate thread blocks that mostly accesses the data that is local to that card. Otherwise, cross-card data movement is an expensive operation. In this work we explore hardware support to efficiently parallelize a single GPU code for execution on multiple GPUs. In particular, our approach focuses on minimizing the number of remote memory accesses across the off-chip network without burdening the programmer to perform data partitioning and workload assignment. We propose a data-location aware thread block scheduler to schedule the thread blocks on the GPU that has most of its input data. The scheduler exploits well known observation that GPU workloads tend to launch a kernel multiple times iteratively to process large volumes of data. The memory accesses of the thread block across different iterations of a kernel launch exhibit correlated behavior. Our data location aware scheduler exploits this predictability to track memory access affinity of each thread block to a specific GPU card and stores this information to make scheduling decisions for future iterations. To further reduce the number of remote accesses we propose a hybrid mechanism that enables migrating or copying the pages between the memory of multiple GPUs based on their access behavior. Hence, most of the memory accesses are to the local GPU memory. Over an architecture consisting of two GPUs, our proposed schemes are able to improve the performance by 1.55× when compared to single GPU execution across widely used Rodinia [17], Parboil [18], and Graph [23] benchmarks.     

Quantization Methods for Equal Gain Transmission With Finite Rate Feedback

We consider the design and analysis of quantizers for equal gain transmission (EGT) systems with finite rate feedback-based communication in flat-fading multiple input single output (MISO) systems. EGT is a beamforming technique that maximizes the MISO channel capacity when there is an equal power-per-antenna constraint at the transmitter, and requires the feedback of t-1 phase angles, when there are t antennas at the transmitter. In this paper, we contrast two popular approaches for quantizing the phase angles: vector quantization (VQ) and scalar quantization (SQ). On the VQ side, using the capacity loss with respect to EGT with perfect channel information at transmitter as performance metric, we develop a criterion for designing the beamforming codebook for quantized EGT (Q-EGT). We also propose an iterative algorithm based on the well-known generalized Lloyd algorithm, for computing the beamforming vector codebook. On the analytical side, we study the performance of Q-EGT and derive closed-form expressions for the performance in terms of capacity loss and outage probability in the case of i.i.d. Rayleigh flat-fading channels. On the SQ side, assuming uniform scalar quantization and i.i.d. Rayleigh flat-fading channels, we derive the high-resolution performance of quantized EGT and contrast the performance with that of VQ. We find that although both VQ and SQ achieve the same rate of convergence (to the capacity with perfect feedback) as the number of feedback bits B increases, there exists a fixed gap between the two     

On a joint temporal-spatial multi-channel assignment and routing scheme in resource-constrained wireless mesh networks

Use of multiple orthogonal channels can significantly improve network throughput of multi-hop wireless mesh networks (WMNs). In these WMNs where multiple channels are available, channel assignment is done either in a centralized manner, which unfortunately shows a poor scalability with respect to the increase of network size, or in a distributed manner, where at least one channel has to be dedicated for exchanging necessary control messages or time synchronization has to be utilized for managing the duration of data packet transmission, causing excessive system overhead and waste of bandwidth resource. In this paper, we first formulate multi-channel assignment as a NP-hard optimization problem. Then a distributed, heuristic temporal-spatial multi-channel assignment and routing scheme is proposed, assuming every wireless node in the network is equipped with a single-radio interface. Here the gateway node is set to use all the channels sequentially in a round-robin fashion. This temporal scheme ensures all the nodes that need to directly communicate with the gateway node shall have a fair access to it. For those non-gateway nodes, a spatial scheme where channels are assigned based on their neighbors’ channel usage is adopted to exploit parallel communications and avoid channel interference among nodes. Furthermore, since the routing factors, including channel usage of neighbor nodes, node hop count, node memory size, and node communication history, are all considered along with the channel assignment, network performance, measured by packet delivery latency, channel usage ratio, and memory usage ratio, tends to be considerably enhanced. The simulation results have confirmed that, compared with a couple of well-known multi-channel assignment schemes, such as LCM [21] and ROMA [15], the proposed scheme shows substantial improvement in network throughput with a very modest collision level. In addition, the proposed scheme is highly scalable as the algorithm complexity is only linearly dependent on the total number of channels that are available in the network and the number of neighbors that a network node directly connects to.     

Design of low-density parity check codes for MIMO systems with doubly-iterative receivers

In this paper, the Multiple Input Multiple Output (MIMO) doubly-iterative receiver which consists of the Probabilistic Data Association detector (PDA) and Low-Density Parity-Check Code (LDPC) decoder is developed. The receiver performs two iterative decoding loops. In the outer loop, the soft information is exchanged between the PDA detector and the LDPC decoder. In the inner loop, it is exchanged between variable node and check node decoders inside the LDPC decoder. On the light of the Extrinsic Information Transfer (EXIT) chart technique, an LDPC code degree profile optimization algorithm is developed for the doubly-iterative receiver. Simulation results show the doubly-receiver with optimized irregular LDPC code can have a better performance than the one with the regular one.     

Multi-Lookup Table FPGA Implementation of an Adaptive Digital Predistorter for Linearizing RF Power Amplifiers With Memory Effects

This paper presents a hardware implementation of a digital predistorter (DPD) for linearizing RF power amplifiers (PAs) for wideband applications. The proposed predistortion linearizer is based on a nonlinear auto-regressive moving average (NARMA) structure, which can be derived from the NARMA PA behavioral model and then mapped into a set of scalable lookup tables (LUTs). The linearizer takes advantage of its recursive nature to relax the LUT count needed to compensate memory effects in PAs. Experimental support is provided by the implementation of the proposed NARMA DPD in a field-programmable gate-array device to linearize a 170-W peak power PA, validating the recursive DPD NARMA structure for W-CDMA signals and flexible transmission bandwidth scenarios. To the best of the authors' knowledge, it is the first time that a recursive structure is experimentally validated for DPD purposes. In addition to the results on PA efficiency and linearity, this paper addresses many practical implementation issues related to the use of FPGA in DPD applications, giving an original insight on actual prototyping scenarios. Finally, this study discusses the possibility of further enhancing the overall efficiency by degrading the PA operation mode, provided that DPD may be unavoidable due to the impact of memory effects.     

对海探测雷达多目标跟踪技术综述

多目标跟踪(MTT)是雷达数据处理领域的难点。相较于一般场景,海上多目标跟踪(MMTT)面临的挑战更大。一方面,复杂的海洋环境和较低的信杂比使得海面小型目标的检测性能受限,检测得到的点迹存在漏检并包含大量虚警,导致多目标跟踪处理的难度大大增加;另一方面,当海面目标以多群形式编队运动,或采用高分辨率雷达对海探测时,目标量测容易呈现跨单元分布的特征,这种情况下,采用常规的多目标跟踪方法效果不理想。目前,国内外关于海上多目标跟踪方面的研究文献还不多,且大都侧重于单一情形。该文从常规多目标跟踪方法、幅度信息辅助的多目标跟踪方法、多目标检测前跟踪方法以及多扩展目标跟踪方法等4个方面对海上多目标跟踪技术进行了梳理,并对海上多目标跟踪的未来发展方向进行了展望。