基于突触可塑性的SNN随钻陀螺仪漂移处理

针对随钻振动引起MEMS陀螺仪的数据漂移问题,文中提出了一种脉冲神经网络算法.首先根据陀螺仪漂移误差的时间特性,利用脉冲网络的脉冲时间编码陀螺仪的信息强度.然后利用Izhikevich神经元模型的突触可塑性,调节激发性突触电导并抑制性突触电导,增强网络的鲁棒性,从而提高陀螺仪信号对噪声的抗干扰能力.在不同振动频率下,分...     

低功耗非对称性可调STDP突触电路设计

基于65 nm CMOS工艺设计了一种可用于脉冲神经网络系统的低功耗、高能效、结构紧凑的突触电路.突触电路采用开关电容电路结构,直接接收来自神经元电路的脉冲信号,根据脉冲时间依赖可塑性(STDP)学习规则调节突触权重,并实现了权重学习窗口的非对称性调节,使突触电路可以适应不同情况.仿真结果表明,突触电路耗能约为0.4 ...     

一种阈值转换忆阻器仿真电路设计

人工神经元和突触是处理神经形态系统中复杂信息的关键单位。忆阻器经常被设计为人工突触,因为它们具有简单的结构,逐渐变换的电导和高密度的集成度。文章设计的这种阈值转换忆阻器(TSM)模型的仿真电路不仅可以尽可能高效的仿真出TSM的基本性能,并通过调节模型相关设置值来计算得出所需要模拟的器件对应参数值,同时解决了现有的仿真模型在仿真过程中仿真时间长、产生仿真文件大、设计难度大、应用的元器件多等技术难题。该模型仅包括电阻、小型运算放大器、滞回比较器、压控开关以及电源,其使用的元器件相对于传统设计电路大幅度减少,模型仿真时间以及仿真文件大小大幅度降低。同时本模型能够较为真实地拟合出TSM相应的电气特征,可以通过调节参数改变TSM的电阻值大小,从而在不同的外界工作条件下,实现记忆或者信息存储的功能。在应用于整合-激发神经元电路(IF)中,也可以实现预期所需要的模拟生物突触间的脉冲信号特性。这些结果表明,良好的人工突触电路可以实现生物突触的基本形态描述,并表现出神经形态计算的巨大潜力。     

IF神经网络电路模型及其分析

为了对IF神经网络模型向实际运用转化,具体对IF模型进行电路模拟,并从电路特性来解释生物神经元的主要特征,该模型满足一般的生物神经元的特性,如:具备非线性空间特性、时间特性(绝对不应期)、EPSP(Excitatory Post-synapitc Potentials)I、PSP(Inhibitory Post-synapitc Potentials),及胞体在阈值下活动期间能够完成EPSP和IPSP之间的非线性相互作用。从而使模型向具体电路转化接近,为人工模拟的理想模型向实际运用提供了方法。     

具有非线性突触联接和适应性的人工神经元电路模型

基于匹配神经细胞突触、胞体和轴突的电神经生理特性,本文提出了一种接近真实神经细胞输入-输出关系和强度-时间曲线的,且具有非线性突触联接和适应性的非线性压控电阻电路模型,并分析了它的一些基本特性。结果表明,该模型比现存的神经网络更接近真实神经计算原理。由这一模型组成的胜者为王(WTAWinner-Takes-All)网络能够比模拟网络更有效地检测出最大输出,并引入了强度竞争和时间竞争两个新概念。     

一种改进FHN神经元滤波模型研究

突触滤波是神经元处理和传递信息的重要过程,有助于生物在复杂环境中获取所需信息.针对当前人工神经元模型中较少考虑到突触滤波机制,本文以FitzHugh-Nagumo(FHN)人工神经元模型为基础构建基于膜电势增量变化的神经元数学模型,在此基础上模拟突触滤波机制,从而提出一种改进FHN神经元滤波模型.而后,对该模型的稳定性...     

一种基于侧抑制的IF模型及其分析

根据视觉神经元中的侧抑制特性对传统IF模型进行简化,简化了活动方程、降低了运算量。在神经元的点火方法上采用不同以往的异步的离散值的点火机制,提高了网络的适应性。仿真结果表明。对图像的处理比较接近人的行为活动,说明对此模型的改进是成功的。     

具有非线性突触联接和适应性的人工神经元电路模型

基于匹配神经细胞突触、胞体和轴突的电神经生理特性，本文提出了一种接近真实神经细胞输入－输出关系和强度－时间曲线的，且具有非线性突触联接和适应性的非线性压控电阻电路模型，并分析了它的一些基本特性。结果表明，该模型比现存的神经网络更接近真实神经计算原理。由这一模型组成的胜者为王（ＷＴＡ－Ｗｉｎｎｅｒ－Ｔａｋｅｓ－Ａｌｌ）网络能够比模拟网络更有效地检测出最大输出，并引入了强度竞争和时间竞争两个新概念。     

具有艾宾浩斯遗忘规则的忆阻联想记忆电路

忆阻器由于具有低功耗、记忆能力和纳米尺寸等特点,是实现人工神经突触的理想器件。为构建简洁、高效、功能全面地联想记忆电路,该文首先提出一种简单的神经元电路和基于压控阈值忆阻器的突触电路。然后根据巴甫洛夫联想记忆模型,设计了相应的联想记忆电路。电路结构简单,仅包含3个神经元电路和突触电路,可有效降低网络复杂度和功耗。尤为重要的是该电路可以模拟全功能的联想记忆行为,不但实现了学习、遗忘、加速学习、减速遗忘以及减速自然遗忘等功能,而且学习速率和自然遗忘速率能够根据学习的次数自动调整,使电路更具仿生性。此外,该电路与艾宾浩斯遗忘曲线相吻合,扩大了电路的适用范围。     

一种神经元模拟电路的分析和应用

本文研究神经元的模拟电路实现，分析一种可编程突触电路、变增益细胞体电路，用它们构成一个８×７单层感知机网络。以３μｍＣＭＯＳ工艺第二类模型参数进行ＰＳＰＩＣＥ模拟，实现了预期的模式识别功能。     

Neuromorphic Synapses for Artificial Dendrites

We describe neuromorphic, variable-weight synapses onartificial dendrites that facilitate experimentation with correlativeadaptation rules. Attention is focused on those aspects of biologicalsynaptic function that could affect a neuromorphic network'scomputational power and adaptive capability. These include sublinearsummation, quantal synaptic noise, and independent adaptationof adjacent synapses. The diffusive nature of artificial dendritesadds considerable flexibility to the design of fast synapsesby allowing conductances to be enabled for short or for variablelengths of time. We present two complementary synapse designs:the shared conductance array and the self-timed synapse. Bothsynapse circuits behave as conductances to mimic biological synapsesand thus enable sublinear summation. The former achieves weightvariation by selecting different conductances from an on-chiparray, and the latter by modulating the length of time a constantconductance remains activated. Both work with our interchip communicationsystem, virtual wires. For the present purpose of comparing variousadaptation mechanisms in software, synaptic weights are storedoff chip. This simplifies the addition of quantal weight noiseand allows connections from different sources to the same dendriticcompartment to have independent weights.     

用于数字电视调谐芯片的高线性CMOS VGA

设计了一种适于DVB-C标准的中频可变增益放大器。该放大器由三部分构成:电流调节型可变增益单元、基于差分对管传输特性的指数控制电压产生电路以及一高线性输出级。采用Chartered0.25μm RFCMOS工艺库下流片。测试结果表明,4～49dB的连续增益范围,100MHz的3dB带宽,50Ω负载下的OIP3为16.8dBm。     

Photoconductivity and photomagnetoelectric effect of SiGaAs in the presence of different surface recombination regimes

We present an analysis of the photoconductivity (PC) and photomagnetoelectric effect (PME) under the conditions of linear and squarelaw recombination at the surface and linear recombination in the bulk. Analytical results for the above conditions and a discussion for different excitation conditions due to strongly absorbed radiation are presented. Application of this model to PC, PME and magnetophotoconductivity (MPC) for SiGaAs leads to satisfactory agreement with experimental results.     

神经元电突触前向传输电导的时变特性

电突触前膜在高于阈值的刺激下,产生一个大小不变的动作电位,突触后细胞可直接被突触前轴突中伴随动作电位产生的电紧张所激活,突触膜相当于整流器,电流只能从突触前神经元流向突触后神经元。电流流过神经元之间结合部时,连结电导将随时间变化。     

An adaptive Generalized S-transform for instantaneous frequency estimation

The Generalized S-transform (GST) applied in Instantaneous Frequency (IF) estimation is addressed in this paper. The analytical model for IF estimation error using the GST is established, which reveals that the estimation bias is determined directly by two variables: the standard deviation of the Gaussian window and the Instantaneous Frequency Rate (IFR) of the test signal. Based on the mathematical model, performance of the GST as an IF estimator is analyzed by three steps. Firstly, the time-varying expression for IF estimation Relative Error (RE) using the GST is derived. Secondly, the relationships between the RE and the signal’s frequency characters are found. Thirdly, the effects of the window parameters on the RE are discussed. Consequently, a novel Adaptive Generalized S-transform (AGST) is developed, in which the window parameters are automatically adjusted in order to process signals with fast changing frequency characters. The proposed method is evaluated through several kinds of test signals to show its efficiency in IF estimation.     

Auditory motivated level-crossing approach to instantaneous frequency estimation

We address the problem of estimating the instantaneous frequency (IF) of a phase signal using its level-crossing (LC) information based on front-end auditory processing motivation. We show that the problem of IF estimation using LC information can be cast in the framework of estimation from irregularly sampled data. The formulation has the generality of estimating different types of IF without the need for a quasistationary assumption. We consider two types of IF-polynomial and bandlimited; we use polynomial interpolating functions for the former, and for the latter, we propose a novel "line plus sum of sines" model. The model parameters are estimated by linear regression. Considering the noisy case, LC data for different levels is analyzed, and methods for combining different estimators from LCs are discussed. Theoretical and extensive simulation results show that the performance of the zero-crossing (ZC) based IF estimator and the level-crossing based IF estimator with smaller level values is better than those obtained with higher level values or their combinations. The new technique reaches the Crame/spl acute/r-Rao bound (CRB) roughly above 4 dB signal-to-noise ratio (SNR), and its performance does not deteriorate rapidly with mismatch in the IF order compared with the other techniques in the literature.     

基于二阶采样和二次混频的全数字正交解调

在获取有采样时间差的数字I／Q信号过程中,二阶采样在采样频率、载频和采样时间差的选择上比一阶采样有更大的自由度。建立了单中频二阶采样正交解调模型,通过获取有采样时间差的数字复指数信号和利用二次混频技术,提出了一种新的采样频率与载频关系式,极大地拓展了二阶采样的采样频率、载频和采样时间差的选择范围,简化了ADC前端雷达系统设计和ADC采样过程。最后通过实验仿真与分析进行了验证。     

BER for optical heterodyne DPSK receivers using delay demodulationand integration detection

Previous work on the bit error rate (BER) performance of differential phase-shift keying (DPSK) including the effects of additive white Gaussian noise (AWGN) and phase noise has concentrated on the delay demodulator with narrow-band intermediate frequency (IF) bandpass filtering (BPF) and sampling detection. No similar analysis has yet been performed for the delay demodulator with wideband IF bandpass filtering and integration detection. Phase noise is an important consideration in coherent optical communication systems and the most widely accepted model is a Brownian motion process. A closed-form BER expression along with detailed Monte Carlo simulation results are presented for the DPSK delay demodulator with wideband IF bandpass filtering and integration detection filtering including phase noise effects using the Brownian motion model. Analytic expressions are also obtained for the moments of the phase noise component of the decision variable. Using these moments, estimates of the phase noise BER floor are produced. It is found that this receiver has noise performance comparable to receivers with narrowband IF bandpass filtering and sampling detectors for signal-to-noise ratio (SNR) and phase noise in the range of practical interest, but with potentially less degradation due to intersymbol interference (ISI)     

Cramer-Rao bounds for wavelet transform-based instantaneous frequency estimates

We use an asymptotic integral approximation of a wavelet transform as a model for the estimation of instantaneous frequency (IF). Our approach allows the calculation of the Cramer-Rao bound for the IF variance at each time directly, without the need for explicit phase parameterization. This is in contrast to other approaches where the Cramer-Rao bounds rely on a preliminary decomposition of the IF with respect to a (usually polynomial) basis. Attention is confined to the Morlet wavelet transform of single-component signals corrupted with additive Gaussian noise. Potential computationally and statistically efficient IF extraction algorithms suggested by the analysis are also discussed.