基于能量最小的统计三级截值模型

从能量函数出发，用统计学方法建立了统计三级截值模型，克服了用光学技术实现Ｈｏｐｆｉｅｌｄ模型时难以表示其高动态范围互连权重的弱点。与光学中通常采用的模型相比，统计三级截值模型提高了网络的存储容量和寻址能力。     

高容量的单极二值互连神经网络联想存储模型

由于光学固有的数值精度低,难以表示负值等弱点,用光学方法实现神经网络时存在着许多困难。针对光学的弱点,本文提出并建立了具有单极二值互连的适应截值模型,这一模型避开了光学实现时难以表示负值和互连精度差等弱点,计算机模拟及光学实现结果表明,这种单极互连神经网络模型同其他的单极模型相比具有高的存储容量及较强的寻址能力。     

分区适应截值模型及其在交通标志识别中的应用

针对Ｈｏｐｆｉｅｌｄ网络模型在存储模式不满足０和１状态的均匀分布及数目对等的条件下存储容量及寻址能力下降的缺点，提出并用光束方向编码光学实现了三值（１，０，－１）互连的分区适应截值模型，并把这一模型应用到交通标志的识别中，结果表明该模型及光学系统有很好的稳定性。     

一种实现MSD加法的光学方法

三值光学计算机系统结合光强与光的偏振方向表示三值信息,其核心器件——三值逻辑光学处理器是按照降值设计理论完成的,该处理器能完成所有19683种二元三值逻辑运算.本文旨在提出一种实现加法运算的新方法——用三值逻辑光学处理器实现加法.为了解决加法的串行进位延时问题,使用改良符号数表示进行数据编码,从而实现全并行无进位加法.用三值光学计算机与改良符号数表示相结合的方法实现加法既能够充分发挥三值光学计算机位数巨大的优势以及三值逻辑光学处理器能完成所有二元三值逻辑运算的特性,同时又发挥了改良符号数加法的无进位特点.经实验证明该方法具有可行性和正确性,是实现光学加法器的一种新思路.     

基于多区域椭圆参数化描述的形状扩散光学层析成像方法研究

获取生物组织体的光学结构是扩散光学成像模式研究(如扩散光学层析技术和荧光分子成像技术)的核心内容.当假设组织器官中具有均匀光学参数时,光学结构的获取方法可归为基于形状的扩散光学层析成像(DOT)重建技术,该技术致力于同时重建组织器官的边界和内部光学参数.实现了二维的边界元正向模型,并与有限元正向模型的结果进行比较.反演问题中使用了截断奇异值(truncated-SVD)优化算法,并在40 dB噪声水平下用内部具有3个异质体的目标模型进行验证.数值结果表明,提出的图像重建算法具有良好的收敛性能,并能从噪声数据中准确恢复相应区域的形状参数和光学参数.     

一种实现MSD加法的光学方法

三值光学计算机系统结合光强与光的偏振方向表示三值信息,其核心器件——三值逻辑光学处理器是按照降值设计理论完成的,该处理器能完成所有19683种二元三值逻辑运算.本文旨在提出一种实现加法运算的新方法——用三值逻辑光学处理器实现加法.为了解决加法的串行进位延时问题,使用改良符号数表示进行数据编码,从而实现全并行无进位加法.用三值光学计算机与改良符号数表示相结合的方法实现加法既能够充分发挥三值光学计算机位数巨大的优势以及三值逻辑光学处理器能完成所有二元三值逻辑运算的特性,同时又发挥了改良符号数加法的无进位特点.经实验证明该方法具有可行性和正确性,是实现光学加法器的一种新思路.     

磁性纳米线反磁化过程的截椭球链模型

依据实际的纳米线形貌，提出了描述磁性纳米线反磁化过程的截椭球链模型.基于此模型，推导了纳米线反磁化过程中不可逆磁化的临界场和矫顽力在一致转动和对称扇形模式下的表达式.同时，还研究了椭球间接触角、单轴磁晶各向异性、外场与链轴方向夹角、截椭球个数和椭球形状因子对临界场和矫顽力的影响.利用该模型解释了镍纳米线的反磁化过程可能是对称扇形的模式. 关键词： 截椭球链模型 反磁化 一致转动 对称扇形     

生物组织光学参数与热物性参数的相关性研究

基于组织光学和人工神经网络理论,提出了一种全新的确定生物体热物性参数的方法.利用积分球系统测量介质组织光学特性参数,以该特性参数作为人工神经网络输入变量,以热物性参数为目标变量,利用神经网络的自学习系统,最终建立光学特性参数与热物性参数间的非线性映射一黑箱模型.为验证模型,测量了多种热物性参数已知物质的光学特性参数,然后由黑箱模型得到热物性参数计算值,计算值和已知真实值的相对误差均在5%以内.该方法可用于测量生物体的各种热物性参数.     

基于改进Phong模型的天基光学观测GEO目标姿态估计方法

针对于缺少基于天基光学观测的空间目标姿态估计的问题,提出将无味卡尔曼滤波应用于基于天基光学观测的地球同步轨道目标姿态估计中.以光学横截面积代替星等作为观测量,提高观测方程的线性化水平,实现观测方程的简化;同时建立基于改进Phong模型的光学横截面积观测模型,在提高观测方程描述能力的同时,以角度作为模型的输入,极大地简化了计算,增强了实时性.最后,以贴敷聚酰亚胺薄膜及朗伯体材质的长方体为例,对比分析了姿态估计情况,验证了所提方法的有效性;并进一步对正六棱柱、正八棱柱进行了姿态估计.     

在轨空间目标光学特性宏观表征模型的反演重构

为了在天基远距离条件下反演三轴稳定空间目标表面的光学特性参数,提出了基于可见光时序光度信号分析的光学特性宏观表征模型的反演重构方法.首先,综合考虑空间目标的结构特性、表面材料特性、帆板的对日指向运动特性、光照观测几何以及光学系统特性,完善了面向在轨观测的空间目标可见光时序光度建模方法;其次,将光度模型等效为双面模型,并利用双向反射分布函数(BRDF)的多级融合模型表征复杂材料表面的光学反射特性,将BRDF对应的面积反射率乘积作为待反演参数;最后,以时序光度信号的测量值与模型值之间的差异最小为优化目标,建立线性优化方法,实现模型参数的反演.仿真实验表明,提出的模型在轨重构方法对于近轨观测条件下的本体、帆板信号的重构精度达到97%以上,验证了方法的正确性.     

Adaptive Clipping for Neural Associative Memories

ＡｄａｐｔｉｖｅＣｌｉｐｐｉｎｇｆｏｒＮｅｕｒａｌＡｓｏｃｉａｔｉｖｅＭｅｍｏｒｉｅｓＣＨＡＮＧＳｈｅｎｇｊｉａｎｇＳＨＥＮＪｉｎｙｕａｎＳＯＮＧＺｈｕａｎｇＺＨＡＮＧＹａｎｘｉｎ（ＩｎｓｔｉｔｕｔｅｏｆＭｏｄｅｒｎＯｐｔｉｃｓ，ＮａｎｋａｉＵｎｉ...     

大气激光通信非对称限幅光正交频分复用技术

将一种新的正交频分复用调制技术运用到大气激光通信系统中,分析了采用这种非对称限幅光正交频分复用调制技术的大气激光通信系统在大气湍流信道下的性能.在此基础上,把非对称限幅光正交频分复用调制方案与传统的开关键控及直流偏置光正交频分复用调制方案进行了比较,并在大气湍流信道中进行了仿真.仿真结果表明:非对称限幅光正交频分复用调...     

All-optical method of developing half and full subtractors by the use of phase encoding principle

In this paper the authors proposed all-optical half subtractor and full subtractor using phase encoding technique. The processing part of the proposed model is consists of numbers of optical coherent mixers and mirrors. Optical switches are used only in the encoding of inputs, but no such switches are used in the processing part. So real time operation based on this proposed encoding may be achieved. The output of the system always carries same intensity and therefore it is suitable for using in a combinational and sequential logic system.     

Optical binary logic gate-based modified signed-digit arithmetic

A new design approach for a three-step modified signed-digit (MSD) adder is presented that can be optically implemented using binary logic gates. The proposed scheme depends on encoding each MSD digits into a pair of binary digits using a two-state and multi-position encoding scheme. The proposed design algorithm depends on constructing the addition truth table of binary-coded MSD numbers and then using Karnaugh map to achieve output minimization. The optical binary logic gates are obtained by simply programming the decoding masks of a shadow-casting-based optical logic gate system. The proposed scheme results in a simple, compact, and efficient optical binary gate-based parallel addition system.     

Alternating approach of implementing frequency encoded all-optical logic gates and flip-flop using semiconductor optical amplifier

In conduction of parallel logic, arithmetic and algebraic operations, optics has already proved its successful role. Since last few decades a number of established methods on optical data processing were proposed and to implement such processors different data encoding/decoding techniques have also been reported. Currently frequency encoding technique is found be a promising as well as a faithful mechanism for the conversion of all-optical processing as the frequency of light remains unaltered after refection, refraction, absorption, etc. during the transmission of light. There are already proposed some frequency encoded optical logic gates. In this communication the authors propose a new and different concept of frequency encoded optical logic gates and optical flip-flop using the non-linear function of semiconductor optical amplifier.     

基于正交化算法的三值互连神经网络模型

提出了以信息损失最少为原则的三值(±1)互连权重编码方法，这种编码方法比以前的三值权重编码方法显著地提高了神经网络的性能。由于互连权重只有三值，恰恰弥补了光互连精度不高的缺点，易于光学实现。     

Multifactor authentication reinforces optical security

A new optical method to obtain multifactor image encoding and authentication is proposed. The encoded complex-amplitude image function fulfills the general requirements of invisible content, extreme difficulty in copying or counterfeiting, and real-time automatic verification. This optical technique is attractive for high-security purposes that require multifactor reliable authentication. A demonstration using a combination of biometric images and key codes is provided. Retina images, which are very effective for authentication, are used as biometric signals. To the best of our knowledge, this is the first report on combined multiple signal encoding and simultaneous AND authentications for optical security reinforcement.     

An all optical method of implementing a wavelength encoded simultaneous binary full adder-full subtractor unit exploiting nonlinear polarization rotation in semiconductor optical amplifier

Full adders and full subtractors are the basic circuit elements of any digital data processor in electronics as well as in the all optical domain. Again the wavelength dependent encoding/decoding techniques have established itself as a very promising and efficient tool having some inherent and unique advantages relative to the other well known intensity or polarization or phase dependent optical data encoding mechanisms. In this communication, the authors therefore propose a new scheme of implementing a wavelength encoded complete binary full adder-full subtractor unit in the all optical domain using the wavelength conversion by the nonlinear polarization rotation in a single semiconductor optical amplifier. The interacting signals are counter propagating in the semiconductor optical amplifier and hence can be set at the same wavelength. To realize the binary logic wavelength dependent encoding/decoding mechanism is exploited in our proposed scheme of full adder-full subtractor unit. Also the optical add/drop multiplexing employing the special filtering property of the semiconductor optical amplifier is utilized for the designing of the all optical system.