一种基于CMOS的电流模式线性分类器

提出了一种对线性不可分数据集进行分类的电流模式线性分类器.该分类器电路结构简单,仅由梯形激活函数电路和线性加权电路组成,其中线性加权电路采用全平衡差分跨导电路实现,梯形激活函数电路主要由阈值电路组成.为了实现对线性不可分数据集的分类,通过MATLAB软件采用Fisher线性判别法计算得到权重系数,并运用PSPICE对所提出的电路进行仿真分析.结果表明:提出的电路结构简单、准确度高、功耗低,可以广泛地应用于模式识别、神经网络、人工智能等领域.     

PCA+4点算法在手写体数字特征识别中的应用

提出一种改进手写字体特征的提取方法:将传统的PCA特征方法与13点特征方法进行综合,得到一种PCA+4点的特征提取算法,然后通过BP神经网络进行训练识别.实验仿真表明这种改进的方法比PCA特征提取及13点特征提取的识别率高,特别在手写变化大、手写速度快等方面优势更加明显.     

An Incremental Time—Delay Neural Network for Dynamical Recurrent Associative Memory

An incremental time-delay neural network based on synapse growth,which is suitable for dynamic control and learning of autonomous robots,is prooposed to improve the learning and retrieving performance of dynamical recurrent associative memory architecture.The model allows steady and continuous establishment of associative memory for spatio-temporal regularities and time series in discrete sequence of inputs.The inserted hiddewn units can be taken as the Long-term memories that expand the capacity of network and sometimes may fade away under certain condition.Preliminary experiment has shown that this incremental netwrok may be a promising approach to endow autonomous robots with the ability of adapting to new data without destroying the learned patterns.The system also bendfits from its potential chaos character for emergence.     

神经网络在引信智能信息处理中应用的研究

介绍了应用神经网络进行引信智能信息处理的探索性研究。采用BP算法建立了无监督特征提取的图象识别系统,包括三层网络自动联想机与两层网络信息提取器,实现了对输入图象的训练、记忆与分类。研究结果表明该方法技术可行,在一定程度上体现了信息处理的智能性。     

Chemical pattern recognition applied to materials optimal design and industry optimization

The multi-dimensional spaces, spanned by the dimensionless numbers describing the equilibrium states or the macroscopic chemical kinetics of complicated chemical reaction systems, have been used for the data processing of the processes in materials preparation or industrial production. It has been found that there are four common topological types of the relations between the optimization zone and the operation zone in the multi-dimensional spaces. Various computation methods using pattern recognition, artificial neural network and genetic algorithm are used for model-building and optimization strategy searching. An integrated system named KDPAG has been built for materials optimal design or industrial optimization. Application to chemical, petrochemical and metallurgical industries and materials optimal design gives good results.     

基于DNA链置换反应的编码器逻辑运算模型研究

作为自组装DNA计算领域中一门新技术,DNA链置换反应在分子计算领域得到了广泛的应用.基于自组装DNA计算原理,设计了对应不同逻辑门的DNA分子电路.基于DNA链置换反应机理构建了编码器逻辑电路的分子计算模型.当输入DNA分子信号链时,将不同分子浓度比的DNA分子逻辑门电路混合,借助分子间的特异性杂交反应及分子间链置换反应,最终可输出信号链分子.Visual DSD仿真结果表明了本文设计的编码器逻辑计算模型的可行性与准确性.为拓展分子逻辑电路的应用做出有益的探索.     

基于数据融合的多生物特征身份识别

针对传统的单一生物特征身份识别由于传感器的噪音以及特征的破损、匹配等缺陷,往往不能正确识别的情况,提出了一种基于语音和人脸的多生物特征身份识别方法.分别提取语音特征和人脸特征作为识别的依据,并用神经网络在特征层上进行融合识别.实验证明,这种方法可以充分挖掘特征之间的关系,在同等条件下,具有更高的识别率.     

紫杉醇类似物抗癌活性构效关系的神经网络模式识别研究

经对MM3、MM＋、MNDO、PM3几何优化结果进行比较，选用速度最快且精确度较好的MM3分子力学方法计算43个紫杉醇类似物的优势构型，应用MNDO法计算了化合物的电子结构，并用回归分析和BP神经网络模式识别方法寻找其电子结构与抗癌活性的关系．结果表明：（1）紫杉醇类似物及C13侧链的油水分配系数与活性参数问呈抛物线关系，最适油水分配系数Popt=3．14；（2）2-OBz中Bz基团的负电荷密度越大，C1、C3原子的正电荷密度越大对活性越有利；（3）R1、R2、1-OH和2-OBz基团可能是药物与受体作用的重要部位．四参数的定量构效关系显著性较好，神经网络模式识别总识别率为98％，可较精确地预测化合物的抗癌活性．     

利用人工神经网络判别陶瓷原料种类

利用三层 BP人工神经网络 ,对陶瓷原料进行模式识别 .以原料特征参数作为判别指标 ,由训练样本集得到最佳网络参数 ,然后对待测样本的种类进行预测 .研究结果表明 ,人工神经网络用于陶瓷原料的模式识别 ,其结果和实际一致 .该方法有助于配方中陶瓷原料的选择 ,值得推广     

A molecular cryptography model based on structures of DNA self-assembly

With the progress of DNA computing, DNA- based cryptography becomes an emerging interdisciplinary research field. In this paper, we present a novel DNA cryptography that takes advantage of DNA self assembled structure. Making use of the toehold strands recognition and strand displacement, the bit-wise exclusive-or （XOR） operation is carried out to fulfill the information encryption and decryption in the form of a one-time-pad. The security of this system mainly comes from the physical isolation and specificity of DNA molecules. The system is con- structed by using complex DNA self-assembly, in which technique of fluorescent detection is utilized to implement the signal processing. In the proposed DNA cryptography, the XOR operation at each bit is carried out individually, thus the encryption and decryption process could be con- ducted in a massive, parallel way. This work may dem- onstrate that DNA cryptography has the great potential applications in the field of inRwmation security.     

一种新的基于支持向量机的自动调制识别方案

为了解决在合作或非合作的通信应用领域中（如软件无线电，电子侦察系统等）多种调制信号之间的切换问题，提出1种基于多类别支持向量机（SVM）的模拟和数字信号的调制识别的新方案。SVM将特征向量非线性地映射到高维特征空间中，并建立1个最优超平面来实现信号调制方式的分类。这种方法避免了在人工神经网络中的过学习、欠学习以及局部最小化的问题。仿真中将应用于调制识别的SVM算法与人工神经网络算法（ANN）做了比较，结果表明SVM自动调制识别方法结构简单，识别率高，解决小样本的能力强，在信噪比SNR不低于5dB时，正确识别率达到94％以上，适于在工程中应用。     

基于注意的图像识别模型及其在人像检测中的应用

提出了一种基于注意机制的图像识别模型。其基本想法是：在进行复杂的场景分析或目标识别时,首先通过注视控制机制,获得视景中的关键特征区域,并将注视点按照一定的顺序对这些关键特征区域进行串行扫描。在扫描到每一个关键特征区时,将对该区域附近的局部模式进行记忆或匹配。对一个复杂目标的记忆将包括2部分,一部分是对局部模式的记忆,即组成该目标的各“部件”的模式;另一部分是对各局部模式之间的空间关系的记忆,即组成该目标的各“部件”之间的结构关系。与记忆过程对应,识别过程也包括2部分,一部分是对局部模式的匹配,另一部分是对各局部模式之间结构关系的匹配。当匹配上的局部模式足够多且其空间关系正确时,目标即得到识别。用该模型对复杂背景中的人像进行了检测。实验结果表明,模型较好地解决了不变性识别的问题,即识别结果与目标物的平移、旋转和尺度变化无关,并且具有良好的鲁棒性和速度,是一种具有认知意义并且可以实用化的模型。     

Sheep don't forget a face.

The human brain has evolved specialized neural mechanisms for visual recognition of faces, which afford us a remarkable ability to discriminate between, remember and think about many hundreds of different individuals. Sheep also recognize and are attracted to individual sheep and humans by their faces, as they possess similar specialized neural systems in the temporal and frontal lobes for assisting in this important social task, including a greater involvement of the right brain hemisphere. Here we show that individual sheep can remember 50 other different sheep faces for over 2 years, and that the specialized neural circuits involved maintain selective encoding of individual sheep and human faces even after long periods of separation.     

A hybrid neural network model for consciousness

A new framework for consciousness is introduced based upon traditional artificial neural network models. This framework reflects explicit connections between two parts of the brain: one global working memory and distributed modular cerebral networks relating to specific brain functions. Accordingly this framework is composed of three layers, physical mnemonic layer and abstract thinking layer, which cooperate together through a recognition layer to accomplish information storage and cognition using algorithms of how these interactions contribute to consciousness: (1) the reception process whereby cerebral subsystems group distributed signals into coherent object patterns; (2) the partial recognition process whereby patterns from particular subsystems are compared or stored as knowledge; and (3) the resonant learning process whereby global workspace stably adjusts its structure to adapt to patterns' changes. Using this framework, various sorts of human actions can be explained, leading to a general approach for analyzing brain functions.     

基于MLP神经网络的超宽带信号自适应识别算法的研究

提出一种利用人工神经网络进行超宽带信号调制模式自适应识别的方法。采用三层MLP神经网络进行调制模式识别。首先对超宽带数字调制信号进行统计特征参数提取,特征参数作为MLP网络的输入层神经元参数,隐含层是双层结构。实验证明,当中间层采用正切型激活函数、输出层采用线性激活函数时,MLP分类器的识别性能最好。在5 dB信噪比环境下,算法的正确识别率高于95%。与传统的统计判决方法相比,神经网络分类器不需要设定判决门限就能实现自适应识别,并且达到更好的识别率,解决了软件无线电系统中的超宽带信号自动识别的问题。     

Odour-plume dynamics influence the brain's olfactory code

The neural computations used to represent olfactory information in the brain have long been investigated. Recent studies in the insect antennal lobe suggest that precise temporal and/or spatial patterns of activity underlie the recognition and discrimination of different odours, and that these patterns may be strengthened by associative learning. It remains unknown, however, whether these activity patterns persist when odour intensity varies rapidly and unpredictably, as often occurs in nature. Here we show that with naturally intermittent odour stimulation, spike patterns recorded from moth antennal-lobe output neurons varied predictably with the fine-scale temporal dynamics and intensity of the odour. These data support the hypothesis that olfactory circuits compensate for contextual variations in the stimulus pattern with high temporal precision. The timing of output neuron activity is constantly modulated to reflect ongoing changes in stimulus intensity and dynamics that occur on a millisecond timescale.     

通过DNA聚合剪切放大体系提高RNA的检测灵敏度

基于循环的DNA剪切循环放大分子机器构建了一个RNA传感器。该分子机器以RNA为输入,产生大量的DNA片段,并替换报告探针上的荧光DNA从而产生荧光信号,实现对靶RNA浓度的放大检测。本分子机器分为两部分,反应部分和报告部分。在反应部分,以靶RNA为输入条件,以一个特殊设计的探针为反应模板引发一个自发连续的DNA聚合-剪切反应网络,重复产生大量信号DNA链;这些信号DNA链进入报告部分,通过杂交替换反应从一个报告探针上替换下带有荧光DNA序列,释放到溶液中。这样通过剪切产生的大量DNA适体序列被释放到溶液中,并替换报告探针上的荧光DNA,实现信号的放大。     

利用BP算法实现集装箱编号识别

研究集装箱字符识别,采用数字图像处理,模式识别与神经网络技术及改良BP算法实现集装箱编号的自动识别,该方法有较强的适应性,可在不同的光照条件下工作,在识别倾斜,变形,磨损,污染的集装箱时表现出较强的抗干扰能力,改良BP算法具有分类功能,可以实现字符的识别,并且具有较强的鲁棒性。     

语音识别中的BP神经网络改进

语音信号是受说话人、环境等因素影响的复杂函数,人工神经网络（ANN）模型进行语音识别具有自适应、自组织及联想等特征,适合用于语音识别。主要研究了传统BP神经网络在实际语音识别中的改进方案。利用改进后的方案实现语音识别、网络结构简化、运算速度加快、训练时间缩短。