Two-tier architecture for unconstrained handwritten character recognition

In this paper, we propose an approach that combines the unsupervised and supervised learning techniques for unconstrained handwritten numeral recognition. This approach uses the Kohonen self-organizing neural network for data classification in the first stage and the learning vector quantization (LVQ) model in the second stage to improve classification accuracy. The combined architecture performs better than the Kohonen self-organizing map alone. In the proposed approach, the collection of centroids at different phases of training plays a vital role in the performance of the recognition system. Four experiments have been conducted and experimental results show that the collection of centroids in the middle of the training gives high performance in terms of speed and accuracy. The systems developed also resolve the confusion between handwritten numerals.     

CDMA时变色散信道的两级盲辨识算法

本文通过对时变信道的扩展函数在时频域进行正则采样，得到了CDMA时变系统的离散正则模型。针对该模型提出了一种具有较低运算量的两级盲辨识算法。为避免两级算法产生的误差传播效应，文章将理想SIMO模型等效为加性噪声模型来处理。与基展开模型方法相比，该算法具有不需利用观测数据的高阶统计量估计展开基频率的优点。文章通过仿真验证了该算法是可行的。     

The smooth fractionator

A modification of the general fractionator sampling technique called the smooth fractionator is presented. It may be used in almost every situation in which sampling is performed from distinct items that are uniquely defined, often they are physically separated items or clusters like pieces, blocks, slabs, sections, etc. To each item is associated a ‘guesstimate’ or an associated variable with a more‐or‐less close – and possibly biased ? relationship to the content of the item. The smooth fractionator is systematic sampling among the items arranged according to the guesstimates in a unique, symmetric sequence with one peak and minimal jumps. The smooth fractionator is both very simple to implement and so efficient that it should probably always be used unless the natural sequence of the sampling items is equally smooth. So far, there is no theory for the prediction of the efficiency of smooth fractionator designs in general, and their properties are therefore illustrated with a range of real and simulated examples. At the cost of a slightly more elaborate sampling scheme, it is, however, always possible to obtain an unbiased estimate of the real precision and of some of the variance components. The only real practical problem for always obtaining a high precision with the smooth fractionator is specimen inhomogeneity, but that is detectable at almost no extra cost. With careful designs and for sample sizes of about 10, the sampling variation for the primary, smooth fractionator sampling step may in practice often be small enough to be ignored.     

基于ANN实现的供应需求预测方法

基于人工神经网络技术，建立了后勤供应需求的网络预测方法，并以某区域高校后勤保障需求为例，给出了后勤供应需求预测的步骤与实现过程。     

Variability in stepping direction explains the veering behavior of blind walkers.

Walking without vision results in veering, an inability to maintain a straight path that has important consequences for blind pedestrians. In this study, the authors addressed whether the source of veering in the absence of visual and auditory feedback is better attributed to errors in perceptual encoding or undetected motor error. Three experiments had the following results: No significant differences in the shapes of veering trajectories were found between blind and blindfolded participants; accuracy in detecting curved walking paths was not correlated with simple measures of veering behavior; and explicit perceptual cues to initial walking direction did not reduce veering. The authors present a model that accounts for the major characteristics of participants' veering behavior by postulating 3 independent sources of undetected motor error: initial orientation, consistent biases in step direction, and, most important, variable error in individual steps. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

二次型时频分析技术的开发与应用

基于二次型的时频分析是高分辨率时频分析技术的有效方法之一，该项技术能够有效反映时变信号的频谱特性，具有分辨率高、能量集中和跟踪瞬时频率的特点。从时变信号的特点出发，阐述了二次型时频分析技术的基本方法原理，采用正演模型分析验证了该方法的有效性，并在实际地震资料中应用于识别物源方向、划分沉积相带、预测储层结构以及相对厚度等诸多方面，均取得了较好的效果。     

饱和软土天然含水量估算指标及ω一e回归方程的建立

天然含水量ω是土的最基本物理量指标之一。采用数理统计方法，建立经验方程来修正土的天然含水量ω的测定值，使修正后的天然含水量数值与实际中的真值符合程度更为接近，效果甚好。     

用分数延迟改进基音预测的CELP编码方案

在ＣＥＬＰ编码器中，通常用延迟为抽样间隔整数倍的长项预测器表征浊音语音的准周期性，然而在低比特率，这种限制降低了编码器的性能。本文在介绍了ＣＥＬＰ编码器原理及激励码本构成后，重点研究了一种新型的基音预测方法；分数延迟基音预测，计算机模拟结果表明，这种方法能对浊音进行更准确的表达，尤其对女性讲话者明显改善了语音质量。     

Zero aliasing ROM BIST

Signature analyzers are very efficient output response compactors for BIST design. The only limitation of signature analysis is the fault coverage reduction (aliasing) due to the information loss inherent to any data compaction. In this article, in order to increase the effectiveness of ROM BIST, we take advantage from the simplicity of the error patterns generated by ROMs and we show that aliasing free signature analysis can be achieved in ROM BIST.This work was performed when the author was on leave from Minsk Radio Engineering Institute, Computer Department, Belorus.     

基于神经网络的河道浅滩演变预测模型

河道浅滩演变是一个复杂的非线性动力学过程, 作者借助神经网络处理非线性问题的优势, 在分析影响河道浅滩演变因素的基础上, 建立了预测河道浅淮演变的BP网络模型, 并对模型中的输入因子和样本的提取进行了探讨. 以闽江竹岐至侯官河段为实例,用“试控法”给出了BP网络模型的建模方案, 用正交设计原理选取相应的训练样本集, 利用该样本集对网络进行学习和训练, 并用训练好的BP网络模型预测浅滩上年内最小水深和年平均淤积厚度. 计算结果表明: 模型预测结果与实际值吻合良好. 这为河道浅滩演变预测研究提供了新方法.