Numerical simulations to support the normalization data reduction technique

The normalization data reduction (NDR) technique is an analytical methodology for characterizing the upper shelf fracture toughness of steels in the ductile regime, both in terms of critical toughness (J_Ic) and resistance to ductile crack extension (J-R curve). It represents an alternative to the more commonly used multi-specimen or single-specimen (unloading compliance and potential drop) techniques.Finite element analyses of a growing crack are executed to evaluate the performance of the technique. This approach has the advantage to remove large uncertainties entailing experimental results. Results demonstrate the precision of the method.     

Optimisation of cobalt-based hardfacing in carbon steel using the fuzzy analysis for the robust design

In this paper, we present the application of the fuzzy logic analysis to a Taguchi orthogonal experiment for developing a robust model with high efficiency in multiple performance characteristics (MPCs) of the plasma transfer arc welding (PTAW) hardfacing process. It eliminates uncertain information and is a simple, effective, and efficient approach. A fuzzy logic system is used to simultaneously investigate relationships between various MPCs and to determine the efficiency of each trial of the Taguchi experiments. From the fuzzy inference process, we are able to determine the optimal setting of factor-levels for the MPCs. In addition, we also use the analysis of variance (ANOVA) to identify the significant factors, which coincide with findings from the fuzzy logic analysis and are found to account for about 79% of the total variance. Furthermore, a confirmation experiment of the optimal process is conducted, and it verifies that both individual performance characteristics and MPCs are successfully optimized and satisfy our desired levels of MPCs.     

Speaker Normalization and Model Selection of Combined Neural Networks

This paper introduces bootstrap error estimation for automatic tuning of parameters in combined networks, applied as front-end preprocessors for a speech recognition system based on hidden Markov models. The method is evaluated on a large-vocabulary (10 000 words) continuous speech recognition task. Bootstrap estimates of minimum mean squared error allow selection of speaker normalization models improving recognition performance. The procedure allows a flexible strategy for dealing with inter-speaker variability without requiring an additional validation set. Recognition results are compared for linear, generalized radial basis functions and multi-layer perceptron network architectures.     

Class normalization in centroid-based text categorization

Centroid-based categorization is one of the most popular algorithms in text classification. In this approach, normalization is an important factor to improve performance of a centroid-based classifier when documents in text collection have quite different sizes and/or the numbers of documents in classes are unbalanced. In the past, most researchers applied document normalization, e.g., document-length normalization, while some consider a simple kind of class normalization, so-called class-length normalization, to solve the unbalancedness problem. However, there is no intensive work that clarifies how these normalizations affect classification performance and whether there are any other useful normalizations. The purpose of this paper is three folds; (1) to investigate the effectiveness of document- and class-length normalizations on several data sets, (2) to evaluate a number of commonly used normalization functions and (3) to introduce a new type of class normalization, called term-length normalization, which exploits term distribution among documents in the class. The experimental results show that a classifier with weight-merge-normalize approach (class-length normalization) performs better than one with weight-normalize-merge approach (document-length normalization) for the data sets with unbalanced numbers of documents in classes, and is quite competitive for those with balanced numbers of documents. For normalization functions, the normalization based on term weighting performs better than the others on average. For term-length normalization, it is useful for improving classification accuracy. The combination of term- and class-length normalizations outperforms pure class-length normalization and pure term-length normalization as well as unnormalization with the gaps of 4.29%, 11.50%, 30.09%, respectively.     

Spectral unmixing of normalized reflectance data for the deconvolution of lichen and rock mixtures

In subarctic regions the ubiquitous presence of rock encrusting lichens compromises the ability to map the reflectance signatures of minerals from imaging spectrometer data. The use of lichen as an endmember in spectral mixture analysis (SMA) may overcome these limitations. Because lichens rarely completely occupy the Instantaneous Field of View (IFOV), it is difficult to define a lichen endmember from an image using visual or automated endmember extraction tools. Spectral similarity of various crustose/foliose lichen species in the short wave infrared (SWIR) suggests that spectral unmixing of rock and lichens may be successfully accomplished using a single lichen endmember for this spectral range. We report the use of a spectral normalization method to minimize differences in SWIR reflectance between five lichen species (U. torrefacta, R. bolanderi, R. geminatum, R. geographicum, A. cinerea). When the normalization is applied to reflectance spectra from 2000-2400 nm acquired for a lichen encrusted quartzite rock sample we show that only a single lichen endmember is required to account for the lichen contribution in the observed mixtures. In contrast, two such endmembers are required when the normalization is not applied to the reflectance data. We illustrate this point using examples where endmembers are extracted manually and automatically, and compare the SMA results against abundances estimated from digital photography. For both the reflectance and normalized reflectance data, SMA results correlate well (R²>0.9) with abundances estimated from digital photography. The use of normalized reflectance implies that any field/laboratory lichen spectrum can be selected as the lichen endmember for SMA of airborne/spaceborne imagery.     

一种快速有效的印刷体文字识别算法

为了利用低成本的硬件来实现对印刷体文字的快速识别,提出了一种基于多级分类的印刷体文字快速识别算法,该算法从预处理、特征提取,到模式匹配各个阶段,都对传统方法作了合理的改进.该算法首先是采用36×36,而不是传统的48×48点阵进行归一化,从而有效地减少了计算量和字典容量;其次是采用改进的粗外围特征,并进行二重分割,以提高特征的稳定性;最后在各级分类中采用了不同的判别准则,包括绝对值距离、欧氏距离及相似度准则,以适应于时间、准确性的不同要求.同时用该算法对一级汉字7000个样本进行了实验,其结果表明,实际正确识别率(正识率)达95%,前5位累积正识率可达98%,从而为“电子阅读笔”的开发与研制打下了坚实的理论基础.     

一种测量固体材料含水率的水分传感器

一种用于测量固体材料含水率的水分传感器已研制成功．它的原理及特点是通过大量实验数据的理论分析得到了各种固体材料含水率的归一化的线性数学德型，运用单片机软件功能实现的．传感器测量范围为4％～50％，精度高于0.5％.     

Normalization of discrete planar objects

The case is made for normalization of discrete planar objects prior to comparison with test objects and an expression for normalization is derived using a Euclidean distance function based on the underlying continuous boundaries of the objects and their prototypes. The results are given in both the spatial and the frequency domain; an analysis of errors due to the quantization introduced by using a discrete grid is also given.     

基于相似度聚类分析的动态背景建模与减除

对复杂动态背景建模技术进行了研究,提出了一种基于像素相似度聚类分析的动态背景建模算法.该算法首先建立融合亮度和色度信息的像素相似度理论,然后用像素相似评价标准对图像各像素点像素值的时间序列进行聚类分析,以建立动态背景模型.最后,利用该背景模型进行了多场景前景检测对比实验和算法内存消耗以及算法单帧处理时间等性能测试.实验结果表明,该算法前景检测准确度高,时间和空间上的复杂度低.