Boosting the Expense and Performance of Ann／Hmm Approch for on-line Handwriting Recognition

This paper focuses on a state sharing method for an artificial neural network (ANN) and hidden Markov model (HMM) hybrid on-line handwriting recognition system. A modeling precisionbased distance measure is proposed to describe similarity between two ANNs, which are used as HMM state-models. Limiting maximum system performance loss, a minimum quantification error aimed hierarchical clustering algorithm is designed to choose the most representative models. The system performance is improved by about 1.5% while saving 40% of the system expense. About 92% of the performance may also be maintained while reducing 70% of system pararfieters. The suggested method is quite useful for designing pen-based interface for various handheld devices.     

Hybrid Genetic Algorithm Based Optimization of Coupled HMM for Complex Interacting Processes Recognition

Coupled Hidden Markov Model (CHMM) is the extension of traditional HMM, which is mainly used for complex interactive process modeling such as two-hand gestures. However, the problems of finding optimal model parameter arc still of great interest to the researches in this area. This paper proposes a hybrid genetic algorithm (HGA) for the CHMM training. Chaos is used to initialize GA and used as mutation operator. Experiments on Chinese Tai‘Chi gestures show that standard GA (SC, A) based CHMM training is superior to Maximum Likelihood (ML) HMM training. HGA approach has the highest recognition rate of 98.0769%, then 96. 1538% for SGA. The last one is ML method, only with a recognition rate of 69.2308 %.     

Gender Recognition with Face Images Based on Partially Connected Neural Evolutionary

In this paper,a new type of neural network model - Partially Connected Neural Evolutionary (PARCONE) was introduced to recognize a face gender. The neural network has a mesh structure in which each neuron didn't connect to all other neurons but maintain a fixed number of connections with other neurons. In training,the evolutionary computation method was used to improve the neural network performance by change the connection neurons and its connection weights. With this new model,no feature extraction is needed and all of the pixels of a sample image can be used as the inputs of the neural network. The gender recognition experiment was made on 490 face images (245 females and 245 males from Color FERET database),which include not only frontal faces but also the faces rotated from-40°-40° in the direction of horizontal. After 300-600 generations' evolution,the gender recognition rate,rejection rate and error rate of the positive examples respectively are 96.2%,1.1%,and 2.7%. Furthermore,a large-scale GPU parallel computing method was used to accelerate neural network training. The experimental results show that the new neural model has a better pattern recognition ability and may be applied to many other pattern recognitions which need a large amount of input information.     

Sparse Representation for Face Recognition Based on Constraint Sampling and Face Alignment

Sparse Representation based Classification (SRC) has emerged as a new paradigm for solving recognition problems. This paper presents a constraint sampling feature extraction method that improves the SRC recognition rate. The method combines texture and shape features to significantly improve the recognition rate. Tests show that the combined constraint sampling and facial alignment achieves very high recognition accuracy on both the AR face database (99.52%) and the CAS-PEAL face database (99.54%).     

Wear Debris Analysis： Fundamental Principle of Wear-Graphy

A new wear-graphy technology was developed, which can simultaneously identify the shape and composition of wear debris, for both metals and non-metals. The fundamental principles of the wear-graphy system and its wear-gram system are discussed here. A method was developed to distribute wear debris on a slide uniformly to reduce overlapping of wear debris while smearing. The composition identification analyzes the wear debris using the scanning electron microscope (SEM) energy spectrum, infrared-thermal imaging and X-ray imaging technology. A wear debris analysis system based on database techniques is demonstrated, and a visible digitized wear-gram is acquired based on the information of wear debris with image collection and processing of the wear debris. The method gives the morphological characteristics of the wear debris, material composition identification of the wear debris, intelligent recognition of the wear debris,and storage and management of wear debris information.     

Robust Hierarchical Framework for Image Classification via Sparse Representation

The sparse representation-based classification algorithm has been used for human face recognition.But an image database was restricted to human frontal faces with only slight illumination and expression changes.Cropping and normalization of the face needs to be done beforehand.This paper uses a sparse representation-based algorithm for generic image classification with some intra-class variations and background clutter.A hierarchical framework based on the sparse representation is developed which flexibly combines different global and local features.Experiments with the hierarchical framework on 25 object categories selected from the Caltech101 dataset show that exploiting the advantage of local features with the hierarchical framework improves the classification performance and that the framework is robust to image occlusions,background clutter,and viewpoint changes.     

基于Gasa-BPNN的水轮机系统辨识与预测控制

Based on the characteristics of nonlinearity, multi-case, and multi-disturbance, it is difficult to establish an accurate parameter model on the hydraulic turbine system which is limited by the degree of fitting between parametric model and actual model, and the design of control algorithm has a certain degree of limitation. Aiming at the modeling and control problems of hydraulic turbine system, this paper proposes hydraulic turbine system identification and predictive control based on genetic algorithm-simulate anneal and back propagation neural network (GASA–BPNN), and the output value predicted by GASA–BPNN model is fed back to the nonlinear optimizer to output the control quantity. The results show that the output speed of the traditional control system increases greatly and the speed of regulation is slow, while the speed of GASA–BPNN predictive control system increases little and the regulation speed is obviously faster than that of the traditional control system. Compared with the output response of the traditional control of the hydraulic turbine governing system, the neural network predictive controller used in this paper has better effect and stronger robustness, solves the problem of poor generalization ability and identification accuracy of the turbine system under variable conditions, and achieves better control effect.     

Input Split Hydraulic Hybrid Bus and Control Strategy

Simple and efficient energy management strategy is the key to ensure hybrid vehicle performance. Based on hybrid dynamical system theory and the concept of finite state mechanism,power split hybrid connected structure hydraulic hybrid system was proposed and described. In order to meet the demand for driving,the layered advanced control strategy was proposed in this paper,which referred to vehicle driving experience. Using Matlab /Simulink /Stateflow hybrid modeling method, the economy performance and the acceleration performance of the vehicle under the typical city driving cycles were carried on the simulation analysis. The results show that the proposed topology and control strategy can obviously improve engine output characteristic,effectively enhance the vehicle＇s instantaneous power performance and economy,and also has a better adaptability in different traffic environments.     

若干人脸识别算法的比较研究

Some of face recognition methods based on Principal Component Analysis （PCA）, Two-dimensional Principal Component Analysis （2DPCA） and Fisher＇s Linear Discriminant Analysis （FLDA） are comparatively studied in this paper. On the basis of the analysis of characteristics, application occasion and limitations, a new method of face recognition based on combining 2DPCA and FLDA is proposed. Finally, comparison simulations are performed with the methods and their combined methods by means of AT＆T face database. The results show that the methods based on 2DPCA can have better recognition precision than the ones based on PCA; the method based on 2DPCA and FLDA improves the real-time characteristics in the condition of keeping the recognition precision and has the best performance.     

Rank sum method for related gene selection and its application to tumor diagnosis

Tumor diagnosis by analyzing gene expression profiles becomes an interesting topic in bioinformatics and the main problem is to identify the genes related to a tumor. This paper proposes a rank sum method to identify the related genes based on the rank sum test theory in statistics. The tumor diagnosis system is constructed by the support vector machine (SVM) trained on the set of the related gene expression profiles. The experiments demonstrate that the constructed tumor diagnosis system with the rank sum method and SVM can reach an accuracy level of 96.2% on the colon data and 100% on the leukemia data.     

基于隐马尔可夫链和支持向量机人脸识别混合模型的视频节目聚类标注

人脸在视频节目中代表了重要语义信息 ,提出使用支持向量机和隐马尔可夫链混合模型对人脸进行识别 ,然后把识别结果进行高斯聚类 ,实现视频节目的内容标注 .具体步骤如下 :首先建立人脸肤色模型 ,对视频图像中可能的人脸区域进行定位 ;从定位区域提取人脸各个器官的独立基特征 ,然后使用支持向量机和隐马尔可夫链混合模型对定位区域进行人脸识别 ,最后由高斯聚类完成视频节目的语义标注     

基于离散余弦变换和支持向量机的人脸识别算法

提出了一种基于DCT提取人脸特征技术和支持向量机分类模型的人脸识别方法。利用离散余弦变换可提取人脸可识别的大部分信息,而支持向量机作为分类器,在处理小样本、高维数等方面具有独特的优势,且泛化能力很强,无需先验知识。从ORT人脸库上的实验结果可以看出,DCT特征提取是很有效的,且SVM的分类性能优于最近邻分类器,同时提高了整个系统的运算速度。     

一种基于小波神经网络混合模型的说话人识别方法

提出一种混合模型,即将隐马尔可夫模型(HMM)和小波神经网络(WNN)相结合应用于说话人识别的模型.该方法利用HMM的时序建模能力以及小波神经网络较强的模式分类能力,进行与文本无关的说话人的识别.实验表明,采用这种混合模型可以提高系统的识别率,特别在噪声环境中具有一定的噪声鲁棒性,提高了识别性能.     

一种基于离散小波变换和支持向量机的人脸识别新方法

为提高人脸识别系统的性能，提出了一种基于离散小波变换DWT（discrete wavelet transform）特征提取和支持向量机（SVM）分类的人脸识别方法。首先，采用DWT对人脸图像进行降维和去噪，然后，对小波低频子图像进行核辨别分析（KDA）提取人脸特征，最后，结合SVM进行分类识别。基于该方法，对ORL人脸库进行分类识别，采用39个特征识别率达到98．2％。仿真结果表明，该方法明显减少了高频干扰对人脸特征的影响，增强了特征的辨别能力。而且，SVM有效地提高了分类器的分类和推广能力。     

基于LWSFLA-SVM的人脸识别算法*

目的用线性调整惯性权重的蛙跳算法(linear decreasing inertia weight shuffled frog leaping algorithm,LWSFLA)训练支持向量机(support vectors machines,SVM),解决人脸识别中SVM在训练样本数较多且维数较高时,识别效果不理想的缺陷。方法该算法先用反向学习法产生初始群体提高初始解的质量,再修改最差青蛙的更新策略,并引入线性递减的惯性权重,最后应用于人脸识别中。结果与结论 ORL和CAS-PEAL-R1人脸库的仿真实验表明,LWSFLA-SVM方法的人脸识别时间短,识别率高,在训练样本不足时,识别效果良好。     

基于主元分析与支持向量机的人脸识别方法

基于支持向量机（SVM）在处理小样本，高维数及泛化性能等强方面的优势，提出了一种基于主元分析（PCA）与SVM的人脸识别方法，利用PCA方法对人脸图像进行特征提取，再利用SVM与最近邻分类器相结合的策略对特征向量进行分类识别，剑桥ORL的人极数据库的仿真结构验证了本算法是有效的。     

基于图片特征与人脸姿态的人脸识别方法研究

针对人脸识别中在非限定条件下（如背景、光照等因素发生变化时）人脸多角度多姿态识别精度低的问题与现有基于识别模型的方法无法快速更新人脸类别,提出了基于图片特征与人脸姿态的识别方法,通过对人脸姿态的识别,最大程度的匹配人脸数据库中的人脸信息,使用VGG16卷积神经网络训练模型提取图片特征,生成特征向量,再使用支持向量机分别训练提取出的特征,与人脸数据库中信息进行比对,从而精确识别人脸。通过在Pubfig与FERET人脸库上实验结果表明,所采用的算法精度较高。     

基于P2DHMM的人脸识别算法

基于隐马尔科夫模型（HMM）人脸识别算法,将改进的伪二维隐马尔科夫模型（P2DHMM）算法应用于ORL人脸库。在VisualC＋＋6．0平台下进行的实验表明,P2DHMM算法在保证较高识别精度的同时,极大地提高了人脸识别速度。     

隐马尔可夫模型及在人脸识别算法中的应用

介绍了隐马尔可夫模型(HMM)及其三大算法,并将其引入人脸识别的研究中,描述了一种基于隐马尔可夫模型的人脸识别方法.一幅正面人脸图像的重要特征具有一定的顺序,它可以通过一维的HMM来建模,每个特征区域被指定为一个状态,通过K-L变换将降维以后的特征矢量作为观察矢量.和其他人脸识别的方法比较,隐马尔可夫模型更能为人脸检测和识别提供灵活的框架.     

基于主成分分析和支持向量机的鲁棒稀疏线性判别分析方法

为了提高人脸图像的识别率、识别效率和鲁棒性,提出一种基于主成分分析（Principal Component Analysis, PCA）和支持向量机（support Vector machine,SVM）的鲁棒稀疏线性判别分析方法,通过ORL和YaleB人脸库、COIL20物体库和UCI机器学习库中部分数据集,将本文方法与线性判别分析、鲁棒线性判别分析、基于 范数和巴氏距离的鲁棒线性判别分析、鲁棒自适应线性判别分析和鲁棒稀疏线性判别分析等六种方法进行比较。实验结果表明,在ORL人脸库、COIL20物体库和UCI机器学习库的部分数据集中,在原始图像条件下,本文方法的识别率均值依次为92.80%,97.76%和89.61%,均高于其它5种方法。在YaleB人脸库加入椒盐噪声的条件下,本文方法的识别率均值为81.35%,比其它五种方法高1.37%以上。