深度连续卷积神经网络模型构建与性能分析

为了提升卷积神经网络特征提取能力,设计了一种基于连续卷积的深度卷积神经网络模型.该模型采用小尺度的卷积核来更细致地提取局部特征,并借助连续的两个卷积层增加模型的非线性表达能力,结合Dropout技术降低神经元之间的相互依赖,利用抑制网络过拟合对模型进行优化.人脸表情、手写数字字符和彩色图像的目标识别实验表明,在图像较为复杂时,该模型在识别的准确性和泛化性能上比手工特征提取方法及一般的2、3层卷积结构具有明显的优势.     

基于Filter Wrapper模式的特征选择算法*

特征选择是数据挖掘、机器学习和模式识别中始终面临的一个重要问题。针对类和特征分布不均时,传统信息增益在特征选择中存在的选择偏好问题,本文提出了一种基于信息增益率与随机森林的特征选择算法。该算法结合Filter和Wrapper模式的优点,首先从信息相关性和分类能力两个方面对特征进行综合度量,然后采用序列前向选择（Sequential Forward Selection, SFS）策略对特征进行选择,并以分类精度作为评价指标对特征子集进行度量,从而获取最优特征子集。实验结果表明,本文算法不仅能够达到特征空间降维的效果,而且能够有效提高分类算法的分类性能和查全率。     

基于群稀疏约束的语音识别特征混合判别分析

为了克服因数据不足,而造成较难提取稳定的长时特征的问题,提出了一种基于群稀疏约束的混合判别分析方法。该方法首先采用高斯混合模型描述数据的分布,在此基础上利用二次变分的形式进行群稀疏的表示,得到基于群稀疏约束的混合判别分析目标函数。接着,通过定义模糊响应矩阵（blurred response matrix）,有效地结合最优化得分方法求解判别分析变换矩阵。最后,拼接相邻帧梅尔滤波器组输出组成超矢量,采用变换矩阵进行变换降维,提取时频特征。实验结果表明,本文方法能够得到稀疏的变换矩阵,相比于PLDA(Penalized LDA)和SLDA（Sparse LDA）判别分析方法,识别准确率分别提高了0.71%和1.53%,且在数据不足的条件下,本文方法能获得更高的识别性能。     

自由活塞汽油直线发电机燃烧过程特性

为了获得自由活塞汽油直线发电机工作过程中的燃烧特性,根据试验测得的活塞运动曲线对其燃烧过程进行了CFD仿真研究,并对比传统发动机分析了两种发动机活塞运动规律差异对燃烧过程的影响规律。研究发现:与传统发动机相比,自由活塞汽油发电机缸内最高燃烧压力和平均温度均较小,且最高燃烧压力保持时间较短,这有助于减少NO污染物;自由活塞直线发电机燃烧放热率呈现"缓和长"的特点,燃烧放热过程等容度较低;自由活塞直线发电机较少的混合气在上止点前完成燃烧,压缩行程负功较小,但由于后燃较严重,其指示效率较传统发动机低。     

基于PSO的模拟电路故障信息特征提取

为了提高容差电路故障诊断的诊断效率,改善电路故障征兆的特征描述,采用基于信息熵的方法生成被测容差电路的故障特征,并提出了基于离散粒子群的容差电路故障信息的特征提取方法,由此确定用于故障诊断的最佳特征子集,并以此训练故障分类器,对容差电路的软故障进行诊断定位。实验证明了基于粒子群和信息熵的特征描述和特征提取方法的有效性。     

采用路径IRP的Windows恶意进程检测方法

针对程序在同一操作系统的不同环境下运行产生的IRP(I/O request packets)序列不完全相同,对检测结果有一定影响的问题,提出了采用路径IRP的Windows恶意进程检测方法.单独提取每一个操作路径的IRP请求序列,应用朴素贝叶斯、贝叶斯网络、支持向量机、C4.5决策树及改进的人工免疫算法(IAIS)进行检测,并比较了各种算法在不同特征选择方法下的检测效果.实验结果表明,本文所提出的采用路径IRP的Windows恶意进程检测方法是有效可行的,在所有方法中,采用Fisher Score进行特征选择的朴素贝叶斯方法得到了最高的检测率99.2%,优于基于IRP序列的恶意进程检测方法.     

Segmentation of the image with multi-visual features for a traffic scene

Scene segmentations based on the pixel classifying calculation are complicated, and they use insufficient features, thus resulting in a low accuracy, so a new model is proposed to overcome these shortcomings,which is to learn these geometric classes based on multi-visual features of super-pixels. First, various features are extracted from the super-pixels of an input image. These features are used for classifying the super-pixels. Then the difference between the adjacent super-pixels is calculated to predict their consistency. The initial classification result and the consistency are synthesized to the Markov Random Field energy function, which is then minimized based on the graph-cuts algorithm to get the final labels of the super-pixels. Experimental results prove the effectiveness of the multi-visual features and the optimization method proposed, with superior performance achieved for traffic scenes.     

Adaptive key SURF feature extraction and application in unmanned vehicle dynamic object recognition

A new method based on adaptive Hessian matrix threshold of finding key SRUF( speeded up robust features) features is proposed and is applied to an unmanned vehicle for its dynamic object recognition and guided navigation. First,the object recognition algorithm based on SURF feature matching for unmanned vehicle guided navigation is introduced. Then,the standard local invariant feature extraction algorithm SRUF is analyzed,the Hessian Metrix is especially discussed,and a method of adaptive Hessian threshold is proposed which is based on correct matching point pairs threshold feedback under a close loop frame. At last,different dynamic object recognition experiments under different weather light conditions are discussed. The experimental result shows that the key SURF feature abstract algorithm and the dynamic object recognition method can be used for unmanned vehicle systems.     

Novel Method Fusing (2D)2LDA with Multichannel Model for Face Recognition

A fusion method of Gabor features and (2D)²LDA for face feature extraction is proposed in this paper. Gabor filters are utilized to extract multi-direction and multi-scale features from facial image to employ its robust performance for illumination, expressional variability and other factors. The extracted features have the defect of high dimension and redundancy data. (2D)²LDA is implemented to reduce the dimension of Gabor features and select effective feature data. Finally, the nearest neighbor classifier is used to classify characteristics and complete face recognition. The experiments are implemented by using ORL database and Yale database respectively. The experimental results show that the proposed method significantly reduces the dimension of Gabor features and decrease the influence of other factors. The proposed method acquires excellent recognition accuracy and has light architectures as well.     

基于分布估计算法的人脸Haar特征选择

通过研究人脸检测算法中Ada Boost算法,针对算法中的haar特征维数过高、训练耗时过长,检测效率过低等问题.提出基于分布估计算法(Estimation of Distribution Algorithm,EDA)的人脸haar特征选择人脸检测.EDA采用类内类间比作为适应度函数,通过统计学习的手段建立解空间内个体分布的概率模型,然后对概率模型随机采样产生新的群体,进行反复计算,实现群体的进化,最终得到全局最优解,以此来实现haar特征选择.实验结果表明:检测率(DR)与误检率(FDR)优于传统算法,而且检测速度得到了提升.