一种改进的Fuzzy-HMMS算法及其收敛性

对D at T ran和M ichae lW agner等人提出的FCM-FE-HMM S算法作了进一步的补充和改进,提出了改进的FCM-FE-HMM S算法,并给出了算法收敛性分析,得出在给定初值的情况下该算法将收敛到一个局部最优解。     

Future of biological aerosol detection

Biological aerosol detection in real time is an urgent civilian and military requirement. Such detection capability will be useful in environmental monitoring, for example, in gathering information in perceived hazardous areas such as housing developments downwind of sewage treatment plants. To be truly functional, the instrument has to operate continuously, 24 h a day and 7 days a week with minimal maintenance and few false alarms. A novel concept is proposed. The system employs a rapid front-end warning/alarming mechanism based on optical technologies that provides useful information for protection decision makers. This is connected to a sample collector that feeds a slower back-end liquid chemistry system that provides analytical results to the medical personnel to assist in prophylaxis and therapy decisions. Experience gained from measuring fluorescence signals of single bacterial spores under flow cytometry (FCM) using UV excitation at 340-360 nm, was applied to concept testing of a prototype instrument, built to do the same for aerosols. This machine was capable of resolving particle size as well as fluorescence intensity of each particle under laboratory and field conditions; it was called the fluorescent aerodynamic particle sizer (FLAPS). This paper describes practical aspects of measuring biological aerosols when the results must be compared to reference samplers that provide culturable or “live” data. Treatment of particle size and fluorescence information is discussed with respect to FLAPS and reference data fidelity. Along with an objective method to evaluate FLAPS data correlation to reference data, an approach for determining limit of detection in the field is discussed. In addressing the back-end detector chemistry, we have prioritized a number of important biological characteristics that must be given to a clinician to help in prophylaxis and therapy decisions. A series of biochemical measurements are proposed to define the threat of a sample and different solutions are given to implement these tests. We predict that the future for biological detection looks promising for fluorescence in situ hybridization (FISH) techniques in identifying microorganisms. A conceptual instrument based on merging FCM and microchip-based analysis is described.     

基于改进模糊C均值聚类的弥散张量成像图像分割

为了避免随机选取初始聚类中心点的缺陷,利用最大最小距离的方法确定初始聚类中心点.实验结果表明,和传统的模糊C均值聚类(FCM)算法相比,所提聚类算法具有较高的稳定性和准确性,所分割的胼胝体图像边缘信息更加清晰.     

基于AIC方法的切换神经网络模型设计

将切换系统设计中的切换思想与神经网络相结合,构建了切换神经网络模型.根据模糊C均值（FCM）聚类方法将样本数据分为多组训练数据,每组数据对应训练一个单一神经网络模型,再利用赤池信息准则（AIC）制定相应的切换规则.根据输入数据特性,选择单一网络或多网络组合的输出作为模型输出,从而达到函数逼近目的.本模型更好地利用了各个子网络在特定区域具有较高逼近精度的特点.仿真结果表明,切换神经网络模型有较高的逼近精度.     

一种模糊高斯混合说话人识别模型

为了研究模糊聚类算法在高斯混合模型（GMM）参数获取方面的应用,采用模糊C均值算法（FCM）进行语音特征矢量的聚类,并结合Tabu搜索算法得到全局最优的聚类结果,进一步用EM算法得到GMM模型参数.使用TIMIT数据库中的语音进行测试,开集和闭集说话人辨认实验都表明,该方法获取的GMM参数比普通EM算法获得的GMM模型参数性能更优,能有效降低说话人辨认系统的误识率.     

基于改进神经网络的模糊聚类算法

文章提出了一种新的聚类方法NFC,首先用模糊逻辑神经元网络的聚类算法和Cauchy训练的模拟退火算法相结合的局部算法得到初始聚类中心,然后用FCM算法进行模糊聚类;实验证明,NFC算法在一定程度上解决了FCM局部极值问题且有效性非常高。     

基于自适应动态时间规整(DTW)的GA-FCM多阶段间歇过程故障诊断

多时段是间歇过程的固有特征,对间歇过程划分阶段可以提高故障诊断的精度。采用模糊C-均值聚类（FCM）算法划分阶段存在对初始聚类中心敏感、易于陷入局部极优值的问题。提出遗传算法与FCM算法相结合的方法（GA-FCM）,用于克服FCM易于陷入局部极优值的问题,以达到全局最优。同时,针对间歇过程数据不等长问题,提出自适应动态时间规整（DTW）算法。随后,用GA-FCM方法完成阶段划分,再建立多向核主元分析（MKPCA）模型完成故障检测。最后将此算法应用于青霉素发酵过程,仿真结果验证了所提方法的可行性和有效性。     

A segmentation method for noisy photoelectric image

In order to improve the ability of noisy photoelectric image segmentation and satisfy the requirement of human visual apperception, a noisy photoelectric image segmentation method is proposed in this paper. Firstly, the basic principle of FCM algorithm is analyzed in detail. Secondly, on the basis of PCM algorithm, only such pixels affected by noise lesser that centralized in the diagonal area of two-dimensional histogram is actualized image processing, and the probability of the pixel that lie in such area is obtained. Thirdly, objective function, membership matrix and cluster centers are renovated on the basis of PCM algorithm. Finally, experiment contrast between our method and other methods is executed, the results of experiment indicate that our method has relatively better segmentation quality and faster segmentation speed.     

基于FCM和标记分水岭的粘连岩石颗粒图像分割

模糊C均值算法可利用图像的多种特征值进行准确的图像分割,但不能分割粘连物体;传统的分水岭分割算法能够获得准确的物体边缘轮廓,但容易造成过分割.为了解决这个问题,提出基于FCM和标记分水岭的粘连图像分割.该方法首先对原始彩色图像中值滤波后进行基于LUV颜色空间的FCM聚类;对聚类后的图像用形态学方法去杂质、空洞填充后进行距离变换;然后根据距离变换图像找出局部最大值,得到种子图像;最后对距离变换图像进行基于标记的分水岭分割,得到最终的分割图像.该方法对粘连岩石颗粒图像进行分割,取得了较好的实验效果.     

基于改进的模糊c均值聚类水凝胶图像分割

基于实验图像中目标与背景区域具有明显的色差和形状差异且其直方图有明显双峰的特性,提出一种改进的模糊c均值聚类和数学形态学相结合的算法,将目标图像有效地分割出来.首先依据像素邻域间的约束属性以及二维直方图对角线元素含有稳定图像信息的特点,在模糊c均值聚类算法基础上对聚类目标函数进行改进,并将二维直方图对角函数引入隶属度函数中,最终将目标区域有效地分割出来.实验结果表明:该方法具有识别精度高、计算量小、去噪能力强的特点.