There exist plenty uncorrelated features in the high dimensional ECG data,so,it is difficult for the classifier based on supervised learning to perform well in both sensitivity and specificity.Pre-processed by baseline wander removing,high-frequency span removing and polynomial fitting,an auto heartbeat classification model is proposed based on supervised MCA dimension reducing.The sequence ECG data is discretized;supervised MCA dimension reducing technology is employed to extract the key features;the ECG data is classified with the common classifiers.The experiment on the PTB database shows,compared with supervised learning method,this approach combining with different classifiers has a better performance on both sensitivity and specificity.

为研究和实现珍珠的自动分选,提出了一种对珍珠的色调及饱和度进行分选的方法.将珍珠图像由RGB颜色模型转换到HIS模型中进行处理,根据,(亮度)分量直方图,通过全局分割提取珍珠光亮区,掩模后得到光亮区H(色调信息)、S(饱和度信息)均值.其中针对色调分选要求,运用了基于RPROP算法的人工神经网络学习方法;针对各色系内色彩的深浅差异,运用了K-均值聚类方法对S均值进行分类.从而实现了先按照珍珠表面色彩进行分类,再将同一色系内的珍珠按照饱和度进行再度分选,最终达到颜色分选要求.

This paper proposes a method based on hidden Markov models to detect flames in video.In addition to motion and color analysis,flame flicker process is detected by using a hidden Markov model.Results show that Markov models representing the flame and flame colored ordinary moving object are effective to distinguish flame flicker from motion of flame colored moving object.False alarms are reduced when using this method,it may have a practical application.

This paper proposes a method about image correspondence based on a speed robust feature descriptor.This relies on integral images and Fast-Hessian detector to extract interest points.For each feature point,the dominant orientation is assigned by computing Haar-wavelet responses.Based on these orientations constructing a window around the feature point,the feature points are decrypted by the sum of Haar-wavelet responses on the domination orientation and the verticality.Image correspondences are made based on distance between these descriptors.

Neural networks is used to estimate three-dimensional(3D) rigid motion parameters based two constraints.The one is based on 3D correspondences.The points' coordinates before motion are updated by the presupposed motion parameters and then compared to those after motion.The other is based two-dimensional(2D) motion fields.The motion fields from presupposition are compared to those computed from image sequences.Both network updates its weights by newton-raphson procedure for minimizing the error measures.Experimental results are presented for validating the proposed scheme.