基于感知评价的三维信号识别的一种在线算法  

An Online Algorithm for 3D Signal Recognition Based on Ethesis Evaluation

作　　者：龚元浩[1] 杨晨晖[1] 

机构地区：[1]厦门大学计算机科学系,福建厦门361005

出　　处：《计算机工程与科学》2009年第5期56-58,61,共4页Computer Engineering & Science

基　　金：国家自然科学基金资助项目(40627001)

摘　　要：目前,各种传感器广泛应用于各种智能系统中,但对这些传感器信号的识别大多以传统的模式识别方法来完成,而这些方法一般都需要训练样本集进行学习或者需要一些先验知识。本文提出了一种新的基于感知评价的方法。该方法通过对加速传感器的三维信号进行提取、分割和识别来完成动作的识别。在信号识别阶段,我们把曲线转为一种加权图结构,我们称之为简单线形图。通过这个结构,构造了线性简单图之间相似性的一个评价函数,并用来对两个图结构进行相似性评价,最终完成动作的识别。该方法使用了三维信号的空间拓扑结构信息和极小样本,而不依赖于先验知识和大量训练样本。而且该方法是一个在线算法,能够满足很多应用的实时性要求。最后,给出了该方法的实验结果。Nowadays, there are so many kinds of sensors which have been widely used in intelligent systems. The signals of these sensors are recognized by the traditional pattern recognition methods, which need huge training sample sets or experts＇ knowledge. In the article, we propose a new method which is mostly based on ethesis evaluation. It obtains original signals, does segmentation and recognition of the signals and then recognizes the action what the signal means. In the stage of signal recognition, it uses comparison and similarity of two curves, We turn the curves to weight graphs which we call simple linear graphs. Using this structure, we construct an evaluation function for comparison. Our method uses the space topological information and very little samples instead of huge training sets or experts＇ knowledge. Moreover, it is an online algorithm, which benefits many real-time applications. Finally, we present the results of our experiment.

关 键 词：模式识别 动作识别 相似度 在线算法 简单线性图 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]