基于小波的时间序列的形状匹配算法

目的提出利用小波的方法对时间序列进行形状匹配，提高匹配的精度．方法将目标形状转换为时间序列的形式后，利用小波对时间序列的逐层降维处理能力，结合欧几里德距离公式准确地计算出不同目标之间的相似度，达到了满意的匹配结果．结果采用真实数据集，分别对特殊点方法、Fourier描绘子方法、形态描绘子方法以及小波方法等4种形状匹配方法进行了实验．实验结果表明，与传统形状匹配算法相比，小波方法在不同序列长度情况下匹配精度都有较大提高．结论此方法具有良好的稳定性和可靠性，并且结合了小波快速降维的优点，保证了对闭合轮廓形状的整体匹配实时处理的能力．

Shape Matching Algorithm of Time Series Based on Wavelet

With the method based on wavelet, a study on the shape matching of time series was carded out to improve the accuracy of the shape matching problem. After the objective shapes were transformed into time series, we used the dimensionality reduction and the Euclidean distance function which reflected the distance between objective shapes precisely to fulfill the query purpose. Using the real dataset, we tested the four shape matching methods of special point method, Fourier describe sub-method, orthography sub-method and wavelet method. The result shows that the matching precision of the method based on wavelet improves greatly under the different series length than the traditional shape matching algorithm. The method has a high efficiency and scalability and guarantees real time ability of fast wavelet algorithm used in closed contour shape of the overall match.