中文摘要：

本文在Lo等（2000）研究的基础上，给出了基于局部极值点刻画的4种技术形态的数学定义，并基于局部多项式估计方法给出了这些形态的自动识别算法。应用所提出的算法对我国沪市A股进行了图形技术分析的实证研究。研究结果表明本文定义的技术形态包含一般统计意义上显著的附加信息，从而证明了图形技术分析的有效性。

英文摘要：

Based on Lo etc＇s work（2000）, we propose four mathematical definitions for technical pattern by local extreme methods and give an automatic algorithm the technical pattern recognition by using nonparametric local polynomial method. Then we do empirical study on Shanghai stocks for the chart-based technical analysis by applying the proposed method. Empirical results show that the chart- based technical analysis contains general significant additional information in statistical sense, which illustrates its effectiveness.