Time Series Similarity Based on Wavelet Transformation and Directional Line Element

Based on the wavelet transform, an efficiency methodology for measuring the similarity of time series is presented in this paper. Firstly, the wavelet transformation is applied to the input series to decompose it into approximations and details. Next, the approximations and the details at the highest level are partitioned into S segments with equal length. Thirdly, based on the approximation series and the detail series, the features from directional line elements amplitudes are extracted. Finally, the similarity of the two input series is calculated based on the extracted features. The simulation results indicate that the proposed algorithm enjoys high accuracy and low computational load.