In the data analysis of dynamic process test and measurement,the sampled data need to be pretreated to identify and eliminate the abnormal data caused by accidental interference. An improved algorithm based on Singular Spectrum Analysis is presented. By means of singular spectrum analysis smoothing method,the algorithm establishes a benchmark to identify the abnormal data,and then with the 3σ criterion,the abnormal data is eliminated and replaced by a reasonable one. The feasibility of this algorithm has be...