Abstract: To realize correction of retention time shift of Chinese medicine fingerprint,a preliminary study on the ordinary least fitting algorithm is carried out,which is not robust enough to effectively handle the problem of retention time shift of Chinese medicine fingerprint because of the complicated multi-component signals.By studying the ordinary least squares fitting according to the inherent features of Chinese medicine fingerprint signal,one practical line fitting method,i.e.local least squares fitting was putforward.A brief description of this method is made to solve the retention time shift of Chinese medicine fingerprint under different experimental conditions.Three real-life samples are corrected and the compared results illustrate the improvement of the similarity degree of the Chinese traditional medicine fingerprint.