Waveform matching extension is considered to be an effective method to improve the end effect of empirical mode decomposition (EMD). According to those deficiencies of existing matching extension methods, this paper proposes a self-adaptive triangular waveform matching extension method for EMD，which ameliorated both the matching degree algorithm of triangular waveform and the searching algorithm for optimal triangular waveform. The new matching degree algorithm not only promoted the extension’s smoothness but enhanced the relationship between the data near the end-point and waveform inside. By the new searching algorithm, we firstly searched for intercepting time to cut out the local optimal wavelet in each wave band corresponding to the fixed extreme point, and then searched for the global optimal wavelet in the local optimal wavelets. Based on the simulated signals and experimental signal analysis, it is suggested that the method could effectively mitigate the end effects of EMD and significantly improve the precision of decomposition.