Aimed at the disadvantages of present monitoring methods for engine misfire, a wavelet analysis method for engine misfire monitoring is studied. Firstly, by analysis of vibration acceleration signal spectrum on cylinder head, it can be known that amplitudes of 0.5 and 1-order have much more growth under single-cylinder misfiring situation than normal one. Collecting the signal around the time of engine misfiring occurred under constant-speed condition. Then,multi-scale approximate decomposition is performed by wavelet analysis method for the vibration acceleration signal, and the low-frequency approximate signal is obtained. And the fault can be easily identified by vibration amplitude fluctuation. Similarly, the same analysis results are gained under acceleration and deceleration conditions. The result shows that wavelet analysis method can effectively monitor the occurrence of engine misfire fault.