Iris recognition is an effective method of biometrics recognition. Empirical mode decomposition(EMD),a multi-resolution decomposition technique,is adaptive and appears to be suitable for nonlinear,non-stationary data analysis. We adopt the EMD approach to decompose the iris images and select the intrinsic mode functions with proper frequency range for iris recognition. The experimental results indicate that the recognition rate can achieve 99.44%; meanwhile,the complexity of the algorithm can be reduced because the effect of high frequency noise and illumination can be eliminated during our feature extraction process.