Wavelet denoising method was introduced to improve the validity of identifying the guide rail irregularity in elevator. The horizontal acceleration signal measured from a test elevator was processed by wavelet denoising. Then the denoised signal was decomposed up to four levels using Daubechies 4 mother wavelet to identify the guide rail irregularity. The results indicate that the impulses due to the guide rail irregularity are prominent in wavelet decompositions. And the causes resulting in those characteristics of the vibration signal can be well revealed. So discrete wavelet transform (DWT) can be used as an effective tool for denoising the horizontal vibration signal of elevator and diagnosing faults in the guide rail.