Abstract: Metal magnetic memory testing technology can test the broken wires and evaluate the stress concentration area of mining steel rope. However, the test of rope always needs real-time online detecting; the tested signal contains a lot of noise. To improve the recognition rate of broken wires, the signal needs to be denoised. The denoising result shows that using Ensemble Empirical Mode Decomposition (EEMD) to denoise the signal is better than wavelet denoising. A new method is proposed to filtrate Intrinsic Mode Functions (IMF).