Explosive nuclear quadrupole resonance (NQR) signals have nonlinear and non-stationary characteristics. In order to solve the NQR signal de-noising problem, a de-noising method based on empirical mode decomposition (EMD) and wavelet transform was proposed. The original NQR signals of RDX detected by experiment were used to analyze the de-noising performance. The results indicate that the proposed method can eliminate the noise effectively and preserve the effective information of the original signals well. Meanwhile, the method can overcome the shortcomings of wavelet threshold de-noising and direct EMD de-noising, and improve the signal-to-noise ratio. The method has excellent adaptability, and proved to be an effective de-noising way for NQR signals.