This paper introduces a new approach based on Empirical Mode Decomposition (EMD) for explicitly denoising signals. The EMD decomposes a noisy signal into several intrinsic mode functions(IMFs), and the estimated signal is reconstructed by using the processed IMFs. In this paper, a piecewise EMD thresholding approach for denoising signal with strong noise is proposed. Simulation results show that the proposed approach has good performance, especially in the cases where the noise is very strong.