Electrocardiograph(ECG) signal could directly reflect the human condition. The signal was contaminated by various of high-frequency and low-frequency noisy interferences in the process of collection, while the effects of the traditional wavelet denoising methods were not very ideal. This paper presented a novel thresholding method based on discrete wavelet transform(DWT), which could effectively inhibit all kinds of noises. Compared with the traditional thresholding methods, the novel approach was applied to perfectly protect the characteristics and amplitudes of the ECG signal. At last, MIT-BIH arrhythmia database was used to validate this new method which was put forward. The experimental results show the proposed algorithm can maintain the geometrical characteristics of the original signal and have high signal noise ratio(SNR), which has achieved perfect denoising effect.