Impedance cardiogram signal is an important parameter of cardiac function evaluation. And impedance cardiogram signal can be used to reflect the health of human body. The diagnosis of cardiovascular disease has been the important research significance and clinical value. Due to the external environment changes, the signal is vulnerable to noises, such as baseline drift, power frequency interference and so on. This paper presents a novel threshold method that is verified by clinical medical signal based on discrete wavelet transform. Compared with the traditional methods, The experimental results show that the proposed new algorithm can keep the characteristics of original signal waveform. So this new algorithm has achieved good denoising performance.