Abstract: Body fluttering signal is weak and often involves a lot of noise arising from instrument and environment. The translation-invariant de-noising method based on improved threshold function was therefore studied to de-noise the body fluttering signal. A proper wavelet base suitable to body fluttering was selected, and the estimate method of noise variance on each and every scale after wavelet decomposition was analyzed. In addition, the capability and characteristics of improved threshold function were analyzed, based on which the translation-invariant de-noising method was applied to de-noising body fluttering signal with relevant algorithm given to implement the de-noising process. Compared the de-noised results by both hard and soft threshold functions, the body fluttering signal presents smooth wave shape with no attenuated amplitude at characteristic points. The comparison results of power spectral density between translation-invariant de-noised signal and original body fluttering signal revealed that the former can keep the characteristics of original body fluttering signal when de-noising.