Abstract: By using wavelet decomposition, filtering and reconstructing, noise signal could be deleted from thermogravimetry signal. It can help researchers to gain true parameters about kinetic characteristics. Different process methods were compared; several rules of threshold selection for wavelet noise removal were studied. Some parameters of processed signal curve were calculated by different threshold selection rules. By utilizing this information, the characteristics of this processed signal by different threshold selection rules were confirmed. Power-spectrum analysis was used to change the time- varying information into frequency distribution information. Quantitative comparison of signals could be made. Wavelet transform is an effective and accurate method, which can be used in thermogravimetry signal processing.