Abstract：The traditional evaluation index cannot meet the requirements of wavelet de-noising quality evaluation if the truth value is unknown. Using rate-change characteristics, this paper proposes a new evaluation index that reconstructs the root mean square error variation evaluation index and smoothness variation evaluation index. We use the normalized entropy weight method to linearly combine the two normalized evaluation indices. This new index is the composite evaluation index. The method decides the optimal de-noising layer by means of the obvious change of the rate of change of the index as the number of decomposition layers increases. Experimental data analysis shows that this method can guide wavelet decomposition accurately, and accurately determines the optimal de-noising layer to achieve the optimal de-noising effect when the truth value is unknown.