To enhance the dynamic signal de-noising effects of cable-stayed bridge health monitoring, an improved wavelet correlation de-noising algorithm is proposed in order to deal with defects of traditional one. The improved algorithm uses the main energy periods of wavelet coefficients of each layer as the correlation detection neighborhood sizes, and sets a de-noising threshold on the basis of the characteristics of cable-stayed bridge dynamic response signal. Finally the de-noised signal is obtained by a new algorithm on signal reconstruction. Two algorithms, called the traditional wavelet correlation de-noising and wavelet default threshold de-noising algorithms, are compared with the improved one in series of simulation experiments. The results show that improved wavelet correlation de-noising algorithm has better de-noising effect, thus it can be used in dynamic signal de-noising of cable-stayed bridge health monitoring.