In reconstructing impact forces, the impulse response function (IRF), which characterizes the relationship between the impact force (cause) and the resulting structural response (effect) has always been considered an important factor and requires prior determination. Unfortunately, as shown in previous studies, the IRF is usually unknown in advance and is also ill-posed when determined. Most previous studies have either implicitly estimated the IRF or focused solely on selecting the response measurement location to obtain the IRF that yields a more accurate reconstruction of the impact force. This paper focuses on investigating the effect of IRF on the reconstruction of impact forces by applying wavelet deconvolution techniques to both the estimation of the IRF and the reconstruction of the impact force. Considerations of appropriate scales and shifts of wavelets in estimating the IRF and reconstructing the impact force can be regarded as a kind of robust regularization to improve the nature of ill-posed problem. Numerical and experimental data are used to validate the feasibility of the proposed method. The results demonstrate that the application the wavelet deconvolution technique with appropriate consideration of scales and shifts during IRF estimation is more effective in enhancing the accuracy of the impact force reconstruction.