A novel denoising method for acoustic signal
Qiuju Li, Zhaotong Zhu, Chuan Xu, Yinzhou Tang · 2017
A novel denoising method which introduces guided filter form image processing filed to denoise acoustic signal is proposed in this paper. It is important to choose a right guidance signal for the guided filter and wavelet denoising will be used to get guidance signal in this paper. For acoustic signal, getting the actual signal without noise is difficult, and it will affect the identification of the optimal wavelet denoising result. In other words, it is difficult for us to get a better guidance signal. On that account, a novel validity assessment method of wavelet denoising, which combines root mean square error with smoothness, will be introduced to help get the better guidance signal. Guided filter with better guidance signal is good at not only denoising but also preserving the sharp part of the acoustic signal. Experiments show that the novel denoising method, which first uses wavelet denoising with the novel validity assessment method to get the better guidance signal, then uses guided filter to denoise the acoustic signal, is an effective acoustic signal denoising method.