Wavelet shrinkage denoising applied to real audio signals under perceptual evaluation

This paper addreses wavelet denoising of audio recordings from a practical viewpoint. First, a set of real high-quality audio signals with distinct characteristics is artificially corrupted by pseudo-white noise. Then, the classical wavelet shrinkage denoising method is applied to them under varied settings, including different wavelets, numbers of scales, threshold application and computation methods etc. At last, an adapted version of the Perceptual Audio Quality Measure (PAQM) is used as an objective quality index to compare the obtained results. This way, the work provides some insight on the performance of wavelet shrinkage applied to corrupted high-quality signals, the performance of the PAQM in this application and the relative importance of different parameters in the final quality determination.