Wavelet analysis techniques of digital audio processing are considered. Tonality formalization and corresponding proof-of-concept detection algorithm are provided. Wavelet approach is applied to the clustering problem on the example of Beethoven's sonatas. The method for effective signal denoising and analysis in both time and frequency domains is constructed based on the properties of discrete wavelet transform.