The topic of my thesis is the sound source separation in mixed audio signals, which is still an active research field of digital signal processing. After the introduction I present the audio mixing technics and the difficulties of the source separation. Additionally I present some methods, which are able to solute the problem.
Afterwards I expand particularly the Azimuth Discrimination and Resynthesis algorithm, its advantages and disadvantages, as well as practicability and conditions of applicability using different sources. The synthesis of the time domain signals happens in the frequency domain, the connection is provided by the well-known Fourier transform. The Wavelet transform is mentioned and used as an alternative variation of the Fourier analysis. After the presentation of its theory I resemble these methods, and explicate my expectations with it.
Then I compare the issues of the implemented algorithms using the different transforms and evaluate the results attending to the previous hopes. The paper is closed by summary, where I summarize the results and describe further development opportunities.