Responsible Professor: Prof. Gerald Schuller Supervisor: Oleg Golokolenko M. Sc.

Project description

This project aims development of a novel algorithm for sound source separation based on the analysis of phase shift between frequency bins in Short-Time Fourier Transforms of a stereo signal. The main task is to extract individual sound sources from mixed audio stream determining which frequency bins belong to the certain sound source. This algorithm should in the future be used for filtering out of undesired signals and determination of the perceived azimuth directions in a stereo signal. The project requires basic prior programming skills (Python) and interest in Signal Processing.

Project goal Novel algorithm for sound source separation based on the analysis of phase shift between frequency bins in Short-Time Fourier Transforms of a stereo signal.

Tasks - Literature study on methods for analysis of phase shift between frequency bins in Short-Time Fourier Transforms - 20% - Novel robust method for analysis of phase shift between frequency bins in FFT domain - 80%

Prerequisites - Good knowledge of Python programing - Interest in Signal Processing

References [1] Scott Rickard, "The DUET Blind Source Separation Algorithm," in Blind Speech Separation, p. 217-241. Springer, 2007 1 students / theory / programming / hardware / evaluation (Advanced Research Project / Master Thesis / Media Project)

Backback