M.Sc. Tezi Görüntüleme

Student: Fatih Yavuz ILGIN
Supervisor: Asst. Prof. Dr. Yusuf SEVİM
Department: Elektrik-Elektronik Müh.
Institution: Graduate School of Natural and Applied Sciences
University: Karadeniz Technical University Turkey
Title of the Thesis: Single Channel Blind Source Separation Using Nonnegative Matrix Factorisation
Level: M.Sc.
Acceptance Date: 25/6/2013
Number of Pages: 63
Registration Number: i2686

      Blind source separation that can be described as estimating each signal from a data set comprising a mixture of more than one signal. This process is called as blind means that there is no additional information on resources. When blind source separation process has mixture signal as the number of signals to be estimated, single-channel blind source separation process has only one mixture signal and because of that estimating resources is very costly operation. This method where in the bookmarked any unwanted noise, distort or otherwise signal source separated from other source or 2 different instrumental point recorded with one microphone can be separated from each other as in this study. In this study, two instrumental points recorded as a single-channel were separated from each other with a non-negative matrix factorization using a continuous wavelet transform. To evaluate the proposed method’s performance analysis, results were evaluated in terms of signal noise ratio.

      Key Words: Blind Source Separation, Cocktail Party Problem, Nonnegative Matrix Factorization, Continuous Wavelet Transform