Title :
Instrument sound separation in songs
Author :
Youssef, Khalid ; Woo, Peng-Yung
Author_Institution :
Dept. of Electr. Eng., Northern Illinois Univ., DeKalb, IL
Abstract :
This paper proposes a new solution to the audio source separation problem. The objective is to separate the original audio source signals generated by various musical instruments in one mixture. Existing approaches to solving this problem that humans easily cope with still have little success. In this paper, the blind source separation (BSS) is approached from a new point of view and is dealt with as a machine learning problem.
Keywords :
audio signal processing; blind source separation; electrical engineering computing; learning (artificial intelligence); musical instruments; BSS; audio source separation problem; audio source signals; blind source separation; instrument sound separation; machine learning problem; Blind source separation; Discrete cosine transforms; Frequency domain analysis; Humans; Instruments; Machine learning; Multiple signal classification; Neural networks; Signal generators; Source separation; Blind Source Separation; Discrete Cosine Transforms; Machine Learning; Neural Networks; Time-Frequency Domain;
Conference_Titel :
Electro/Information Technology, 2008. EIT 2008. IEEE International Conference on
Conference_Location :
Ames, IA
Print_ISBN :
978-1-4244-2029-2
Electronic_ISBN :
978-1-4244-2030-8
DOI :
10.1109/EIT.2008.4554343
