Title :
A modified Wiener filter suitable for separation of individual instrumental sounds in monaural music signals
Author_Institution :
Dept. of Electr. Syst. Eng., Hiroshima Inst. of Technol., Hiroshima, Japan
Abstract :
A FIR filter which is suitable for distilling sound of a part played by a particular instrument from a music ensemble is discussed in this report. The filter separates a monaural music signal into individual tones of every pitch with a particular timbre, the combination of which provides melodies played by an instrument. A mixture of strongly correlated signals such as the same pitch tones by different instruments can be processed by the proposed FIR filter to separate them which is modified from the discrete-time Wiener filter combined with a window and an evaluation method of current statistics of the mixture in the frequency-domain.
Keywords :
Wiener filters; correlation methods; frequency-domain analysis; music; source separation; FIR filter; correlated signals; discrete-time Wiener filter; evaluation method; frequency-domain; individual instrumental sound separation; monaural music signal separation; monaural music signals; music ensemble; sound distillation; Correlation; Instruments; Maximum likelihood detection; Multiple signal classification; Music; Standards; Wiener filters; monaural signal separation; music signal processing; statistical least squares; time-domain design;
Conference_Titel :
Norbert Wiener in the 21st Century (21CW), 2014 IEEE Conference on
Conference_Location :
Boston, MA
DOI :
10.1109/NORBERT.2014.6893934