Application of FRM filters for musical notes separation

In developing an automatic recognition system to extract musical notes from a single signal source, several problems are encountered. One of the challenging problems is the separation of multiple notes played simultaneously when the harmonics of the notes are close to one another. In this paper, the application of frequency response masking (FRM) filters to separate notes that have harmonics close to one another is presented. FRM filters are employed for their high selectivity and narrow bandwidth. The method was tested with recording from an acoustic piano. Results show that FRM filters can enhance the performance of systems for musical notes recognition.