Title :
Melody recognition system
Author :
Adamska, Katarzyna ; Pelzynski, Pawel
Author_Institution :
Inst. of Electron., Tech. Univ. of Lodz, Łódż, Poland
Abstract :
The purpose of the presented research was the development of a melody search system that would allow to find a song in a music database, based on humming the tune of a known song fragment. The melody recognition in the developed system is based on comparing vectors of voice pitch values. The best match of the humming pitch vector against all the recordings in the database is searched. An original frequency and time scaling approach was implemented to improve recognition accuracy. The system performance was tested with the use of live humming recordings of four volunteers with a small database.
Keywords :
audio signal processing; music; humming pitch vector; melody recognition system; melody search system; music database; time scaling approach; voice pitch values; Accuracy; Algorithm design and analysis; Databases; Feature extraction; Instruments; Signal processing algorithms; Time-frequency analysis; melodic contour; melody search; note frequency; query by humming;
Conference_Titel :
New Trends in Audio & Video and Signal Processing: Algorithms, Architectures, Arrangements, and Applications (NTAV/SPA), 2012 Joint Conference
Conference_Location :
Lodz
Print_ISBN :
978-8-3728-3502-4