Software development of blind source separation of music
Loading...
Date
2019-06
Authors
Mohamad Rasyidi Yazid
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
In popular music, a cover version or cover song is a new performance or recording by someone other than the original artist. However, it is impossible to retrieve a piece of single music from the song (instrument music and singer’s vocal). Therefore, the goal of this study is to deliver a program that separates a record into several tracks by using Blind Source Separation (BSS) method which is portable by using Android-based smartphone. Android operating system (OS) is chosen because it is a popular open source OS whereby a developer can customize the OS and create their own application (App). An android app is developed for the smartphone to select the song, processing the separation and receive the result back of separation of sound from a server through the internet. A computer server is used in this project to solve the limitation memory and processing power of the smartphone. In this study, FastICA with kurtosis and FastICA with negentropy have been used for separation of music. Performance evaluation on the estimated signal are based on waveform characteristics as subjective evaluation. From the results, based on average SIR, SNR, and SDR measurements, FastICA with kurtosis obtain 42.245 dB, 56.629 dB, and 40.148 dB respectively while the performances of FastICA with negentropy are 44.089 dB, 44.820 dB and 43.739 dB respectively. In conclusion, the sseparation of music by BSS approach is viable and the use of the separation software on android application(App) is such a practical approach.