Title :
Unsupervised singing voice separation from music accompaniment using robust principal componenet analysis
Author :
Umap, Priyanka K. ; Chaudhari, Kirti B. ; Joshi, Madhuri A.
Author_Institution :
Dept. of Electron., AISSMS Coll. of Eng., Pune, India
Abstract :
Singing voice separation from monaural recording is important for many real world applications. Various algorithms have been put forward for audio separation such as feature based, beamforming and computational model based separation. In following paper we have used Robust Principal Component Analysis for separation purpose which decomposes the audio signal into sparse and low rank components. The musical accompaniment can be assumed as a low rank subspace as musical signal pattern is repetitive in nature. Likewise singing voices contained in song can be considered as relatively sparse in nature. We examine performance of the algorithm by various performance measurement parameter such as source to distortion ratio (SDR), source to artifact ratio (SAR), source to interference ratio (SIR) and GNSDR in unsupervised systems.
Keywords :
audio signal processing; music; principal component analysis; speech processing; audio separation; audio signal; low rank components; music accompaniment; musical signal pattern; robust principal component analysis; source to artifact ratio; source to distortion ratio; source to interference ratio; unsupervised singing voice separation; unsupervised systems; Radio access networks; Augmented Lagrange Multiplier(ALM); Robust Principle component Analysis(RPCA); Singing Voice Separation;
Conference_Titel :
Industrial Instrumentation and Control (ICIC), 2015 International Conference on
Conference_Location :
Pune
DOI :
10.1109/IIC.2015.7150974