Publication: Blind source separation using non-negative matrix factorisation
Loading...
Date
2022-08-01
Authors
Rohaimy, Nurakasyah Qistina
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Blind Source Separation (BSS) is the process of recovering source signals from a given mixture of unknown source signals with no prior knowledge of the source or mixing approach. The BSS can be used to identify the species of bird in a noisy environment by detecting the source from the mixing signal. Since bird audios recorded in the wild frequently contain overlapping signals, such as bird dawn chorus, audio feature extraction, and bird identification can be difficult. In this study, NMF and CNMF were used for sound source extraction. The mixtures of Bird sound signals are used as input in this experiment. Performance evaluation is described in subjective evaluation based on estimated signals and objective evaluation based on BSS_Eval toolbox metrics parameter. The both system was evaluated using the BSS_Eval toolbox, which compared the original source to the estimated source. the
signal to distortion ratio (SDR), signal to interference ratio (SIR), and signal to artifacts ratio (SAR). By using two set of data taken from internet, the average valuefor two datasets measurements is calculated. From the experiment result, based on SAR, SIR and SDR measurement, CNMF shows better result than NMF. So, CNMF is recommended in source separation to get better result.