Role of source separation using combined RPCA and block thresholding for effective speaker identification in multi source environment

Speaker Recognition is the technique used for validation of a user's identity by extracting certain features from his/her voice. This method is the most utilitarian one amongst the different biometric based recognition methods. Efficiency of speaker recognition system falls under noisy multi source environment like background noise and music. This paper uses an efficient front-end processing system to improve the performance of speaker identification in multi-source environment. In this work RPCA (Robust Principal Component Analysis) algorithm is used on mixed speech signal to separate the musical background from the signal. Other background noise is then removed using block thresholding algorithm. The separated speech from the mixed signal is then used as input for effective speaker identification.