Batch Learning Techniques for Blind Signal Separation
Yongjian Zhao, Manlan Hao · 2017 International Conference on Smart Grid and Electrical Automation (ICSGEA) · 2017
Simultaneous blind signal separation (BSS) methods aim to estimate all source signals at a time. In many applications, only one or a few source signals are desired and the rest are considered to be noise. Many natural signals such as biomedical signal exhibit typical temporal structures. After time delayed correlation about desired source signal is exploited, a constrained optimization problem based on linear autocorrelation of the desired signal is introduced. Then a batch learning approach, which may separate source signal with linear autocorrelation, is presented correspondingly. Someone can select to estimate one or a few source signals from a large number of sensor signals, thus saving a lot of computing resources. Computer simulations demonstrate the efficiency of the proposed approach.