Kurtosis-Switching Based Blind Source Separation for Linear Mixed Signals
This algorithm addresses blind source separation of linearly mixed signals when source probability density functions and activation functions are challenging to determine - particularly when signals contain both super-Gaussian and sub-Gaussian components. By utilizing kurtosis as a statistical measure of signal PDFs, the method adaptively configures activation functions through a switching mechanism, implementing an effective kurtosis-based blind separation approach with robust mixed-signal handling capabilities.