I have recently worked toward applying compressed sensing technique to the case of which sensing matrix is sampled from biased Bernoulli distribution.
So far I have found literature demonstrating that the sensing matrix sampled from gaussian distribution or symmetric Bernoulli distribution more or less satisfies RIP condition.
However, I haven't came across with the works demonstrating for the matrix sampled from biased Bernoulli distribution. Any comments/feedback/reference to the related works will be much appreciated.
Also, I have got some feedback that If I construct the sensing matrix by subtracting the average from the initial matrix(which is sampled from biased Bernoulli distribution), that matrix will likely to satisfy RIP.Is there a paper/work out there that I can make sense of this (non-math, non-CS major), please do let me know.
Thank you very much!
