Assumes as input a constant complex signal with complex noise s[i]=(A+j B) + (wr[i]+j wi[i]) Estimates the signal power on the real part A^2 and real noise variance sigma^2.
More...
Assumes as input a constant complex signal with complex noise s[i]=(A+j B) + (wr[i]+j wi[i]) Estimates the signal power on the real part A^2 and real noise variance sigma^2.
First convert the input stream from complex type to real type, so r[i]=A + wr[i].
Then split the stream into two branches:
1) Pass the first branch through a single-pole IIR filter (averaging), square the result, and obtain an estimate of the signal power A^2.
2) Square the second branch, pass the result through a single-pole IIR filter, subtract the estimated signal power and obtain an estimate of the noise power sigma^2.
For the IIR filter, the parameter alpha controls the averaging length. See equation below:
y[i] = (1-alpha)*y[i-1] + alpha*x[i].