How does MATLAB built in autocorrelation function work without knowing the distribution?

242 Views Asked by At

Since the autocorrelation function is defined as : $$R(s,t)={\frac {\operatorname {E} [(X_{t}-\mu _{t})(X_{s}-\mu _{s})]}{\sigma _{t}\sigma _{s}}}\,,$$ and requires the expected value and therefore, the distribution of the signal, how do MATLAB built in functions like autocorr work without knowing the distribution? How does it calculate the expected value? Thanks in advance.

1

There are 1 best solutions below

2
On

AFAIK, MATLAB's autocorr function is not dealing with a "theoretical" random process but with a time series, which may be a sample from this random process. So it's not using expected values of the distribution, but rather averages of the observed values.