I have a 2D matrix 60 row and 5000 columns in a .mat file . each row is a registration of 5000 data points (received power values) during 5 seconds.
I want to do the following for each line of the matrix:
I wan to apply a moving average filter to smooth the data. The filter length should smooth the data but not alternate its trend. Once the data is smoothed , I should substruct the original data from the filterd data to obtain the Diffrence_data . Later, I will use a sliding window on the Diffrence_data to do some statistical anaylsis such as calculating the autocorrelation fucntion and so on. Take into consideration that, the filter length should be larger than the sliding window length.
My issues are: 1- How to determine the length of the moving filter and the sliding window for each row of the matrix?. I only know that the data should be stationary inside the the sliding window. 2- is there a way to identify the stationarity zones of each row of Diffrence_data ?
In the litterature reasearchers assume that the propagation channel is stationary during a traveled ditance of 20-100 lambda and that the sliding window length should be less than the filter length. I have tried this assumption but visually, I can depict that the 100 lambda is not stationary in some matrix rows.