# eigen values of a correlation matrix

Started by October 31, 2002
 hello ppl, i had a question regarding eigen values....i have a 1000x1 matrix....i want to compute the eigen values of the autocorrelation of that matrix....but the problem is that eigen value can only be computed of a square matrix and the autocorrelation function in matlab gives a 1999x1 matrix for mu input matrix...so i cannot find out eigen values of that... the second problem is that i have an input noise signal and i want another noise signal which is correlated with the first one (so i can use it as a reference in adaptive noise cancelling)...so how do i generate this 2nd signal being sure that its correlated with the first one... thanks in advance for all the help, cya.
 You would need to create a correlation matrix from the correlation values you have got. But usually you wont need all the 1999 correlation values. If this analysis is for your adaptive filter, it is enough to compute the correlation values upto a maximum lag equal to the length of the filter. Then form a square matrix using the correlation values you have, probably using the toeplitz function. If R is the correlation matrix then the value R(i,j) is the correlation at lag i-j (assuming a stationary signal). Pass that to the eig function to find the eigen values. To get another correlated noise signal just pass the first signal through some linear filter. If you have simulink and DSP Blockset take a look at the model 'lmsdemo' in the demos. HTH Navan --- haresh29 <> wrote: > hello ppl, > > i had a question regarding eigen values....i have a > 1000x1 > matrix....i want to compute the eigen values of the > autocorrelation > of that matrix....but the problem is that eigen > value can only be > computed of a square matrix and the autocorrelation > function in > matlab gives a 1999x1 matrix for mu input > matrix...so i cannot find > out eigen values of that... > > the second problem is that i have an input noise > signal and i want > another noise signal which is correlated with the > first one (so i can > use it as a reference in adaptive noise > cancelling)...so how do i > generate this 2nd signal being sure that its > correlated with the > first one... > thanks in advance for all the help, > cya. __________________________________________________