Linear Prediction with cross-correlation


supandey
10-16-2013, 11:03 AM
I would like to generalize equations 13.6.10-12 to use cross-correlations. Has anyone done something similar?

I assume 13.6.11 goes to (k loops over dimensions)

y_n^k = sum_m_k( d_j^k * y_(n-j)^k) + x_n

and 13.6.12 similarly goes to a double summation. However, I am not sure how to modify the memcof and predic code to accomplish this.

Presumably d is 2 dimensional (m,k). But how does one calculate num and denom in the loop in memcof? Is num the sum of all cross terms in wk1 and wk2?

Thanks,
Sanjeev