Correlation for finding peaks

I have a 32x25 matrix with 800 elements. There are 50 Training bit values, whose locations are known and are fixed in the matrix. The 800 elements are indices which can take up any value between 1 and 64 (decimal) or hexadecimal or binary. Data given to me has 37 assigned as the training bit value. So, of the 800 elements, I will have 50 37's and the rest 750 elements can take up any value between 1 and 64. I convert this 2D matrix into a 800x1 vector and keep this as my transmitted copy. At the receiver end, I have the same matrix of size 32x25 but i have only the training bit values at the respective fixed positions and all the other elements as zeros. When i correlate the transmitted copy with the received copy, i must get a peak when the training bit values coincide. But i am getting a lot of false peaks may be because my training bit value is not maximum. I have also tried time averaging but i'm still not able to notice anything significant. Could any of you help me find what i'm doing wrong? Any help would be greatly appreciated. Start a New Thread