Finite length discrete matched filters (Q1078146)

The matched filter (MF) is a usually occurring element in many communication systems. It is known that the MF gives the maximum signal- to-noise ratio (SNR) for a given signal corrupted by white noise. For continuous-time signals the realization of the MF can lead to noncausal solutions. For finite-length discrete-time signals this situation is of no interest. Also, for non-white noise the SNR can be maximized by a proper choice of the signal. A procedure to obtain optimal versions of MFs is presented. First, upper and lower bounds for the SNR are obtaied. Because its iterative structure in finding the MF the utility of the Levinson algorithm is limited when a suitable filter length is to be found. So, looser, but easier to compute, bounds are derived in the paper. The parameter required for the application of the proposed procedure is the so-called equivalent rectangular time duration \(\Delta\) T of the noise autocovariance. Some numerical examples show the validity of the procedure.