Application of B-splines in digital filtering

Summary form only given, as follows. In many digital signal processing applications, the input signal rate is higher than the output signal Nyquist rate. In the case of an FIR (finite-impulse-response) digital filter, it has been found that a B-spline representation of the filter response leads to minimal time and circuit complexities. The number of filter coefficients required in this implementation is reduced and corresponds to the lower sampling rate of the filter function. The aliasing usually associated with this lower sampling rate is substantially reduced by processing the input data with a prefilter which can be interpreted as an interpolator. The prefilter is implemented using a cascade of running average filters which requires no multipliers. The B-spline approach is suited to VLSI implementation and leads to substantial reductions in circuit complexity