An efficient method for generating unaliased samples of certain signals

In simulating a continuous-time system being driven by a continuous-time signal f(t), we may wish to generate a discrete-time signal whose spectrum agrees with that of f(t) for frequencies up to the Nyquist frequency. One important example is provided by the simulation of the vocal tract driven by a triangularly shaped glottal pulse, although a similar problem arises whenever we simulate a known continuous-time input to a system. We define classes of signals, called H-synchronous signals, which, after sampling, can be "de-aliased" with a fixed digital filter depending on H. The McClellan-Parks-Rabiner algorithm is used to find FIR designs for such de-aliasing filters for the important special cases of piecewise-constant and piecewise-linear signals.