A new nonuniform frequency sampling technique for multidimensional FIR filter design

Summary form only given. A nonuniform frequency sampling method for designing multidimensional finite impulse response (FIR) filters has been derived. The derivation makes use of the fact that a multidimensional polynomial is uniquely reconstructible from its samples on irreducible curves, provided there are enough samples on each curve. The technique entails choosing an appropriate number of samples of the desired frequency response on an appropriate number of straight lines with different rational slopes in such a way that the resulting filter design problem has a unique solution. The major advantage of the technique over a window and transformation methods is its simplicity, and its advantage over existing frequency sampling methods is its increased flexibility. Efficient algorithms for determining the filter coefficients have been developed