In this paper, we discuss methods for preprocessing and digital filter design with application to 3-D sound. It is clear that binaural filter design can be optimized by taking into account such properties of human hearing as non-uniform frequency resolution and a logarithmic magnitude scale. We show that by choosing proper spectral and time-domain preprocessing techniques the choice of filter design method becomes a relatively trivial task. The validity of preprocessing methods is proved using a binaural auditory model. Filter design techniques based on optimizing the Hankel, Least squares and Chebyshev norms will be discussed.