This paper describes the transputer implementation of two different front-end processors suitable for use in a speech recognition system. The two preprocessors are a physiologically-based cochlear model and a parallel filter bank based on the mel frequency scale. The structure of the cochlear model is a cascade of 128 digital filters. The parallel filter bank consists of 19 independent channels. The inherent structure of the preprocessors is reflected in the topology of the transputer configuration used to implement them. The two preprocessors developed, were tested, for their suitability for use as a front-end, to a two-layer Kohonen neural network based speech recognition system.