This paper describes a frequency-domain method of extracting the fundamental frequency of voiced speech which has been band-limited to 300 Hz to 3. 4 KHz. The method uses a linear auditory model into which non-linearity has been introduced. Two methods for introducing the non-linearity into the model are described. Harmonic product spectra are derived from the outputs of the linear and non-linear auditory models. Results show that the spectrum derived from the output of the non-linear auditory model is superior to that obtained from the output of the linear model. Keywords: auditory modelling, speech processing, pitch extraction.