Summary In this thesis, a time-domain binaural auditory model is developed for the assessment of high quality coded audio signals. Traditional objective measurements of sound quality yield misleading results when applied to psychoacoustic based codecs. Whilst human subjective opinion must be the ultimate arbiter of perceived sound quality, an objective measurement is sought that accurately predicts human perception. The auditory model described in this thesis is designed to match the processing within the human auditory system as closely as possible. The outer and middle ear transforms are simulated by linear filters; the basilar membrane response is simulated by a bank of amplitude dependent gammachirp filters, and the compressive action of the inner hair cells is simulated by an adaptation circuit. The whole model is calibrated using known human performance data. This monophonic model is used as the basis for a model of binaural hearing. Signals from the inner hair cell simulation are compared with those from the opposing ear, and an internal binaural image is formed. Changes in this image are shown to reflect audible changes in the binaural properties of the sound field. The binaural model predicts binaural masking and localisation data using a single detection criterion, which suggests that both phenomena may be due to the same internal process. Finally, the model is used for audio quality assessment. The monophonic model correctly predicts human perception for some coded audio extracts, but not all. A specific problem in the hair cell simulation is identified, and solutions to this problem are suggested. The binaural model accurately predicts human perception of spatial qualities. In particular, the model can detect changes to the stereo soundstage within real music stimuli. Published Papers