Evaluation of a physiological ear model for the simulation of nonlinear masking effects

Summary A physiological model of the human auditory system with the aim at generating masked thresh - olds for arbitrary complex sound signals is evaluated with respect to nonlinear mechanisms in masking. In (1, 2) it was shown that the ear model is able to rebuild the level-dependency of spectral and temporal masking patterns, the asymmetry of masking between noise and tone, and the dif ferent detection thresholds for amplitude and frequency modulation. The simulation results in this paper indicate that nonlinear mechanisms such as the additivity of masking, suppression, and distortion product detection are reproduced as well by the physiological ear model. Since the model represents a unified approach comprising the most important masking effects of complex sounds, it is suitable for applications in audio signal processing. Audio coding as well as audio quality assessment represent application examples for perceptual models which require the processing of complex audio signals. In contrast, most psychoacoustical measurements are carried out with simple signal configurations like combinations of narrow-band noises or pure tones. Therefore, the task of such perceptual models consists not only of rebuilding the required psychoacousti - cal results, but also of performing an extrapolation for more complex signals. A common goal in these applications is the generation of masked thresholds for the evaluation of the audibility of signal distortions. Currently used perceptual models for masked threshold generation widely ignore the highly nonlinear properties of sound processing in the auditory system. Therefore, the influence of the nonlinearity on masking can not be considered sufficiently. In this paper, the modelling of three nonlinear mechanisms af fecting masking will be investigated: (1) The perception of distortion products created in the cochlea. (2) The suppression of a test signal by a suppressor signal. (3) The additivity of masking in case of two superimposed maskers. All three mecha - nisms can influence the amount of masking but are not considered in currently applied perceptual models. A physiological ear model published in ( 1, 2) is used for the simulation of these nonlinear mechanisms. The main part of this model is a nonlinear cochlear model, which rebuilds the behavior of the outer hair cells (OHCs) as ''cochlear amplifiers' ' with saturation. It was already shown, that this physiological model rebuilds masking properties like the classic simultaneous masking patterns, post-masking, and the detec - tion of amplitude and frequency modulation. Due to the physiological approach the modelling results are valid independently from signal complexity. Chapter 2 contains a brief description of the physiological model. A more complete one can be found in (1, 2). The simulation results of the three nonlinear mechanisms are outlined in Chapter 3. Concluding remarks are summarized in Chapter 4.