Evaluation of different 3-channel dynamic compression schemes in a field-test with a wearable DSP prototype hearing aid

Appell, Jens-E. and Hohmann, V. and Kollmeier, Birger
An overview of the general structure of Zwicker's loudness model, (Zwicker, 1958; Zwicker, 1990) is given and modifications are described, with respect to several implementation details by different authors, (Moore, 1996; Launer, 1995:LPL:jens; Marzinzik, 1996; Marzinzik, 1996;, Blum, 1999; Appell, 2001). While all models use the same overall, structure (i.e., pre-filtering to account for outer and middle ear, transmission, separation into critical bands, construction of excitation, patterns, compression of the excitation into specific loudness,, summation of the specific loudness across frequency), and predict the, same loudness function for normal listeners at high input levels, they, differ with respect to the pre-filtering, the definition of the, frequency scale (BARK versus ERB scale), the reference level for the, excitation pattern, and the way in which specific loudness is computed, at low input levels. Other differences refer to how the model is, modified in order to predict loudness perception in hearing-impaired, listeners and how the model output can be transformed to predict, loudness on a different scale (e.g., categorical loudness scale). To, facilitate comparisons across the model versions, a common, implementation of the models has been implemented (DOS-Program). The, details of the models are reviewed and a comparison of the model, predictions for a limited set of input sounds is given.
01 / 2002