Classification Schemes for Step Sounds Based on Gammatone-filters

In this study the classification performance of 2 machine learning methods and 2 sound representations schemes are compared, having the focus on short impact like sounds: Footsteps have been classified according to the material of the floor and the shoe type. The gammatone auditory filter bank is a spectral analyser, that converts a given signal into a multi-channel simulation of the basilar membrane motion. Combinations of the gammatone auditory filter bank with the Hilbert transform and with the Meddis Inner Hair-cell model have been evaluated and compared in classification tasks. The experiments show that the gammatone based representation techniques yield in general promising results in the classification tasks of impact like everyday sounds. The support vector machines outperform the hidden Markov models, where both in general perform equal or better using the inner hair cell model representation.