A Real-Time Implementation Of A Dual Resonance Nonlinear Filterbank
Interpreting various forms of sound waves from the human perspective is associated with the auditory pathway. The auditory pathway consists of a complex network of neurobiological systems extending from the outer ear to the auditory cortex in the brain. From a signal processing standpoint, modeling various stages of the auditory pathway has evolved to a point where every stage can be modeled separately. However, analyzing the auditory pathway from a top-down viewpoint involves the integration of fragmented models of all stages. Furthermore, in order to study the signals generated in the upstream modules of such an ecosystem, it is imperative that computing at lowerstages do not cause a bottleneck especially with a large audio input. Therefore, a real-time implementation of the lower stages of the auditory pathway such as the outer, middle and inner ear consisting of the basilar membrane provides an ideal platform for higher levels of sound perception studies in the fields of computational neuroscience and digital signal processing. In this paper, a real-time implementation of outer and middle ear filters as well as a dual resonance nonlinear filterbank simulating the basilar membrane operations is described. The model is run on a general purpose desktop with Intel dual corePentium processor and is capable on simulating close to 180 channels concurrently.
Index Termsóreal-time, gammatone, DRNL, basilar membrane, cochlea.