A Fast Real-Time Auditory-Nerve Model

We present the details of a real-time design and implementation of an auditory-nerve (AN) model on Motorola's DSP56367, a 150 MIPS audio digital signal processor (DSP). Sensory processing is highly nonlinear and time variant and so are the computational models that are derived to simulate these processes. Complex signal processing models have been derived to predict essential AN properties. However, their use is limited due to large processing time requirements. Our DSP solution makes fast real-time AN simulations possible and thus allows an experimentalist to change model parameters on the fly. This can speed up testing of several sound coding/decoding strategies hypothesized to occur in the human brain.