Central auditory model for spectral processing

In this paper, two central spectra processing models of the central auditory system were proposed to further process the spatio-temporal firing pattern of the auditory nerve as outputs of the periphery auditory model. The temporal processing model is to capitalize on the dominant periodicity in the firing pattern. The spatial processing model is based on the winner-take-all neural network mechanism, and is to detected edges of discontinuities in the firing patterns. Experiments of noisy speech analysis showed that the central auditory model can preserve prominent peaks at the formant resonances, while significantly reduce the output associated with the background noise. The noisy speech recognition experiments for both small scale classification tasks and large scale continuous speech recognition task showed that the model provides a robust speech representation, and outperformed all other representation.<
>