Modelling speech processing and recognition in the auditory system with a three-stage architecture
One approach to the construction of an engineered system for hearing and efficient speech recognition is the modeling of the human auditory system. We applied this approach to our speech recognition tasks using a coupled modeling concept (Fig. 1) which should reproduce this system in a plausible way (Brückner et al. ). Starting with a model of signal processing by the cochlea (Kates ), our coupled modeling concept contains a lateral inhibitory neural network (LIN) system (Shamma ) performing filter operations by spatial processing of the speech evoked activity in the auditory nerve, and a structured formal neural network (Brückner et al. ) for learning and recognition of the spectral representations of the speech stimuli provided by the LIN.
Unable to display preview. Download preview PDF.
- 1.B. Brückner and W. Zander, “Neurobiological modeling and structured neural networks”, Proc. Inter. Conf. Artificial Neural Networks, Amsterdam, Sept. 13–16, 1993, pp. 43–46.Google Scholar
- 2.S. Shamma, “Spatial and Temporal Processing in Central Auditory Networks”, C. Koch and I. Segev (eds.): Methods in Neuronal Modeling, The MIT Press, Cambridge, Massachusetts, pp. 247–289, 1989.Google Scholar
- 3.B. Brückner, T. Wesarg and C. Blumenstein, “Improvements of the modified Hypermap Architecture for Speech Recognition”, Proc. Inter. Conf. Neural Networks, Perth, Australia, Nov.22-Dec.l, 1995, vol. 5, pp. 2891–2895.Google Scholar
- 4.J.M. Kates, “A time-domain digital cochlear model”, IEEE Transactions on Signal Processing, vol. 39, no. 12, pp. 2573–2592, December 1991.Google Scholar
- 5.Teuvo Kohonen, “The hypermap architecture”, In: T. Kohonen, K. Mäkisara, O. Simula, and J. Kangas, editors, Artificial Neural Networks, pp. 1357–1360, Helsinki, 1991. Elsevier Science Publishers.Google Scholar