Modeling of chaotic dynamics in the biological system and application to speech recognition
Koji Shimoide, Walter J. Freeman · 2003
The significant capabilities of a dynamic neural model system, the KIII model, are illustrated for speech recognition. The neural network system presented was derived from the anatomical features of the olfactory system and consists of three layers of primary and secondary nerve units composed of basic oscillators: OB, AON and PC. Chaotic dynamics arises in the KIII model mainly due to differences in the interlayer time delay constants and dissonant characteristic frequencies in the feedback interactions between the OB and the lumped AON and PC layers. The authors discuss the generation of chaotic activity as basic dynamics of the system and the abrupt bifurcation between a basal and different classes of attractors.>