Paper is devoted research and realisation of two-channel structural-holistic model 
for the speech patterns recognition. A theoretical basis of given bionic model 
is the concept about functional asymmetry of brain hemispheres as applied 
to process of oral speech perception. The architecture of the speech recognition 
system on the basis of interaction of structural and holistic recognition 
subsystems is offered. offered. The structural subsystem is realised in  basis 
of neural networks, and the holistic subsystem is realised in basis of fuzzy logic. 
Methods of the decisions coordination of these subsystems are considered.