In this paper, the Dirac symbol is used to represent a neural network,
and a discrete Hamilton neural network model with a 16-state (+/- 1 /- i +/- j +/- k) neuron has been presented. By using signal-to-noise
theory and computer numerical simulation, the stability, the storage c
apacity and the error correction ability of the model are analysed. Th
e storage capacity ratio of the presented model equals that of the Hop
field model, This 16-state neural network can be applied to recognize
16-level gray or color patterns.