| Modern communication systems such as intelligent networks (IN) and integrated services digital networks (ISDN) require real time fault diagnosis and error free transmission. These demands can be met efficiently by the use of parallel expert systems and artificial neural networks. Parallel processing concepts in expert systems has been utilized to speed up the execution time in order to achieve a real time response. The proposed architecture has resolved the memory bottleneck and hypothesis coherence problems. The neural network approach has been applied for error control coding. Simulations have been conducted on Hamming codes, extended Hamming codes, constant weight codes, and Reed-Solomon (RS) codes. Systematic and non systematic sets of constant weight codes along with encoding/decoding methods have been proposed. The proposed codes are suitable in frequency hopped systems and optical ring networks. A sub set of RS codes utilizing natural language redundancy and a modified encoding/decoding method has been suggested. A Comparison of various learning algorithms in the error control coding problem is presented. The proposed techniques have resulted in improved signal to noise ratios, simple encoding/decoding methods, improvement in bandwidth, improved error probability, and data compression. |
