A neural network implementation suitable for correlation decoding of some block codes

Algorithms for the decoding and error correction of codes such as the Hamming and Golay codes are well established. They are generally implemented using either general purpose or dedicated microprocessor systems. Designers of such systems continually seek faster devices that retain flexibility and ease of use. As a consequence some researchers have proposed theoretical systems based on artificial neural network (ANN) architectures for the implementation of decoding algorithms. The authors present a practical proposal for the realisation of an ideal correlation decoder based on an analogue binary associative memory network (BAMNET) ANN prototype, which has been developed and manufactured in a 2.4 μm CMOS technology for rapid pattern recognition