DIGITAL COMPATIBLE RECEIVER USING PRIORITIZED SEQUENCE ESTIMATION BY NERVE NETWORK

摘要

PURPOSE: To attain the high speed of transmission by providing a neural network in which a sampling signal outputted by a matching filter is received, and an output signal being the estimation of a signal transmitted at first through a transmission channel is applied from a receiver. CONSTITUTION: A signal z(t) obtained when the output of a matching filter 56 having an input connected with a timely changing transmission channel 58 is sampling-processed in an interval nT by a switch 60 is generated. A sampling-processed signal sequence Zn} is inputted to a shift register 62 in an M stage, and the M pieces of observation of the symbol sequence Zn} are temporarily stored in shift registers R1 -RM at the time operation. All amplifier inputs ui (i is 1 to M) are set as 0 values at first, and switches Swi (i is 1 to M) are simultaneously closed in a time sufficient for stabilizing the output of a neural network 54 to a stable state. After the neural network 54 is stabilized, an estimated sequence is read by the output of a threshold device 74. Thus, the high speed of a processing can be attained.