NEURAL NETWORK CODE DIVISION MULTIPLE ACCESS SYSTEM

摘要

PURPOSE:To save the loss due to retraining caused by signal level fluctuation by controlling an information signal level into a set allowable signal level range and using a neural network trained so as to perform the demodulation of information bits within that range. CONSTITUTION:The received wave of an antenna 01 is amplified by an RF amplifying part 02 and converted to a base band by a frequency converting part 03, and the signal level range is controlled to the allowable signal level range by an automatic gain controller 04. Matched filter outputs 051 are inputted from matched filters 05 for respective users to a neural network part 08 and parallelly demodulated to information bits 07 considering the influences of mutual correlation and interference between the spread codes of respective users. The part 8 calculates the maximum and minimum levels of this range and is trained with the plural patterns of pseudo matched filter outputs provided by combining all the user signal levels and information bits inside a system and after the training is finished, the part 8 demodulates the information bits while inputting the matched filter output patterns so as to unnecessitate relearning at the time of signal range fluctuation.