A Bayesian network (BN) decoder for direct-conversion microwave and millimeter-wave six-port receivers (SPRS)

In this paper learning algorithms are used io decode QPSK data in a direct- conversion hardware (QPSK-unimode) microwave / millimeter wave receiver usi ng an application-specific six-port module. Two different algorithms, the K -means and Bayesian network, are considered for operation of the decoder to recover IQ data from QPSK modulated signals. Bit-error-rate (BER) results versus noise level (E-b /N-0) are presented after a brief data-collecting p eriod involving fewer than 500 data points. Results of the Bayesian network for Eb / Ng = -3 dB and Eb / Ng = - 9 dB are presented, using the same dat a-gathering period. (C) 2001 John Wiley & Sons, Inc.