Recursive Least Squares adaptive algorithm Maximum Likelihood Sequence Estimation: RLS-MLSE

This paper derives the Recursive Least Squares adaptive algorithm ‘Maximum Likelihood Sequence Estimation’ (RLS-MLSE) which is suitable for fast time-varying frequency selective-fading channels. RLS-MLSE represents the channels by a combination of generating and measurement processes. Three kinds of channel representations are shown. Application of the maximum likelihood theory to the models results in an estimation process. The process simultaneously estimates the channel impulse response and channel state by a combination of the Kalman filter and MLSE. Modification of the Kalman filter to an extended RLS leads to RLS-MLSE. MLSE employs the Viterbi algorithm for efficient processing. Ensemble-average Inverse-matrix Lease Squares is proposed to simplify the extended RLS. A computer simulation for a two-path model of mobile radio channel shows that RLS-MLSE operates up to the maximum Doppler frequency of 160Hz in 40kb/s QPSK transmission. [In Japanese.]