Abstract: During reception of a training signal in a received signal R(n) an estimated impulse response value Hm(n) of an M-channel, and a tap coefficient G(n) of a linear filter 111 is calculated by an adaptive algorithm through the use of the received signal R(n) and the training signal b(n). For an information symbol of the received signal R(n), the received signal R(n) is subjected to linear filtering with the most recently calculated tap coefficient G(n), and the linear filtering output Z(n) and the most recently estimated impulse response value Hm(n) are used to calculate a soft decision value ?1. In the second and subsequent rounds of equalization, the likelihood b?(n) of a soft decision value ?2[b(n)] from a decoder is calculated, and a replica is generated by linear-filtering the likelihood b?(n) with an estimated impulse response value vector HL(n) obtained by approximating intersymbol interference with the current code b(n) to zero.

Abstract: During reception of a training signal in a received signal R(n) an estimated impulse response value Hm(n) of an M-channel channel, and a tap coefficient G(n) of a linear filter 111 is calculated by an adaptive algorithm through the use of the received signal R(n) and the training signal b(n). For an information symbol of the received signal R(n), the received signal R(n) is subjected to linear filtering with the most recently calculated tap coefficient G(n), and the linear filtering output Z(n) and the most recently estimated impulse response value Hm(n) are used to calculate a soft decision value &lgr;1. In the second and subsequent rounds of equalization, the likelihood b′(n) of a soft decision value &lgr;2[b(n)] from a decoder is calculated, and a replica is generated by linear-filtering the likelihood b′(n) with an estimated impulse response value vector HL(n) obtained by approximating intersymbol interference with the current code b(n) to zero.