Equalizer with a cost function taking into account noise energy

A device for processing signals in order to remove waveform distortion by using an equalizer, said device comprising:

• receiving means (5, 10, 15) for receiving and preprocessing an incoming signal (y);
• means (20) for estimating a channel impulse response (h) from said received signal (y);
• means (25) for estimating a noise energy (δ2) from said received signal (y);
• means (30) for calculating filter taps (f, b) by using said estimated channel impulse response (h) and said noise energy (δ2);
• feedforward filter means (35), the filter taps (f) of which are calculated by said means (30) for calculating filter taps;
• and
• an equalizer means (40) which comprises a feedback filter, the filter taps (b) of which are calculated by said means (30) for calculating filter taps, said received signal (y) being supplied to the equalizer means (40) via said feedforward filter means (35) to obtain a resulting signal (x) by a decision in said equalizer means;
• wherein said estimated channel impulse response is partitioned into at least two parts (bff, bdfe), said means (30) for calculating filter taps using one of said parts (bdfe) of said channel impulse response for calculating filter taps by using a weight function having the form of
• ƒ(h,δ2,A)bdfeHbdfe,
• whereby a cost function of said equalizer is formed as
• J=E|xi−yf−xb|2+ƒ(h,δ2,A)bdfeHbdfe,
• wherein J is a cost function result, E indicates an expectation value, xi is a data symbol of the signal currently to be estimated, y is a vector containing the received signal, f is a column vector containing taps of said feedforward filter, x is a vector containing data symbols decided on previously to xi, b is a column vector containing taps of said feedback filter, ƒ(h,δ2,A) is a function estimating an expected squared error vector with A as a user definable constant and the operator H indicates a complex-conjugate transpose of a matrix or a vector.