This paper proposes a new calibrated adaptive frequency domain beamformer for speech enhancement. The beamformer is based on the principle of a soft constraint formed from calibration data, rather than precalculated from free-field assumptions. The benefit is that the real room acoustical properties will be taken into account. The proposed algorithm continuously estimates the spatial information for each frequency band, based on weighting of the received data. The update of the beamforming weights is done recursively where the initial precalculated correlation estimates of the speech constitute a soft constraint. The soft constraint secures the spatial-temporal passage of the desired source signal, without the need of any speech detection. The performance is evaluated in real world scenarios, both in a car-and a restaurant- environment. Interference and noise sup-pression of more than 15 dB is accomplished, while very small distortion is measured for the source signal.