This paper presents a method for the non-contacting moisture prediction of small objects using m-sequences. The objects under test are placed in a free-space transmission path. Their UWB impulse responses are captured using an m-sequence method which provides high signal quality by means of ultra low jitter and noise. The size of the objects is in the range of the coherence length and smaller than the footprints of the antennas. The impulse responses depend on the moisture content (dielectric properties) and the geometry. Multivariate calibration is applied in order separate these influences and to develop a statistical model for the moisture content. For the prediction of moisture content an accuracy of 0.55% was achieved while it was varied in a range between 4.3% and 23.4%.