Soil moisture sensor systems need extensive calibration processes because of the strong dependence on soil properties. Therefore a suitable method for in-situ soil characterization is required. Using impedance spectroscopy, more information can be extracted from real and imaginary part of the complex admittance for several frequencies at the same moisture value. In this paper, an automatic measurement set-up was developed. It allows keeping similar conditions of the soil during measurements. Data consistency was investigated by means of the Kramers-Kronig relation after compensation of cable effects. For soil type characterization suitable signal features were extracted and subjected to principal component analysis (PCA) via singular value decomposition (SVD). The results are correlated with soil composition. The first two principal components were sufficient for classification of unknown soil type by the k-nearest neighbor method