Soil moisture is a derived product of passive microwave remote sensors. But the soil moisture data from passive microwave sensors is often not correct for the places having large open water bodies. The present paper describes a technique of soil moisture determination from passive microwave brightness temperature values, by using ANN. Two ANNs are trained using either brightness temperature values or Polarization Index (PI) values as input with measured actual soil moisture values as target. The trained ANNs are then capable of estimating the soil moisture values with fair accuracy. The technique using PI and two brightness temperature values as input is found to be a better technique, due to less data requirement for both training and estimation.