Evaluation of Sensor Measurement and Determination of Basic Probability Assignment for Context Inference

In a wireless sensor network, measures to rationally evaluate the signal detected and reported by the sensor are needed. To infer the context based on the detected signal, determination of basic probability assignment for multi-sensor data fusion using Dempster-Shafer Theory is proposed. If there is insufficient advance information in a wireless sensor network using different type of sensors, it is helpful to use the difference of the sensed signal values. First is to compare the measurement with the previous measurements and second is to compare the measurement with the filtering reference which is the basis of sensor report. In evaluation through comparison, one can compare the averages in a selected time band, but it may be more rational to compare the variance in evaluation of signals. We applied such evaluation to determination of basic probability assignment function to utilize it for context inference using DST.