Context-dependent trust in data fusion within the possibility theory

In possibility measures theory, several fusion rules exist. Generally these rules can not handle rigorously contradictory data and the reliability of the results appears difficult to evaluate. This paper proposes a method where a vector reflecting the reliabilities of data sources is used. A new fusion rule is then introduced, appearing in a larger frame of application, and the reliability of the result can be evaluated. This approach enables the handling of context dependent parameters in the estimation step.