Abstarct: Data fusion refers to collecting data and knowledge about an event from various sources and then integrating them to achieve more desirable information for the event. The top level of data fusion is called as inference. An inference algorithm accepts ‘decision’ as input and results conclusion and fusion of its input as output. This thesis aims to study on Dempster-Shafer theory as a robust inference algorithm. This algorithm is derived from Bayesian theory however there are some differences in their problem solving consideration. This thesis proposes an efficient approach baxsed on optimization algorithm (e.g. Genetic Algorithm) to improve inference of Dempster-Shafer theory. Using inference can improve performance of tools such as Wireless Multimedia Sensor Network. Overlapping camera network is one of the mostly used tools in Multimedia Wireless Sensor. The reliability of these networks can be improved using inference algorithms. Health smart house is a research plan to support old and imperfective people. Fall detection is one of the challenging subjects in these houses. The case study part of this thesis is an attempt to improve the inference reliability of overlapping camera network for fall detection. Finally, the cameras with highest reliability factor are selected for fall detection using the proposed method.