In wireless sensor networks, data fusion is widely used to merge data obtained from heterogeneous sensors. Since sensors provide uncertain information and containing redundant and irrelevant features, it is imperative to avoid undesired features and select the subset of relevant features for use in data fusion. In this paper, a novel data fusion scheme is proposed based on Pearson correlation and Dempster-Shafer Theory to properly quantify the correlation of features and uncertainty of sensors data. The proposed FCDS (Feature correlation-based fusion using Dempster-Shafer) scheme utilizes the correlated features and available information in the body of evidence (BoE). The BoE obtained from the correlated features of sensor data are then processed using belief entropy to fused the evidence and make a decision. Extensive computer simulation shows that the proposed scheme significantly outperforms the existing schemes for multisensor data fusion in terms of fusion accuracy, degree of uncertainty, and energy efficiency.