Problem of statistical hypothesis testing arises in the context of digital communication and detection systems. These systems employ several receivers/sensors to observe a source of message information/decisions produces a bit 0 or 1. The observed decisions are reported to a data fusion processor which is responsible for combining the received decisions from the various sensors into a final global decision. This approach is called hard-decision approach. The alternative approach for decision fusion is the soft decision approach where each sensor report a measure of uncertainty or confidence value for each hypothesis to the data fusion processor. The soft decision approach has the advantage of better performance over a comparable hard-decision approach. This paper proposes a new soft decision approach in statistical hypothesis testing problems with data fusion using Neyman Pearson criterion. The performance of the proposed approach is evaluated using Monte Carlo simulations and compared to that of a hard-decision approach. The proposed approach is simple and shows better performance.