A new, objective approach for optimal decision regarding signal detection is constructed. Deciding whether some data point is a signal changes the amount of missing information in the data via a feedback mechanism. By minimizing the amount of missing information in the objects one is interested in, one arrives at the optimal decision. This optimal decision is dictated by the data, in fact by the noise, or "background." In a computer experiment, with Gaussian noise, the scheme works. A minimum for the missing information exists, and an optimal decision can be made. A connection between the present approach and entropy on fuzzy sets is shown.