Precoding strategies for decentralized detection of unknown deterministic signals

We study different linear precoding strategies and their asymptotic detection performance for decentralized detection of unknown deterministic signals. Two precoding schemes are proposed: a random precoding scheme which generates its precoding vectors following a certain distribution, and a signassisted precoding scheme which incorporates the sign information of the signal components into the precoding design. The detection performance analysis shows that utilizing the knowledge of the signs of the signal components can considerably improve the detection performance. Also, our numerical results show that precoding-based detectors are more effective than the energy detector in detecting weak signals that are buried in noise.