Discrete signal reconstruction by simulated annealing

The problem of reconstructing an N-point discrete signal from a part of its known sample points and Fourier magnitude is converted into a combinatorial optimization problem. By using a simulated annealing technique, a signal-reconstruction algorithm is developed for effectively solving the problem. In comparison with the existing algorithms in the literature, the authors´ algorithm has the characteristics of fast speed and optimal recovery, and may be applied in various situations. The advantages become more obvious for those difficult problems of reconstructing signals with large N. Computer experiments with the algorithm were conducted, and the results show its advantages