Detection of long-duration narrowband processes

Abstract

Detecting long, weak signals that are narrowband but of unknown frequency structure is an important signal processing challenge, with many applications in remote sensing and process monitoring. An ad hoc scheme is developed. Its stages include the discrete Fourier transform (DFT), a multiresolution decomposition in the frequency domain, and a generalized likelihood ratio test (GLRT). The computational load is light, and the performance is remarkably good. This is so not just in the original narrowband situation, but also, due to an inherent adaptivity to the data, in the detection of signals that are relatively broadband in nature. Generalizations are given to constant false alarm rate (CFAR) operation in both prewhitened and unwhitened cases, and to the detection of multiband signals. As regards the last, it is discovered that there is little loss from overestimating the number of bands.

Periodic signals detection and deinterleaving using Hidden Markov Models

“…This model assumes that the frequency behaves according to a random walk in the time-frequency domain, so that from time t to t + 1, the frequency "moves" in its vicinity according to some given transition rules. HMM based algorithms have proved to be effective under low SNR conditions for various applications, such as underwater acoustics [4] [5], in astrophysics [6] to track drifting frequency peaks, or in radar [8]. However, in [3], a single frequency track is considered.…”