Petar M. Djurić
Stony Brook University Stony Brook New York
In practice, spectrum estimation is usually a preliminary step of many signal processing problems. When we analyze temporal signals and when they satisfy certain statistical conditions, we apply temporal spectrum estimation that yields the distribution of the signal power over frequency. This is important for subsequent signal analysis, which may include developing parametric models that provide improved description of the data or building filters for suppressing noise and passing useful portions of the signal. For example, spectrum estimation of a temporal sequence with unknown spectral contents can reveal if the signal samples contain harmonic signals or not. If they do, one can propose a model for the data that can extract the number of signals in the data as well as their parameters.
Spectrum estimation is also very important in the analysis of signals obtained by an array of passive sensors and is referred to as spatial spectrum estimation. In general, L identical sensors may be deployed in a sensor field in an arbitrary or predefined way with their exact locations being known. The sensors are basically antennas, hydrophones, or seismometers, and they take measurements that can be electromagnetic, acoustic, or vibrational signals. The received signals are possibly ...