Periodogram and parametric methods estimate power spectra.
Estimate power spectral density from finite data: periodogram (biased, noisy), Welch's method (averaged overlapping windowed segments), Blackman-Tukey, parametric AR/MUSIC/ESPRIT with high resolution. Trade-off between variance and resolution.
Welch PSD estimates vibration spectra in condition monitoring.
Spectral estimation finds the power in each frequency band from noisy data, which is how vibration and speech analysers work.
Periodogram and parametric methods estimate power spectra.
Spectral estimation finds the power in each frequency band from noisy data, which is how vibration and speech analysers work.
Averaging several segments reduces the variance of the estimate, so the spectrum looks smoother and more reliable.
The squared magnitude of the DFT of a record, which is simple but has high variance without averaging or windowing.
Typical applications include estimating the vibration spectrum of a machine from a noisy record, finding the dominant harmonics in a power line measurement and comparing two sensor records through their power spectra.
Study "Adaptive filters" first (topic 36 of 101 in Signals, Control & Comms). It covers the groundwork this topic builds on.
Window and overlap choice. Bias and variance. Model order selection.
Welch PSD estimates vibration spectra in condition monitoring.
The next topic is "Wavelets", which builds directly on this one.
It is classed as Intermediate in Signals, Control & Comms (topic 37 of 101). It is an intermediate topic. Work through the beginner topics in Signals, Control & Comms first.
Want to practise? Open the interactive course for quick checks, quizzes, simulations and progress tracking.