Choosing a Power Spectral Density Threshold for Signal Denoising

Summary

The document asks whether a power spectral density threshold can be selected dynamically when using the fast Fourier transform to denoise a signal. It compares this desired selection with retaining enough singular value decomposition components to capture a chosen share of variance. The question specifically seeks a thresholding method that does not depend on visually inspecting a power spectrum.

No threshold rule, algorithm, or experiment is provided, so the text does not establish how to choose a cutoff or whether a particular method works. The underlying challenge is that Fourier power is distributed across frequency bins, while a variance-retention criterion for singular values does not automatically define an equivalent frequency-domain rule. Any practical threshold would depend on assumptions about signal structure and noise, which the document does not specify. The topic is a general signal-processing question rather than a trading strategy or market analysis.

Key ideas

• The question concerns frequency-domain denoising using the fast Fourier transform.
• It compares power spectral density selection with a variance-retention rule for singular values.
• It seeks a threshold that can be computed without visual inspection.
• No method, validation results, or assumptions about noise are supplied.

Tags

# Dynamic PSD threshold for FFT


# Dynamic PSD threshold for FFT












I am using signal denoising as explained by Steve Brunton, who explains that FFT is data data-driven SVD.

I can select important components by selecting singular values that capture 90% of the variance in SVD.

Can I dynamically compute the threshold (without any PSD visualization) of PSD when I decompose a signal using FFT?

Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.