Wavelet Denoising
This tool applies wavelet-thresholding–based noise reduction to the input signal using a Julia implementation of discrete wavelet decomposition.
It automatically separates noise from meaningful signal components and reconstructs a cleaner version of the signal, preserving important transients and features while reducing broadband noise.
Various options can be set directly from the context menu:

Wavelet Type
Selects the wavelet family and filter shape used for decomposition.
- Haar – Fast, simple, blocky transitions; good for step-like signals.
- Daubechies4 / Daubechies8 – Compactly supported wavelets with increasing smoothness; suitable for general denoising.
- Symlet6 – Symmetric, smooth wavelet; often produces very natural-looking denoised signals.
- Coiflet4 – Balanced wavelet with good time–frequency localization.
- Battle4 – Smooth, spline-like wavelet; good for strongly oscillatory signals.
- Beylkin – Very smooth wavelet for detailed feature preservation.
Levels (3–6)
- Defines the number of wavelet decomposition levels.
- Higher levels remove noise on larger time scales but may also smooth out fine details.
- Lower levels focus on removing only high-frequency noise.
Threshold Method
Controls how the denoising threshold is computed.
- Universal – Classic Donoho universal threshold; strong noise suppression, risk of over-smoothing.
- Robust – Uses robust estimation of noise level; suitable for signals with non-Gaussian noise.
- Fixed3Sigma – Applies a fixed threshold of 3× estimated noise sigma; useful for predictable noise levels.
Shrinkage Mode
Determines how coefficients above or below the threshold are modified.
- Soft – Reduces coefficients smoothly toward zero; produces smoother, more natural results.
- Hard – Sets coefficients below threshold to zero while keeping others unchanged; preserves sharp features but may introduce artifacts.