Savitzky-Golay Differentiation

Savitzky-Golay differentiation is a numerical differentiation method based on the Savitzky-Golay smoothing filter, widely applied in signal processing and data analysis. The core concept involves simultaneous data smoothing and derivative calculation through local polynomial fitting, making it particularly suitable for processing discrete signals contaminated with noise. This method employs a sliding window to perform least-squares polynomial fitting on data points, using the analytical derivative of the fitted polynomial as the differential estimate for the window's center point. Compared to simple difference methods, it effectively suppresses the amplification effect of high-frequency noise. By adjusting the polynomial order and window size, users can balance smoothing performance with differentiation accuracy. Typical application scenarios include spectral analysis, biological signal processing (e.g., ECG/EEG), and sensor data calibration, where preserving authentic signal characteristics while extracting derivative information is crucial. Higher-order derivatives can be computed by appropriately modifying the same set of convolution coefficients used for first-order differentiation. Implementation typically involves precomputing convolution coefficients for specific polynomial orders and window sizes. Key algorithmic steps include: 1) Selecting optimal parameters based on signal characteristics, 2) Applying symmetric padding to handle boundary points, 3) Convolving the signal with precalculated differentiation coefficients. The scipy.signal.savgol_filter function in Python provides built-in support for Savitzky-Golay differentiation through its deriv parameter.