Hemoglobin feature extraction based on wavelet packet-fuzzy shrinkage denoising model for improving the accuracy of whole blood NIRS quantitative analysis

• Novel Hybrid Approach : Combines WPT-FS denoising with WOA to enhance NIRS analysis of hemoglobin. • Innovative Denoising Strategy : Uses adaptive fuzzy shrinkage to reduce noise in NIRS data. • Optimized Feature Extraction : WOA reorganizes wavelet nodes to isolate key Hb spectral features. • Robust Quantitative Modeling : Achieves PLS model with RMSEP of 2.0409 and R 2 of 0.9746, outperforming conventional methods. • Clinical Potential : Enables rapid, non-destructive Hb quantification in blood samples for diagnostics. This study introduces a hybrid model that integrates wavelet packet-fuzzy shrinkage denoising (WPT-FS) with the whale optimization algorithm (WOA) to enhance the measurement accuracy of near-infrared spectroscopy (NIRS) for the quantitative analysis of hemoglobin (Hb). First, a novel threshold denoising function is proposed, which employs the fuzzy shrinkage of wavelet packet coefficients to significantly mitigate noise interference in NIRS data. Subsequently, the denoised wavelet packet nodes are optimized using the WOA to reorganize the nodes that correspond to the Hb information band. Finally, a partial least squares regression (PLS) model is developed for the reconfigured spectrum. Actual blood data analysis demonstrates that this method outperforms the traditional preprocessing techniques in effectively capturing Hb spectral features, and yields a root mean square error of prediction (RMSEP) of 2.0409 and a coefficient of determination ( R P 2 ) of 0.9746. These findings suggest that the proposed method substantially enhances the accuracy and precision of quantitative analyses of Hb in near-infrared spectra, offering a novel solution for blood spectroscopic analysis.