BiLSTM-FuseNet: A Deep Fusion Model for Denoising High-Noise Near-Infrared Spectra
Abstract
Near-infrared spectroscopy (NIRS) is widely used in food, pharmaceutical, and agricultural analyses but is highly susceptible to noise. To address this, we propose BiLSTM-FuseNet, a denoising framework that combines temporal modeling and explicit noise estimation. It uses stacked Bidirectional Long Short-Term Memory (BiLSTM) layers for global–local spectral learning and an MLP branch to predict and subtract noise. Evaluated on the Tablet and AnHui soil datasets with various synthetic noise types, the model outperformed the conventional methods, achieving an RMSE of 0.024 and R2 of 0.68 under mixed noise. The downstream regression improved the tablet weight prediction R2 from 0.079 to 0.218. These findings demonstrate the robustness of BiLSTM-FuseNet and its clear advantages for practical downstream NIR applications.
Fields of Study & Subjects
Spectroscopy and Chemometric Analyses
Soil Geostatistics and Mapping
Smart Agriculture and AI
Publication details
- Journal / source
- Electronics
- Publisher
- MDPI AG
- Volume
- 15
- Issue
- 1
- Pages
- 206–206
- Publication date
- 1 January 2026
- Language
- EN