We can't find the internet
Attempting to reconnect
Something went wrong!
Hang in there while we get back on track
A Quantum Neural Network Algorithm for Microplastic Identification Using Raman Signatures Collected During Field Tests
Summary
Researchers developed a Quantum Neural Network Residual Autoencoder framework combining wavelet transform, compressed sensing, and a quantum bottleneck layer for robust Raman spectral reconstruction and microplastic identification from field test data. The method outperformed conventional approaches across RMSE, MAE, R², and Pearson correlation metrics for identifying PE, PP, and PET microplastics, demonstrating strong practical applicability for real-world spectral recovery.
Raman spectroscopy has emerged as a powerful tool for identifying microplastics due to non-destructive nature and less interference with water. However, its effectiveness for field tests is often hindered by spectral noise, baseline drift, and limited data availability. To address these challenges, this paper proposes a Quantum Neural Network Residual Autoencoder framework for robust Raman spectrum reconstruction and few-shot learning. The method integrates wavelet transform and compressed sensing for spectral denoising and smoothing, followed by a residual encoder with a quantum bottleneck layer that enhances feature representation in a low-dimensional latent space. Experimental results on multiple microplastic types (PE, PP, PET) demonstrate that the proposed method achieves superior performance across key metrics including RMSE, MAE, R2, and Pearson correlation, outperforming conventional methods in both accuracy and stability. The reconstructed spectra show strong alignment with standard references, validating the model's effectiveness in realworld spectral recovery tasks.