Improving the Identification of Textile Fibers Using a New Raman–Deep Learning System

Using new technology to accelerate forensic evidence processing is seen as a critical variable in quickening criminal investigations. In a recent study, H. Sun of Shaanxi Police College tested a new automated system combining Raman spectroscopy with deep learning to trace textile fibers. According to findings published in the journal Advanced Electromagnetics, the new automated system identified 98.7% of textile fibers accurately, showcasing its potential for use in criminal investigations.¹

What are microfibers?

Microfibers are thread-like fragments that come from clothing and other textiles.^1,2 They are routinely found at crime scenes and in the environment.² Although microfibers are among the most common types of trace evidence found at crime scenes, identifying their composition is slow and dependent on analyst expertise.¹

Sun sought to address this issue by pairing Raman spectroscopy, which reveals the molecular structure of a material through its interaction with laser light, with an automated classification model.¹

What did the researchers do in their study?

Building the system required Sun to compile a data set of 8,500 Raman spectra that spanned 15 common and blended fiber types.¹ Sun developed a preprocessing pipeline that corrects baseline drift, smooths and denoises signals, and normalizes spectral intensity before feeding the data into a one-dimensional (1D) convolutional neural network trained to recognize the chemical signature of each fiber type.¹

When tested on an independent set of spectra not used in training, the model correctly classified fiber types 98.7% of the time, according to the study.¹

What do the results suggest?

The 98.7% accuracy rate indicates that combining Raman spectroscopy with machine learning can reliably distinguish between textile materials with similar visual or physical properties.¹ These visual and physical similarities are often difficult to detect through conventional microscopy or manual spectral interpretation alone.¹ The capabilities of Raman spectroscopy allow researchers to learn more about these fibers.

All of this is beneficial for forensic laboratories. Acquiring more consistent fiber identification efficiently helps reduce case backlogs and limit variability introduced by manual analysis.¹ The approach may also have applications beyond criminal investigations, including textile quality control and recycling sorting, where rapid, non-destructive material identification is similarly valuable.¹

References

1. Sun, H. Research on Rapid Identification Technology of Microfiber Evidence in Criminal Cases Based on Raman Spectroscopy and Deep Learning. Adv. Electromag. 2026, 15 (3), 6857–6861. DOI: 10.7716/aem.v15i3.3763
2. Brandon, A.; Baechler, B. What is the Difference Between Microplastics and Microfibers? Ocean Conservancy, 2024. https://oceanconservancy.org/blog/2024/08/27/difference-between-microplastics-microfibers/ (accessed August 21, 2026).