Machine learning-based wavelength detection system

Abstract A portable wavelength detection system has potential applications in various fields, including chemistry, biology, physics, and environmental sciences. Conventional systems rely on optical components, such as filters or slits, to separate light wavelengths, leading to complex measurement structures and challenges in miniaturization. Additionally, signals generated by light are susceptible to environmental factors and electrical interference, making traditional programming methods insufficient for accurate signal correction. To overcome these limitations, this study proposes an artificial intelligence-based filter-free wavelength sensor system that identifies wavelengths using the absorption coefficient of silicon. The proposed system consists of an analog circuit that applies signal conversion and noise reduction techniques for photocurrent from the filter-free wavelength sensor, and a microcontroller embedded with machine learning algorithms to process signals and calculate wavelengths in real-time. The system can detect central wavelengths in the 400–700 nm range, even with variations in light intensity, and corrects signals using embedded machine learning data. The system demonstrated the ability to identify wavelengths with a 1.74% error rate, even when light intensities varied between 0.20, 0.25, and 0.30 mW cm −2 . By leveraging the absorption coefficient of silicon and machine learning algorithms, a system has been developed that enables real-time wavelength detection regardless of changes in light intensity.