Effectiveness of Webcam Visible Spectrometer (WeViSpec) technology dissemination at Kolej Dato^ Onn Jaafar Living Lab as an Applied Research Facility Agus Setyo Budi, Lari Andres Sanjaya, and Mohd Amri Md Yunus
Faculty of Mathematics and Natural Sciences, Universitas Negeri Jakarta, 13220 Jakarta Timur, Jakarta, Indonesia
Abstract
Adulteration of honey with starch-based sweeteners is a global challenge that requires reliable yet affordable detection technology. This research disseminates the Webcam Visible Spectrometer (WeViSpec), a low-cost (USD 30.93) spectroscopy instrument that integrates off-the-shelf components with artificial intelligence, at the Kolej Dato^ Onn Jaafar (KDOJ) Living Lab, Universiti Teknologi Malaysia (UTM). This instrument is designed using a 1080p HD webcam sensor and a diffraction grating from old DVD disks with a single-channel inverted optical system. Data analysis was conducted thru differential absorbance feature engineering and the use of a Variational Autoencoder (VAE) model for dimensionality reduction, as well as a Bayesian Neural Network (BNN) for predicting adulterant concentrations. The test results show that WeViSpec has high wavelength accuracy with an average error of 2.93 nm and a spectral resolution of ~5 nm. The developed AI model successfully classified the types of sweeteners with 100% accuracy and predicted concentrations with high precision (R2=0.9978) and a Limit of Detection (LOD) of 6.66% v/v. The dissemination program involving 20 participants showed very high effectiveness with an average score of 4.68 out of 5.00, confirming the successful transfer of WeViSpec technology as a functional field screening tool in the living lab ecosystem.
Keywords: Please Just Try to Submit This Sample Abstract
