MUMBAI, India, Aug. 10 -- Intellectual Property India has published a patent application (202641089085 A) filed by Jeppiaar Institute Of Technologyjit on July 22, 2026, for Novel System, Design And Method Of A Lightweight And Explainable Artifical Intelligence Framework For Geo Forensic Crime Investigation.
Inventors include Mr. R. Elango; Jesimathasneem S; Jayapriya D; Harini S; Vijayasankar G; and Karthikeyan M.
The application for the patent was published on July 31, 2026, under issue no. 31/2026.
Abstract: 6. ABSTRACT OF THE INVENTION The patent disclosure covers Novel System, Design and Method of A Lightweight and Explainable Artificial Intelligence Framework for Geo Forensic Crime Investigation. This Patent relates to a lightweight and explainable artificial intelligence ()CAI) based framework for analysing geological trace evidence, such as soil, sediment, airborne dust, mineral particles and pollen, recovered from a crime scene, suspect, vehicle or physical object, for the purpose of geographic origin determination during criminal investigation. Existing geoforensic analysis relies heavily on manual microscopy, elemental assay and subjective expert interpretation, which is slow, resource-intensive and difficult to defend under judicial scrutiny in the absence of a transparent reasoning trail. Deep- learning based approaches, though comparatively accurate, function as computationally heavy black-box models unsuitable for portable forensic instrumentation and incapable of producing an interpretable rationale acceptable in a court of law. This Patent overcomes these limitations through a compressed, resource-efficient ensemble of shallow learning models coupled with an integrated feature-attribution based explainability engine, deployable on low-power, portable or embedded forensic hardware. The framework acquires multi-modal geological signatures of a questioned sample using portable spectroscopic, elemental and morphological sensing instruments, extracts a normalised feature vector, classifies the probable region of origin by correlating the sample against a reference geo-signature database, and automatically generates a human-readable explanation report describing the specific physical and chemical features supporting the prediction along with a numerical confidence score. This Patent thereby provides investigators, forensic laboratories and courts with a fast, low-cost, transparent and legally defensible tool for establishing the geographic provenance of trace evidence, further embodiments of which incorporate tamper-evident chain-of-custody logging, privacy-preserving collaborative model updates across laboratories, and calibrated uncertainty reporting.
Disclaimer: Curated by HT Syndication.
