MUMBAI, India, July 24 -- Intellectual Property India has published a patent application (202641061835 A) filed by International Institute Of Information Technology, Bangalore on May 15, 2026, for Autoencoder Based Arrhythmia Detection System For Resource Constrained Embedded Devices.
Inventors include Nagarajan S; and Kurian Polachan.
The application for the patent was published on July 17, 2026, under issue no. 29/2026.
Abstract: ABSTRACT AUTOENCODER-BASED ARRHYTHMIA DETECTION SYSTEM FOR RESOURCE-CONSTRAINED EMBEDDED DEVICES An autoencoder-based arrhythmia detection system configured for real-time inference on resource- constrained embedded devices is disclosed. The system comprises an electrocardiogram signal acquisition module, a segmentation module configured to segment electrocardiogram signals into heartbeat-level segments based on R-to-R intervals, and a lightweight autoencoder-based machine learning model trained on normal electrocardiogram morphology. The autoencoder compresses each electrocardiogram segment into a lower-dimensional latent representation and reconstructs the segment, wherein reconstruction error provides an interpretable measure for detecting cardiac abnormalities. The system performs arrhythmia detection entirely on-device using low-cost microcontroller platforms without dependence on smartphones, gateways, cloud servers, or external AI engines. In some aspects, the system further comprises a weighted autoencoder mechanism that assigns higher reconstruction importance to clinically significant electrocardiogram waveform regions and a localized reconstruction-error analysis framework that evaluates reconstruction deviations across segmented electrocardiogram regions. In some aspects, a TinyML-based multiclass arrhythmia classification model classifies a plurality of heartbeat categories based on R-to-R segmented electrocardiogram signals.
Disclaimer: Curated by HT Syndication.
