MUMBAI, India, June 26 -- Intellectual Property India has published a patent application (202641072624 A) filed by Amrita Vishwa Vidyapeetham on June 11, 2026, for A Wearable Seizure Prediction System Using Adaptive Cepstral Weighting Of Electrocardiogram Signals.
Inventors include Kumar, C. Santhosh; and Muralidharan, Pooja.
The application for the patent was published on June 19, 2026, under issue no. 25/2026.
Abstract: The present disclosure provides a wearable seizure prediction system (100) for predicting seizure onset in a subject (102) through adaptive cepstral weighting of electrocardiogram signals. The system (100) includes a skin-contact electrode unit (104) sensing ECG signals, a spectral whitening unit (106) applying inverse filtering using a linear predictive coder representing the interictal ECG (106-4) to extract a seizure-specific residual spectrum, a cepstral transformation unit (108) generating a cepstral coefficient vector (108-4) and a cepstral weighting unit (110) applying a gradient-optimized weight vector with a threshold operation (212) to produce a weighted cepstral coefficient vector (214) for classification by a classification unit (112) as an interictal state (112-6) or a preictal state (112-8) to trigger a timely alert. Unlike conventional approaches that fail to suppress dominant cardiac and sensor-induced signal components, the present disclosure achieves robust preictal state classification through subject-specific spectral whitening and adaptive cepstral weighting for wearable epilepsy monitoring.
Disclaimer: Curated by HT Syndication.
