MUMBAI, India, Jan. 9 -- Intellectual Property India has published a patent application (202541133076 A) filed by Ranjith Pulyala; Prof. T Kishore Kumar; and Dr. S. Siva Priyanka, Warangal, Telangana, on Dec. 29, 2025, for 'multi-scale depth wise separable progressive attentional knowledge distillation transfer driven for epileptic seizure detection on eeg signals.'
Inventor(s) include Ranjith Pulyala; Prof. T Kishore Kumar; and Dr. S. Siva Priyanka.
The application for the patent was published on Jan. 9, under issue no. 02/2026.
According to the abstract released by the Intellectual Property India: "Multi-scale Depth wise separable progressive attentional knowledge distillation transfer driven for Epileptic seizure detection on EEG signals Abstract Epilepsy, a neurological disorder caused by abnormal brain electrical activity, is difficult to diagnose and treat. This paper aims to create a KD-based deep-learning algorithm for detecting epileptic seizures. There are two primary stages: signal pre-processing and feature extraction. Initially, the focus is on denoising EEG signals to reduce errors and extract relevant information. The algorithm examines denoising methods capable of effectively removing noise and retaining essential data. The raw input data is then pre-processed to remove unwanted noises with the Upgraded Kalman Filtering (Up-KalFil) method. Wavelet Packet Decomposition (WPD) approach decomposes the original EEG signal into an approximate rhythm wave. Feature extraction plays a critical role in compressing the dimensions of the original data and capturing information reflective of EEG activity states. Time domain, frequency domain, and nonlinear dynamic features are extracted to facilitate this process. The study presents the Hierarchical Depth-wise Separable Triple Progressive Attentional Knowledge Distillation (HDSTPAKD) model and investigates the application of KD-based deep learning for detection. This model integrates teacher and student models to transfer knowledge at various depths of deep representation learning. Knowledge transfer is facilitated by a triple-attentional knowledge transfer model that combines the activation, spatial, and channel attentional frameworks. Furthermore, both the teacher and student models integrate progressive attentional modules to improve feature extraction for accurate epileptic seizure detection. Subsequently, the proposed method classifies whether the signal is healthy or affected by elliptic seizure. In the evaluation, the model attains 98.82% in the elliptic seizure recognition dataset and 99.92% in BONN dataset."
Disclaimer: Curated by HT Syndication.
