MUMBAI, India, Oct. 5 -- Intellectual Property India has published a patent application (202641115566 A) filed by Vallurupalli Nageswara Rao Vignana Jyothi Institute Of Engineering And Technology on September 26, 2026, for A Deep Learning-Based Automated Sleep Stage Classification System Using Cnn-Lstm Network For Eeg Signal Analysis.
Inventor includes Mr. Bhupesh Deka.
The application for the patent was published on October 02, 2026, under issue no. 40/2026.
Abstract: ABSTRACT [0025] This project focused on the design, development, and evaluation of an automated system for sleep stage classification using EEG signals, leveraging a hybrid deep learning architecture combining Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks. The primary goal was to create an accurate and generalizable model capable of classifying 30-second EEG epochs into five sleep stages: Wake, N1, N2, N3, and REM, following AASM standards. Starting with the Sleep-EDF Expanded dataset from PhysioNet, we extracted and structured single-channel EEG data (Fpz-Cz) and aligned it with corresponding expert-labelled sleep stages. [0026] Preprocessing steps such as epoch segmentation, normalization, and label encoding ensured the data was clean and ready for training. Unlike traditional machine learning approaches that rely on handcrafted features, our deep learning pipeline learned directly from raw EEG signals, improving efficiency and reducing bias. The CNN layers effectively captured spatial patterns from the raw signals, while the LSTM layers modelled the temporal dependencies between consecutive sleep epochs, mimicking the sequential nature of human sleep cycles. This combination enabled the model to understand both the fine-grained signal characteristics and the broader contextual transitions across sleep stages. [0027] Model evaluation demonstrated strong performance in terms of accuracy, precision, and F1-score, especially when validated on subject-independent test data. Confusion matrices and metric visualizations helped identify areas where the model excelled and where further improvement may be needed. Furthermore, the model showed potential for real-time application, with low-latency predictions and compatibility with wearable EEG systems. proposed CNN-LSTM-based system presents a scalable and efficient solution for automated sleep stage classification. It not only reduces the dependency on manual scoring but also lays the foundation for real-time sleep monitoring systems in both clinical and consumer health domains. This project successfully demonstrates how deep learning can be applied to solve complex biomedical signal processing tasks with minimal feature engineering and high interpretability.
Disclaimer: Curated by HT Syndication.
