Asritha Madadi

Built a Smart Video Navigation System using Python and Streamlit that automatically transcribes any video using OpenAI Whisper, enables meaning-based semantic search through the transcript using SentenceTransformer embeddings, and auto-generates adaptive multiple choice quizzes using TF-IDF , all without requiring any pre-existing captions. The system features 7 modules including auto chapter detection, engagement analytics, speech density visualisation, and timestamp-based navigation, processing both uploaded video files and YouTube URLs in real time.

This project presents a cognitive state detection system that predicts mental states Positive, Stressed, and Fatigued from EEG signals using the DEAP dataset. Raw 1D EEG signals are converted into 2D time-frequency representations, enabling deep learning models to capture both temporal and frequency patterns simultaneously. A multi-model ensemble combining SVM, Random Forest, LDA, CNN, LSTM, and CNN-LSTM achieves 70–80% classification accuracy. The system is specifically designed to remain reliable under real-world degraded conditions such as noisy signals, missing channels, and distorted representations, while a confidence scoring mechanism flags uncertain predictions rather than producing silent incorrect outputs.