112 Emergency Video Call Response Pipeline for Car Crashes Using Computer Vision and Natural Language Processing

This paper presents a comprehensive 112 emergency response pipeline that integrates multimodal technologies, such as automatic speech recognition, object detection, natural language processing, and blockchain, to enhance how critical incident information is gathered, analyzed, and acted upon. From real-world video inputs, speech content is transcribed using a speech-to-text algorithm, while computer vision models identify individuals involved. The transcribed dialogue is semantically processed to extract structured question-answer pairs, which are then evaluated to assess the medical urgency of the case. To enrich contextual awareness, the system cross-references vehicle ownership claims made in the transcript with a blockchain-based ledger of profiles linked by Vehicle Identification Number (VIN) identifiers. This verification step enables the retrieval of car owner name, age and critical medical background, such as insurance coverage, chronic conditions, and ongoing treatments. A lightweight web interface presents this information in an accessible format for first responders. The result is an intelligent, end-to-end system that prioritizes car accident emergency cases efficiently and empowers intervention teams with personalized insight, aiming to reduce response time and improve patient outcomes by classifying each emergency accordingly.