Visually impaired people often cannot tell whether the person at the door is a familiar visitor or a stranger, which affects both convenience and personal safety. This paper presents EchoVision AI, a native Android application that acts as a smart safety companion. Familiar people are registered after phone OTP verification, and a FaceNet TensorFlow Lite model converts each face into a 512-dimensional embedding. During monitoring, Google ML Kit detects faces in live camera frames, and the system compares each live embedding with the stored ones using cosine similarity (threshold 0.75, three consecutive results). The result is announced by Text-to-Speech as a registered name or an unknown person. Visitors are logged and shown on a dashboard. When an unknown person is currently detected, pressing Volume Down three times within three seconds sends an SMS with a Google Maps location link to the trusted contacts. All records are kept in a local Room database.

Dr. B Meena Preethi, Lakshana K, "EchoVision AI – Intelligent Assistive Technology for Visual Safety" International Journal of Scientific Research and Engineering Development, V9(5): Page(1568-1572) September - October 2026. ISSN: 2581-7175. www.ijsred.com. Published by Scientific and Academic Research Publishing.