In year 2023, approximately 2.2 billion people globally have near or distance visual impairment. Individuals with visual impairments often face challenges in navigating to destinations and avoiding obstacles. This research introduces ObstaSense, a wearable Electronic Travel Aid (ETA) designed to address these issues through obstacle detection and assistive navigation, enhanced with an integrated voice assistant. The obstacle avoidance system identifies objects including people, cars, bicycles, and motorcycles, providing users with instructions to help avoid obstacles. The RTN system eliminates the need to remember exact location names by recognizing keywords from the user’s speech via Bluetooth earbuds and guiding them in the correct direction. Additionally, ObstaSense integrates Gemini API to handle general queries during navigation and is capable of functioning in low-light conditions. The system was tested with non-visually impaired individuals walking outdoors with their eyes closed. Results from the pyramid chairs obstacle avoidance and navigation experiments demonstrated that ObstaSense successfully guided users from point A to point B while avoiding obstacles with 75% accuracy achieved. The solution is suitable for visually impaired individuals, the elderly, and people with cognitive or amnesia disabilities. Furthermore, it aligns with Sustainable Development Goals (SDGs) 3 and 10 by promoting healthy lives and well-being through independent navigation and reducing inequalities with its low-profile chest bag design.

Invention Technical Description

The ObstaSense is a wearable Electronic Travel Aid (ETA) that addresses these concerns by detecting obstacles and providing assistance navigation with an integrated voice assistant. The obstacle avoidance system employs a Raspberry Pi 5 with 16GB RAM, a Raspberry Pi Camera Module V3, and a TensorFlow Lite model to identify objects such as people, vehicles, bicycles, and motorbikes and delivers instructions to users via Bluetooth earbuds to avoid obstacles in their path. The system delivers instructions (e.g., turn left, right, straight, stop). In contrast, the Real Time Navigation (RTN) function uses a GNSS sensor, a compass sensor, vibration motors, and Plus Codes to provide highly precise navigation instructions that are supported by a voice assistant that includes Google Speech-To-Text (STT) and Text-to-Speech (TTS). The voice assistant is capable of processing over 40 input languages and producing output in over 50 languages. The RTN system eliminates the need to recall exact location names with home-finding navigation and destination searches within a 1 km radius by recognizing terms like “navigate to”, “nearest”, and “home” from the user's speech via Bluetooth earbuds and leading them in the correct direction Navigation routes are constantly updated while navigating, and haptic feedback notifies users when their heading differs more than 60 degrees from their destination. When GPS signals become unreliable indoors, the functionality incorporates a callback system that retrieves the most recent valid coordinates. ObstaSense also integrates the Gemini API to handle general inquiries during navigation and can function in low-light circumstances. The product weights 650g, was 3D-printed, and has tactile braille text for accessibility, metal push buttons, and infrared LEDs for enhanced functionality in low-light environments. The system was tested with non-visually impaired people walking outdoors with their eyes closed. The results of the pyramid chairs obstacle avoidance and navigation experiments showed that ObstaSense successfully instructed participants from point A to point B. Obstacle avoidance performance varied by row, with the first row having the best accuracy (100%), followed by the third row at 66.7% and the second row at 50%.

Demostration/ Presentation Video

Poster/ Broucher/ Invention Photo