PERCEVO: AI-powered Smart Cane with Dynamic Object Recognition to Support People with Disabilities

With the recent advancement of AI technologies, many traditional tools have been developed to improve the quality of life for people with different impairments. One such field is technology for the visually impaired, where current solutions focus more on navigation and less on helping the visually impaired to understand their environment. This paper presents Percevo, an AI-powered smart cane attachment that provides real-time descriptions of the environment. Equipped with a camera and AI algorithms, Percevo detects, recognizes, and provides information about nearby objects, including their distance from the user. Additionally, it incorporates an ultrasonic sensor for collision avoidance, and vibration sensors extend its utility to users with hearing impairments. A physical prototype of Percevo was developed and tested in an indoor environment, demonstrating its potential to enhance quality of life for users with visual and hearing impairments. The object detection algorithm achieved a success rate of 45% indoors and 25% outdoors, highlighting the challenges that occur due to rapidly changing environments, varying lighting conditions, and cluttered backgrounds. While the detection speed itself was adequate for indoor environments, further optimization is required for outdoor environments. Nonetheless, in comparison to existing tools, the integration of the system's components combining all forms of effective feedback from the surrounding environment, achieved an overall 98% indoor accuracy and 90% outdoor accuracy in detecting obstacles and measuring distances within a range of 1.3 meters. This work highlights the potential of AI technologies for assistive devices and encourages continued exploration with larger datasets to further refine the system.