A Smart City Assistive Infrastructure for the Blind and Visually Impaired People: A Thin Client Concept

Abstract

The World Health Organization pointed out that over 285 million people worldwide suffer from loss of vision and blindness, and that the number could drop drastically in just a few years. About 90 % of the blind and visually impaired (B&VI) live at a low income that means these people cannot buy the expensive assistive devices for the spatial cognition. In this work, the new concept of the smart city assistive infrastructure with distributed server-client architecture is presented for the B&VI using the inclusive smart assistive component that interacts with other subsystems of the smart city (smart buildings, smart mobility, smart energy, etc.) via IoT protocols such as MQTT. The main constituents of thin client are as follows: Raspberry Pi 3 B board with camera for the objects detection / recognition, ultrasonic sensor(s) HC-SR04 for the obstacles identification on the short-range distance up to 5 m, GPS module for the global navigation, iBeacon Bluetooth low energy proximity sensing software, MQTT IoT protocol for the mutual communication of clients, Python multithread application, Raspbian OS. The thin client hardware is of affordable price USD 70. The objects detection and recognition are implemented on the thin clients via the Histogram of Oriented Gradients with Euclidean distance classifier (HOG+EDC). The design of the recognition models, the file hosting of the training images and the knowledge base with the recognition rules are done on server(s). The modified Viola-Jones fast face detector with the combination of features “eye” and “nose” is proposed to speed up the image processing, but its detection rate is not 100 %. Hence, it can be applied only with the subsequent recognition using HOG+EDC.