Localization in the MTR for the Visually Impaired by Deep Learning

Abstract

Visually Impaired people (VIP) are the terms refer to the people who have a decreased ability on seeing things and in the worst case of complete or nearly complete vision loss. To help the VIP to access MTR stations easier, some aids are provided by the MTR stations. However, the aids provided by the MTR station such as the tactile guide path tile and tactile station layout map cannot fully support the needs of the VIP to navigate on the MTR platform. Considering this problem, we propose to use deep neural networks to significantly reduce the work-force burden of locating a user position to help VIP navigating at the MTR platform. Therefore, an android app is then developed to help VIP locating their position and the facilities at the MTR platform with the use of a deep learning approach to analyze the Wi-Fi signal in the MTR platform. The model was trained for ten thousand times to get the highest accuracy of the position. Several functions such as distance between the nearest elevator or nearest door entrance would be provided in the apps with voice feedback. The testing experiment would be taken in the real MTR platform to test the reliability of the app.