Walking assistance system for sight impaired people based on a multimodal information transformation technique

A multimodal information transformation system is proposed in this paper to provide sight impaired people with scene information of walking areas and obstacles. The scene information is first acquired as images using a single CCD camera, and then image information is transformed into voice information so that sight impaired people can obtain by hearing instead of seeing. During scene image processing period, the walking area is extracted by the vanishing point and boundary of a sidewalk on edge image using a chain-code line detection algorithm. And obstacles are detected by applying Gabor filter to the vertical lines extracted from the walking area on image. Later, based on the above image information, voice information is constructed in the form of pre-defined sentences by combining a set of template words that represent walking areas, obstacles, directions and distances. With the help of voice instructions provided by this multi-modal information transformation system, sight impaired people are able to reach their destinations safely and conveniently. The proposed algorithm has been implemented and tested in both indoor and outdoor environment, and its superiority in providing exact walking instructions has been verified.