Design and development of a smart blind walking stick using machine learning

Abstract Visually impaired people are often subjugated under extreme circumstances even in their day-to-day life. The daily requirements of a common man appear to be an impediment in their routine life. Simplest of tasks like walking, eating, bathing, conversing and even eating is of utmost difficulty to them. Moreover, with such difficulties their only way-out seems to be dependency on the privileged lot, which further diminishes their confidence in themselves and gradually makes them even more dependent. The conventional devices that are used by visually impaired people include basic walking sticks which fail at the job in hand by not providing adequate stabilisation on rough surfaces and misguiding the users into unfavourable conditions. There is no way for the person to know what the object in front of them is without hitting it with the stick, which could also lead to accidents. To solve these problems, a smart walking stick is developed which not only recognises the object in front of it using Machine Learning (ML) models, but also gives a voice output to alert its user about the particular object thereby limiting the chance of any and all accidents. The concept is realised in hardware and integrated to the walking stick. This helps in stabilisation of phone and to produce better results in object identification. Further an application is developed to alert the user by converting the obtained image into a voice messages.