| dc.description.abstract |
People with motor disabilities are increasing day by day in the world. Many disabled and elderly persons need to use wheelchair for performing regular movements and activities. However, there are a lot of issues and challenges in using the conventional wheelchairs. Safe navigation, independent mobility and low cost are the key issues for wheelchair users. It becomes challenging to run wheelchair manually, specially for those who suffer from dexterity-inhibiting disorders and mobility impairments. Nowadays, many smart wheelchair systems have been proposed and developed using modern technologies consisting of several features. Nonetheless, these types of systems are not so suitable and flexible for users having dexterity problems who face difficulties with the movement of their hands or other organs, which those systems require. Also, the recent hand gesture-controlled wheelchair system doesn’t perform well with good accuracy at outdoor environment under heavy sunlight or if the background matches with skin color. Hence, it is still now a challenging issue to build an efficient gesture recognition technique for wheelchair control that will consider users’ flexibility, high accuracy, and environment’s lighting condition independency. Keeping these issues in mind and aiming to mitigate, this thesis presents the design of an efficient hand gesture recognition algorithm for a smart wheelchair control technique, where a mathematical model is developed. It utilizes the positions of the hand landmarks and the distances among the landmarks along with some threshold ranges. Different hand gestures are recognized when different logical conditions on distances and thresholds are satisfied. The proposed method emphasizes users’ flexibility more by requiring less hand movement so that disabled persons can easily operate it with least physical effort. Users can control the proposed wheelchair by using only movement of the fingers. This method works for hands of different sizes, on any background in indoor or outdoor lighting condition. Therefore, it will resolve the issues of other proposed techniques. The proposed model has been implemented and validated in a real powered wheelchair and we showed that it’s working well. We have compared the results with those of other researchers which demonstrates that it outperforms other existing methods in terms of success rate with excellent Precision, Recall and F1 Score indicating higher relevant results with higher accuracy. |
en_US |