An Intraoral Camera for Supporting Assistive Devices

Abstract

Though much research has been conducted into improving the quality of life for individuals with paralysis, there is relatively limited work in this regard on the tongue-computer interface. The tongue is a muscular organ directly connected to the brain through a cranial nerve known as the hypoglossal nerve which is responsible for the motor functions of the tongue. Hence, tongue movements are not affected during spinal cord injuries, which is one of the major causes of paralysis. We propose in this paper a novel method for capturing inputs from the tongue using an intraoral camera. An endoscope camera is attached to a personalized mouthguard to capture images of the tongue. The intraoral environment is lit up with LEDs carefully placed at angles to ensure the images of the tongue are of adequate quality for the processing stage. The features of the captured images of the tongue are then extracted in real-time using image processing techniques which are primarily aimed at efficiently obtaining the contour of the tongue along with the centroid of the detected region. This helps to obtain the orientation and position of the tongue, and the system was successfully able to classify this information into one of 11 possible categories, thus producing 11 distinct outputs for the system which can be used by paralytic patients as inputs to any assistive device. The number of outputs obtained for the developed system are more than most of the other currently available solutions. After testing the system with a data entry application, an average of 19.34 correct selections per minute was obtained from 5 different experiments. In addition, an average error percentage per trial of 3.96% was achieved. The advantages of using the intraoral camera include the non-invasive nature of the system, the number of options available for the user at a given instance, and not requiring the tongue to establish any contact with a sensor or device.