Meal-assistance by Robot Suit HAL using detection of food position with camera

Meal-assistance is necessary for patients with upper limb paralysis caused by diseases such as cerebrovascular disorder and cervical cord damage. We have developed the upper limb type HAL (HAL-UL) to assist the arm movement during a meal for people with upper limb paralysis. HAL-UL achieves reaching movements according to pre-generated trajectories. Furthermore, for meal assistance it is required to provide the reaching movement for various locations, shapes, and amounts of food. Therefore, the purpose of this research is to achieve the detection the food position by HAL during meal-assistance using image processing and to verify its effectiveness. A camera is installed overhead to detect the image of the food on the table. Then the locations of the food pieces are detected using image processing. A microphone is installed as a command interface for the user to command the dish number. Finally, the optimal trajectory to the location is selected from pre-generated reaching trajectories. To determine the effectiveness of the proposed method, an eating experiment was performed. A subject wore the HAL-UL in front of a table, and 4 dishes with different types of food were prepared. Then, the image of the food was processed, the food locations were detected, and the optimal reaching trajectories were selected. According to the determined trajectories the HAL-UL provided eating movement support for the subject. Therefore, the effectiveness of the proposed method was determined and we confirmed the HAL-UL for meal-assistance may be used successfully for the actual meal-assistance.