Diseño y construcción de un dispositivo electrónico de estimulación temprana motriz gruesa para niños de 4 a 5 años empleando técnicas de visión artificial.

Abstract

The objective of this curricular integration work was to design and build an electronic device for early gross motor stimulation for children from 4 to 5 years old using artificial vision techniques; by researching bibliographical sources and studies that support the favorable development of gross motor skills in early development stages, the exercises taken into account for the tests were identified. When establishing the requirements of the device, a Raspberry Pi model B was selected as the controller, and Pycharm Community Edition as the programming enviroment. The programming was developed in Python language, which used segmentation and recognition of positions and articulations through an estimation model of the human body; PosNet based on Tensorflow and Mediapipe architecture for the detection of the human body. The functions encontrarpose, encontrarpuntos, and findangle, which are the base of the entire programming structure, were programmed. Tests were run using a webcam Connected to the device in both children and adults to establish the initial setting, it was verified that the system is capable of identifying the most relevant points of the strong joints of the human body. In the tests carried out with 15 infants, aged between 4 and 5 years, the device was able to recognize the main joints of the body at an average distance of 241cm. Among the 6 established exercises, every kid satisfactorily performed each activity in an average time of 3 minutes and 40 seconds. Of the 5 attempts executed for each of the 6 exercises, the children on average succeeded in four exercises with an effectiveness of 89%. For a better answer regarding the scalability of the project when adding more exercises to the routine, it is necessary to carry out a sensory survey in the spatial plane.