Optimización e implementación de disposotivo vestible para la medición de las fases al caminar

Abstract

This research work focuses on the implementation of robotic devices in the rehabilitation of patients with mobility problems. The main objective is a reduction of the existing device based on resistive force sensors (FSR) to improve its comfort and practicality when taking measurements, while maintaining the arrangement of the sensors and the second device that employs an MPU6050 inertial sensor thus maintaining consistency in data collection. Additionally, a database of 30 individuals was taken and used to create a new, more general algorithm for detecting walking phases. This algorithm was developed with the purpose of improving the accuracy and efficiency in identifying the different stages of the human gait cycle. Thanks to the diversity of data collected from these 30 people, it was possible to obtain a robust and versatile model that can adapt to different characteristics and variations in people's walking patterns. This advance in the detection of walking phases has the potential to benefit various fields, such as medicine, rehabilitation and assistive technology for people with motor difficulties. In the previous implementation of the device, there were two classification algorithms, Perez's algorithm, which gives an F1-score of 0.64 in the classification of the 5 phases of walking using data from a single person and a neural network (perceptron) which improves the F1-score to 0.93 also using data from a single person and reducing to 4 phases of walking, the device was made with THT technology, which by its welding characteristics and the location of the elements on the plate allowed it to have a size of 10x10 cm, in contrast, in this degree project using the Perceptron to classify a database of 33 people and 5 phases of walking, obtained an F1-score value of 0.22, Therefore, it was decided to implement two fuzzy algorithms which obtain an F1-score of 0.68 for the MFC Clusterin and an F1-score of 0.69 for the Grid Partition algorithm, both algorithms retake the classification of the 5 phases of walking and classifying on a database of 33 people, in addition, the device made with SMT technology allowed to reduce its dimensions by 40%, obtaining a final size of 6x6 cm.