Reconhecimento de gestos de membros superiores utilizando sensores de movimento e fotopletismografia

Abstract

This project aims to evaluate machine learning techniques using low frequency photoplethysmog raphy coupled with motion sensors from wearable devices, such as smart watches, in recognizing wrist and finger gestures. After gesture segmentation based on the identification of motion artifacts in the photoplethysmography (PPG) signal, classification models using Support Vector Machines, Random Forests and XgBoost were trained using statistical attributes extracted from PPG signals and motion sensors. The project indicates that frequencies of 25 Hz are suitable for the recognition process, achieving results of up to 82% accuracy and 82% recall.