| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 5.61 MB | Adobe PDF |
Advisor(s)
Abstract(s)
Este documento descreve a utilização de técnicas de Deep Learning para a criação de um sistema de reconhecimento de movimentos efetuados pelo ser humano, tais como andar, subir e descer escadas, entre outros. Esses movimentos são captados por sensores presentes num smartphone comum, tais como giroscópios e acelerómetros. Esta aplicação torna-se interessante, pois poder-se-á tornar num produto de
monitorização de movimentos para pessoas com dificuldades de mobilidade ou idosos com essas e outras dificuldades. O principal objetivo deste projeto é o de utilizar metodologias de aprendizagem máquina para a resolução do problema, recorrendo a diferentes tipos de redes neuronais que, aproveitando as capacidades de aprendizagem, levarão a uma posterior análise do seu desempenho. Deste modo, o trabalho apresentado terá o propósito de avaliar qual ou quais as redes que melhor desempenho demonstraram no reconhecimento
dos movimentos.
The present document aims to describe the use of Deep Learning techniques to create a human motion recognition system, that can recognize walking, climbing up or down stairs, and so on. The motions are captured by smartphone sensors, like accelerometers and gyroscopes. This approach is interesting, since it can become a product to monitor the motion of people with mobility problems and old people that have those problems and more. The main objective of this project is to use Machine Learning methodologies to solve this problem, using different types of Neural Networks that, taking advantage of learning capabilities, can lead to a posterior analysis of its performance. Therefore, the purpose of this project is to evaluate the best Neural Networks to recognize motions.
The present document aims to describe the use of Deep Learning techniques to create a human motion recognition system, that can recognize walking, climbing up or down stairs, and so on. The motions are captured by smartphone sensors, like accelerometers and gyroscopes. This approach is interesting, since it can become a product to monitor the motion of people with mobility problems and old people that have those problems and more. The main objective of this project is to use Machine Learning methodologies to solve this problem, using different types of Neural Networks that, taking advantage of learning capabilities, can lead to a posterior analysis of its performance. Therefore, the purpose of this project is to evaluate the best Neural Networks to recognize motions.
Description
Keywords
Redes neuronais LSTM CNN FNN Machine learning Deep learning Inteligência artificial Neural networks Artificial intelligence