A carregar...
Publicação
Computing Velocity, Force and Power in Strength Exercises Using a Wearable Device and a Sport Performance Monitoring Machine (MYO-QUALITY)
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 11.87 MB | Adobe PDF |
Autores
Orientador(es)
Resumo(s)
Human activity recognition plays a critical role in fields such as healthcare, sports and rehabilitation, particularly in monitoring physical activity and strength exercises. Inertial Measurement Unit sensors, with accelerometers and gyroscopes are being increasingly more used in wearable devices, such as smartwatches, to track and quantify human activity movements during physical tasks. Velocity, force and power are three fundamental parameters to take in consideration in sports performance and strength exercises. This Masters’ dissertation introduces a novel approach that combines commercial wearable technology with a specialized performance machine for velocity estimation. The high-precision sports performance machine (M1 MYO-QUALITY), engineered for strength training, delivers real-time measurements of force and velocity (power is the multiplication of these two parameters). Concurrently, a commercial wearable device collects data from its embedded IMS sensors through a custom designed application. The data collected from these two devices is then aligned, prepared and segmented in order to be inputted in a deep learning model that combines the feature extraction capabilities of Convolutional Neural Networks with the temporal sequence learning of Long Short-Term Memory networks. In the first part, the four exercises performed were classified by the model due to the fact that it is necessary to identify which exercise is being performed in order to predict any of the parameters. Excellent results were achieved to support the incoming research. Subsequently, our deep learning model estimated the velocity with excellent results and accurately predicted the force. For the last parameter, power, excellent results were also achieved. The deep learning model accurately estimating the velocity in relation to the professional sports machine, facilitates the measurement of these parameters in scenarios where M1 MYO-QUALITY is not accessible to athletes. To build an improved and more robust model, a bigger database and hyperparameter tunning are crucial.
Descrição
Tese de mestrado, Engenharia Biomédica e Biofísica, 2024, Universidade de Lisboa, Faculdade de Ciências
Palavras-chave
Exercícios de força Máquina M1 MYO-QUALITY Sensores Unidades Inerciais Classificação exercícios Estimação de variávies Teses de mestrado - 2024