Publicação
IMUs: validation, gait analysis and system’s implementation
| Resumo: | Falls are a prevalent problem in actual society. The number of falls has been increasing greatly in the last fifteen years. Some falls result in injuries and the cost associated with their treatment is high. However, this is a complex problem that requires several steps in order to be tackled. Namely, it is crucial to develop strategies that recognize the mode of locomotion, indicating the state of the subject in various situations, namely normal gait, step before fall (pre-fall) and fall situation. Thus, this thesis aims to develop a strategy capable of identifying these situations based on a wearable system that collects information and analyses the human gait. The strategy consists, essentially, in the construction and use of Associative Skill Memories (ASMs) as tools for recognizing the locomotion modes. Consequently, at an early stage, the capabilities of the ASMs for the different modes of locomotion were studied. Then, a classifier was developed based on a set of ASMs. Posteriorly, a neural network classifier based on deep learning was used to classify, in a similar way, the same modes of locomotion. Deep learning is a technique actually widely used in data classification. These classifiers were implemented and compared, providing for a tool with a good accuracy in recognizing the modes of locomotion. In order to implement this strategy, it was previously necessary to carry out extremely important support work. An inertial measurement units’ (IMUs) system was chosen due to its extreme potential to monitor outpatient activities in the home environment. This system, which combines inertial and magnetic sensors and is able to perform the monitoring of gait parameters in real time, was validated and calibrated. Posteriorly, this system was used to collect data from healthy subjects that mimicked Fs. Results have shown that the accuracy of the classifiers was quite acceptable, and the neural networks based classifier presented the best results with 92.71% of accuracy. As future work, it is proposed to apply these strategies in real time in order to avoid the occurrence of falls. |
|---|---|
| Autores principais: | Ribeiro, Nuno Miguel Ferrete |
| Assunto: | Falls Gait parameters Inertial measurement units (IMUs) Sensory fusion’s algorithms Calibration Principal Component Analysis (PCA) ASMs Deep learning Quedas Parâmetros da marcha Unidades de medição inercial (IMUs) Algoritmos de fusão sensorial Calibração |
| Ano: | 2017 |
| País: | Portugal |
| Tipo de documento: | dissertação de mestrado |
| Tipo de acesso: | acesso aberto |
| Instituição associada: | Universidade do Minho |
| Idioma: | inglês |
| Origem: | RepositóriUM - Universidade do Minho |
| Resumo: | Falls are a prevalent problem in actual society. The number of falls has been increasing greatly in the last fifteen years. Some falls result in injuries and the cost associated with their treatment is high. However, this is a complex problem that requires several steps in order to be tackled. Namely, it is crucial to develop strategies that recognize the mode of locomotion, indicating the state of the subject in various situations, namely normal gait, step before fall (pre-fall) and fall situation. Thus, this thesis aims to develop a strategy capable of identifying these situations based on a wearable system that collects information and analyses the human gait. The strategy consists, essentially, in the construction and use of Associative Skill Memories (ASMs) as tools for recognizing the locomotion modes. Consequently, at an early stage, the capabilities of the ASMs for the different modes of locomotion were studied. Then, a classifier was developed based on a set of ASMs. Posteriorly, a neural network classifier based on deep learning was used to classify, in a similar way, the same modes of locomotion. Deep learning is a technique actually widely used in data classification. These classifiers were implemented and compared, providing for a tool with a good accuracy in recognizing the modes of locomotion. In order to implement this strategy, it was previously necessary to carry out extremely important support work. An inertial measurement units’ (IMUs) system was chosen due to its extreme potential to monitor outpatient activities in the home environment. This system, which combines inertial and magnetic sensors and is able to perform the monitoring of gait parameters in real time, was validated and calibrated. Posteriorly, this system was used to collect data from healthy subjects that mimicked Fs. Results have shown that the accuracy of the classifiers was quite acceptable, and the neural networks based classifier presented the best results with 92.71% of accuracy. As future work, it is proposed to apply these strategies in real time in order to avoid the occurrence of falls. |
|---|