Gait rehabilitation monitor

• Autor(a) principal: Leite, Paulo
• Data de Publicação: 2018
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10071/18574
• topic: Gait; Classification; Rehabilitation; Android; Engenharia de telecomunicações; Telemóvel; Machine learning; Fisioterapia; Reabilitação; Sensor; Aplicação informática

This work presents a simple wearable, non-intrusive affordable mobile framework that allows remote patient monitoring during gait rehabilitation, by doctors and physiotherapists. The system includes a set of 2 Shimmer3 9DoF Inertial Measurement Units (IMUs), Bluetooth compatible from Shimmer, an Android smartphone for collecting and primary processing of data and persistence in a local database. Low computational load algorithms based on Euler angles and accelerometer, gyroscope and magnetometer signals were developed and used for the classification and identification of several gait disturbances. These algorithms include the alignment of IMUs sensors data by means of a common temporal reference as well as heel strike and stride detection algorithms to help segmentation of the remotely collected signals by the System app to identify gait strides and extract relevant features to feed, train and test a classifier to predict gait abnormalities in gait sessions. A set of drivers from Shimmer manufacturer is used to make the connection between the app and the set of IMUs using Bluetooth. The developed app allows users to collect data and train a classification model for identifying abnormal and normal gait types. The system provides a REST API available in a backend server along with Java and Python libraries and a PostgreSQL database. The machine-learning type is Supervised using Extremely Randomized Trees method. Frequency, time and time-frequency domain features were extracted from the collected and processed signals to train the classifier. To test the framework a set of gait abnormalities and normal gait were used to train a model and test the classifier.