الفهرس 
Wearable Sensors in IoTOnly 14 pages are availabe for public view
 |  | from | 133 |  |  |
| | | from | 133 | | |
Abstract
Using on-body wearable sensors have a significant role in several domains, especially in health care. Monitoring some quantitative analysis of human motion can help patients and improve their quality of life. Moreover the automatic classification is the main computational task to be pursued during the monitoring process. The human physical activity can be classified using on body accelerometers. This thesis considered the elder people, who suffer from PD. PD is a chronic neurological disease that is linked to decrease in dopamine levels, causing abnormal brain activity and other signs. Generally, there are many symptoms of PD, including tremor, slowed movement, loss automatic movement, impaired posture and balance, speech changes, and writing change. The most common negative effect of PD is motor blocks (freezing) which affect the patient’s leg during walking and is gradually called FOG. FOG has substantial social and clinical consequences for patients. It is a common cause of falls, interferes with daily activities, and significantly impairs quality of life. Measuring FOG is complex, as it is highly sensitive to environmental triggers, cognitive input, and medication. For example, FOG occurs frequently at home and much less frequently in the doctor‘s office or in a gait laboratory. Networked sensors, either worn on the body or embedded in our living environments, are an application of the IoT. Such sensors gather rich indicative information about our physical and mental health. In IoT, vast amount of data is generated, which needs to be controlled and managed. Data management plays an important role, which includes data processing techniques, such as data aggregation and reduction. Hence, this thesis proposed a hybrid DWT-FFT feature extraction technique followed by IFS for significant feature selection in PD. Additionally, the choice of the effective specific position of the sensor depends on the significant selected features is studied using the classical ML methods. Also, DL network as LSTM is used for FOG detection by applying data directly without any feature extraction stages. Both methods have been applied to patientindependent and patient-dependent models.