Diagnosis of hand tremor using a smart phone accelerometer and SVM

One of the most common neurological diseases in the elderly is Parkinson's disease. It is very common in the elderly after Alzheimer's disease, affecting about 1% of the elderly worldwide. According to international statistics, the number of patients with Parkinson's is increasing every day, so that by 2030, their population will double. Parkinson's has no definitive cure and on the other hand, its severity increases over time so that the person can not do their daily tasks independently. Therefore, knowing the progression of this disease in the early stages can be helpful in controlling and preventing its progression. The aim of this project is to determine the level of disease severity using a mobile accelerometer and neural network. In this study, 111 people, including 78 men and 33 women with an average age of 31.56 years participate in this experiment. EMD algorithm decomposed signal to 8 IMF. We employed a SFS algorithm to select best feature to increase a speed of classification. our results show that skewness, kurtosis and energy feature have a best discrimination. the best average accuracy is 91.35% for first IMF.