Detection of different levels of Parkinson's disease based on the extraction of nonlinear features from the typing speed signal on The keyboard

Parkinson's disease is a neurological disease that belongs to diseases classified as a subset of movement disorders. Movement disorders are neurological conditions in which a person has difficulty slowly controlling their movements. Activities such as walking or having a cup of tea can be problematic. In some cases, people cannot rest some parts of their body remain in constant motion. Parkinson's is a chronic and progressive complication that focuses more on the elderly, but in some cases has been observed in young people. Due to the prevalence of disorders and complications caused by this disease, patients with it face many problems.In this research, first, specific filters with appropriate cut-off frequency are used to remove noise from the desired signal. After the preprocessing stage, the signal was considered once with windows and once without windows and feature extraction was performed. In window mode, the signal was divided into different intervals and in each interval, features were extracted and then averaged. In order to extract the feature, statistical features such as median, maximum, mean and standard deviation, skewness, elongation, and frequency such as median and mean frequency and nonlinear such as entropy and Lyapunov exponent and fractal dimension and Hurst view were used for all three data columns. Properties were calculated. To reduce the dimensions, LLE and ISOMAP methods were used and classification was performed with the help of SVM, MLP, and TREE classifiers. The results of sensitivity, accuracy, and specificity index for decision tree classification are ۸۸%, ۹۰%, and ۹۵%. And it is convenient. The regression diagram is also suitable for diagnosing the degree of disease for the fuzzy set with PSO optimization algorithm and has a low average error rate.

کلیدواژه ها:

Parkinson's disease ، typing speed signal ، Feature Extraction ، Classification