Freezing of Gait (FoG) Detection Using Logistic Regression in Parkinson's Disease from Acceleration Signals

The detection and diagnosis of Parkinson disease (PD) are very important concerning the treatment of this disease. In this work, the freezing of gait (FoG) from subjects with Parkinson disease has been detected by the logistic regression modeling. To complete this work, first, the acceleration sensor has been placed on the ankle of the patients to get the signals. Second, the features from these acceleration signals have been extracted by the Fast Fourier Transform (FFT) algorithm. With the FFT algorithm, the frequency coefficients have been gotten. To diminish the number of features, the statistical measures including variance, maximum amplitude, minimum amplitude, maximum energy, and minimum energy, have been applied to frequency coefficients of these signals. So, for each class (FoG and no- FoG), five parameters have been extracted. Eight patients are having Parkinson disease in the dataset. After feature extraction, the logistic regression modeling has been used to detect the freezing of gait cases from the dataset. The classification of the accuracy of 81.3% has been achieved in the classification of FoG cases having PD from the acceleration signals. In addition to logistic regression, four different classifiers (Linear SVM, Quadratic SVM, Cubic SVM, and kNN) have been used to classify the FoG cases. The obtained results have shown that the proposed method could be used in the detection and identification of Parkinson disease using only a sensor of acceleration.