Using Multi-Feature Fusion for Detecting Freezing of Gait Episodes in Patients with Parkinson's Disease

This paper proposes a model for detecting Freezing of Gait (FoG) episodes in patients with Parkinson's Disease (PD) using multi-feature fusion. The proposed approach applies two schemes for feature extraction. The first one is time-domain statistical feature engineering and the second one is spectrogram-based time-frequency analysis by Convolutional Neural Network (CNN) feature learning. The two extracted feature sets are fused with applying Principal Component Analysis (PCA) algorithm for dimensionality reduction. Benchmark dataset of three tri-axial accelerometer sensors for patients with PD is tested in both principle-axes and angular-axes. Moreover, performance of the proposed approach is characterized on experiments considering several Machine Learning (ML) algorithms. Experimental results show that using multi-feature fusion with PCA dimensionality reduction outperforms using typical single feature sets. The significance of this study is to highlight the impact of using multi-feature fusion on the performance of FoG episodes detection.