Feature Extraction and Analysis of Plantar Pressure Data Based on Machine Learning

The study focuses on the preprocessing, feature extraction, and in-depth analysis of plantar pressure data. The data is preprocessed through cleaning, signal filtering, noise reduction, and time series correction and synchronization. Subsequently, global, local, and dynamic features are extracted to better analyze the distribution of plantar pressure. Finally, leveraging machine learning techniques, the study delves into the deep exploration of plantar pressure data. Local pressure distribution features and dynamic variation features are extracted using methods such as wavelet transformation for time series analysis. This study also incorporates principal component analysis (PCA) for feature dimensionality reduction and constructs a random forest classifier model. The simulation results demonstrate the model's high accuracy, reaching 95.67%, in distinguishing between healthy individuals and specific disease patients, showcasing its strong diagnostic and health assessment capabilities.