Person Identification using Door Accelerations

First, 1000 entrances by 12 people were recorded as the training data. Second, 266 features were extracted from preprocessed entrance data. Third, feature selection was performed to reduce the number of features to 36. Finally, the classification models were trained using four machine-learning algorithms. To classify an entry, the selected features are extracted from the raw acceleration signal in order to obtain a feature vector, which is finally classified using the most accurate classification model, which predict the identity of the person who entered.