A classification of accelerometer data to differentiate pedestrian state
Pichaya Prasertsung, Teerayut Horanont · 2016
This research is motivated through the demand to create routing in indoor environment based on activity recognition approach. A model to discriminate between walking, climbing up stair, and climbing down stair is introduced. Data was collected from a group of participants performing walking up stairs, walking down stairs, and walking on normal path inside the building. 35 features are considered in order to build a model. The classification is carried out using support vector machine (SVM) using leave one person out cross validation method. From this study, a combination of features extracted from raw accelerometer data and filtered data can be used to improve a classification result. Based on this dataset, a combination of features extracted from body acceleration data and raw accelerometer data itself, provides accuracy up to 85.73% outperform other combination data.