On the Use of Decision Tree for Posture Recognition

The aim of this study is to evaluate the effectiveness of decision tree as classifier for recognition of four main human postures (standing, sitting, bending and lying) since decision trees are well known for their success for prediction, recognition and classification task in data mining problems. Firstly, the eigenfeatures of these postures are optimized via Principal Component Analysis rules of thumb specifically the KG-rule, Cumulative Variance and the Scree Test. Next, these eigenfeatures are statistically analyzed prior to classification. In doing so, the most relevant eigenfeatures that we termed as eigenpostures can be ascertained. Further, we employed decision tree as classifier for posture recognition. Initial results of the experiments are encouraging which suggested that our method can efficiently be applied for posture classification using DT.