11th International Conference on Body Area Networks

Research Article

Eigenwalk: a Novel Feature for Walk Classification and Fall Prediction

  • @INPROCEEDINGS{10.4108/eai.15-12-2016.2267645,
        author={masoud hemmatpour and Renato Ferrero and Bartolomeo Montrucchio and Maurizio Rebaudengo},
        title={Eigenwalk: a Novel Feature for Walk Classification and Fall Prediction},
        proceedings={11th International Conference on Body Area Networks},
        publisher={ACM},
        proceedings_a={BODYNETS},
        year={2017},
        month={4},
        keywords={fall prevention health care accelerometer gyroscope smartphone},
        doi={10.4108/eai.15-12-2016.2267645}
    }
    
  • masoud hemmatpour
    Renato Ferrero
    Bartolomeo Montrucchio
    Maurizio Rebaudengo
    Year: 2017
    Eigenwalk: a Novel Feature for Walk Classification and Fall Prediction
    BODYNETS
    EAI
    DOI: 10.4108/eai.15-12-2016.2267645
masoud hemmatpour1,*, Renato Ferrero1, Bartolomeo Montrucchio1, Maurizio Rebaudengo1
  • 1: Politecnico di Torino
*Contact email: masoud.hemmatpour@polito.it

Abstract

Predicting a fall is one of the most promising approach to avoid it. Different studies strive to classify abnormal and normal walks in order to predict a fall before its occurrence. This study introduces eigenwalk, a novel feature based on the principal components of the accelerometer and gyroscope signals. This feature, in conjunction with a random forest classifier, is able to distinguish walk patterns and to estimate a fall risk. As the accelerometer and the gyroscope embedded in a smartphone are recognized to be precise enough for fall avoidance systems, they have been exploited in an experimental analysis in order to compare the proposed approach with the most recent ones. The results have shown that the new feature in combination with the random forest classification outperforms state-of-the-art approaches, by improving the accuracy up to 98.6%.