1st International ICST Conference on Pervasive Computing Technologies for Healthcare

Research Article

Methods for Detection and Classification of Normal Swallowing from Muscle Activation and Sound

  • @INPROCEEDINGS{10.1109/PCTHEALTH.2006.361624,
        author={Oliver Amft and Gerhard Troster},
        title={Methods for Detection and Classification of Normal Swallowing from Muscle Activation and Sound},
        proceedings={1st International ICST Conference on Pervasive Computing Technologies for Healthcare},
        publisher={IEEE},
        proceedings_a={PERVASIVEHEALTH},
        year={2007},
        month={5},
        keywords={Swallowing detection event detection bolus viscosity classification bolus volume classification sensor collar.},
        doi={10.1109/PCTHEALTH.2006.361624}
    }
    
  • Oliver Amft
    Gerhard Troster
    Year: 2007
    Methods for Detection and Classification of Normal Swallowing from Muscle Activation and Sound
    PERVASIVEHEALTH
    IEEE
    DOI: 10.1109/PCTHEALTH.2006.361624
Oliver Amft1,*, Gerhard Troster1
  • 1: Wearable Computing Lab, ETH Zurich, Switzerland.
*Contact email: amtf@ife.ee.ethz.ch

Abstract

Swallowing is an important part of the dietary process. This paper presents an investigation to detect and classify normal swallowing during eating and drinking from electromyography and microphone sensors. The non-invasive sensors are selected in order to integrate them into a collar-like fabric for continuous monitoring of swallowing activity over a day. We compare methods for the detection of individual swallowing events from continuous sensor data. Furthermore we present a classifier comparison for the swallowing event properties volume and viscosity. The methods are evaluated on experimental data and a performance analysis is shown. Moreover we present a class skew analysis based on the metrics precision and recall