8th International Conference on Pervasive Computing Technologies for Healthcare

Research Article

Pervasive Eating Habits Monitoring and Recognition through a Wearable Acoustic Sensor

  • @INPROCEEDINGS{10.4108/icst.pervasivehealth.2014.255423,
        author={Yin Bi and Wenyao Xu and Nan Guan and Yangjie Wei and Wang Yi},
        title={Pervasive Eating Habits Monitoring and Recognition through a Wearable Acoustic Sensor},
        proceedings={8th International Conference on Pervasive Computing Technologies for Healthcare},
        publisher={ICST},
        proceedings_a={PERVASIVEHEALTH},
        year={2014},
        month={7},
        keywords={eating habit feature extraction knn svm},
        doi={10.4108/icst.pervasivehealth.2014.255423}
    }
    
  • Yin Bi
    Wenyao Xu
    Nan Guan
    Yangjie Wei
    Wang Yi
    Year: 2014
    Pervasive Eating Habits Monitoring and Recognition through a Wearable Acoustic Sensor
    PERVASIVEHEALTH
    ACM
    DOI: 10.4108/icst.pervasivehealth.2014.255423
Yin Bi1,*, Wenyao Xu2, Nan Guan1, Yangjie Wei1, Wang Yi1
  • 1: Northeastern University, China
  • 2: The State University of New York at Buffalo
*Contact email: biyin0125@126.com

Abstract

Eating habits provide clinical diagnosis evidences of lifestyle related diseases, such as dysphagia and indigestion. However, it is costly to obtain eating habit information of common people in terms of both time and expenses. This paper presents a pervasive approach for eating habit monitoring and recognition by a necklace-like device and a smartphone communicating via bluetooth. The necklace-like device acquires acoustic signals from the throat, and the data are processed in the smartphone to recognize important features. With complex acoustic signals collected from the throat, our method comprehensively analyzes and recognizes different events including chewing, swallowing, and breathing in the smartphone. Experiments show that the proposed approach can recognize different acoustic events effectively, and the recognition accuracy with K-Nearest Neighbor (KNN) and Support Vector Machine (SVM) is 86.82% and 98.35%, respectively. Finally, a real eating case study is conducted to validate the proposed approach.