Conference paper Open Access

A Weighted Late Fusion Framework for Recognizing Human Activity from Wearable Sensors

Athina Tsanousa; Georgios Meditskos; Stefanos Vrochidis; Ioannis Kompatsiaris

Citation Style Language JSON Export

  "publisher": "Zenodo", 
  "DOI": "10.5281/zenodo.3507004", 
  "author": [
      "family": "Athina Tsanousa"
      "family": "Georgios Meditskos"
      "family": "Stefanos Vrochidis"
      "family": "Ioannis Kompatsiaris"
  "issued": {
    "date-parts": [
  "abstract": "<p>Following the technological advancement and the<br>\nconstantly emerging assisted living applications, sensor-based activity<br>\nrecognition research receives great attention. Until recently,<br>\nthe majority of relevant research involved extracting knowledge<br>\nout of single modalities, however, when individual sensors performances<br>\nare not satisfactory, combining information from multiple<br>\nsensors can be of use and improve the activity recognition rate.<br>\nEarly and late fusion classifier strategies are usually employed<br>\nto successfully merge multiple sensors. This paper proposes a<br>\nnovel framework for combining accelerometers and gyroscopes<br>\nat decision level, in order to recognize human activity. More<br>\nspecifically, we propose a weighted late fusion framework that<br>\nutilizes the detection rate of a classifier. Furthermore, we propose<br>\nthe modification of an already existing class-based weighted late<br>\nfusion framework. Experimental results on a publicly available<br>\nand widely used dataset demonstrated that the combination of<br>\naccelerometer and gyroscope under the proposed frameworks<br>\nimproves the classification performance.</p>", 
  "title": "A Weighted Late Fusion Framework for Recognizing Human Activity from Wearable Sensors", 
  "type": "paper-conference", 
  "id": "3507004"
All versions This version
Views 3434
Downloads 292292
Data volume 61.6 MB61.6 MB
Unique views 3333
Unique downloads 284284


Cite as