Skip to main content
padlock icon - secure page this page is secure

Optimization of Sensor Placement Combinations and Classification Thresholds for the Accelerometer-Based Activity Recognition

Buy Article:

$106.22 + tax (Refund Policy)

Binary decision tree classification and usage of multiple accelerometers have been suggested to be simple and accurate method for the recognition of daily activities. However, the determination of classification thresholds and sensor placement combinations have been empirical and subjective. The purpose of this study was to find optimal set of sensors and classification thresholds in a systematic way considering both accuracy and decision margin of binary decisions. Sixteen healthy men participated in this study. Wireless tri-axial accelerometers were attached on eight body locations i.e., forehead, arm, chest, waist, wrist, thigh, shank, and ankle. Subjects performed three static activities (standing, sitting, and lying) and two dynamic activities (walking and running). Binary tree classification scheme with four binary decisions was constructed using signal magnitude area, variance of vertical acceleration, and tilt angle as classification features. Optimal thresholds and decision margin for every sensor location were calculated for each binary decision. Then, all sensor combinations (with 1–3 sensors) were evaluated in terms of total accuracy and overall decision margin to derive optimal ones. Three of four binary decisions (static vs. dynamic, walking vs. running, and upright posture vs. lying) showed 100% accuracy at multiple sensor locations. The classification of standing versus sitting was rather difficult, where the best accuracy was 98.75% at two sensor locations (wrist and thigh). The optimal single sensor location was the wrist (accuracy: 98.15%, margin: 22.8%). Optimal combination of two sensor locations was arm and thigh (accuracy: 99.5%, margin: 209.3%) and that of three sensors was arm, thigh, and ankle (accuracy: 99.5%, margin: 214.8%). Wrist would be the best option for single sensor recognition. If accuracy and decision margin are of critical interest, two sensors at arm and thigh are recommended. Three sensors are not recommended because of minor improvement.
No Reference information available - sign in for access.
No Citation information available - sign in for access.
No Supplementary Data.
No Article Media
No Metrics

Keywords: ACCELEROMETER; ACTIVITY CLASSIFICATION; BINARY TREE; COMBINATION; SENSOR LOCATION; THRESHOLD

Document Type: Research Article

Publication date: January 1, 2018

More about this publication?
  • Journal of Medical Imaging and Health Informatics (JMIHI) is a medium to disseminate novel experimental and theoretical research results in the field of biomedicine, biology, clinical, rehabilitation engineering, medical image processing, bio-computing, D2H2, and other health related areas.
  • Editorial Board
  • Information for Authors
  • Subscribe to this Title
  • Ingenta Connect is not responsible for the content or availability of external websites
  • Access Key
  • Free content
  • Partial Free content
  • New content
  • Open access content
  • Partial Open access content
  • Subscribed content
  • Partial Subscribed content
  • Free trial content
Cookie Policy
X
Cookie Policy
Ingenta Connect website makes use of cookies so as to keep track of data that you have filled in. I am Happy with this Find out more