Load Position Estimation Method for Wearable Devices Based on Difference in Pulse Wave Arrival Time

Figures

• 

  Figure 1. System flow of the proposed method.

• 

  Figure 2. Example of peak detection: each • denotes a peak found during the process).

• 

  Figure 3. ECG and pulse wave peaks and time differences for (a) a fingertip, (b) an upper arm, and (c) a toe. The black line represents the ECG, and the red line represents the pulse wave. The green and blue circles indicate the ECG and pulse wave peaks, respectively.

• 

  Figure 4. Position estimation based on the KL divergence between the training and test data distributions.

• 

  Figure 5. Target body positions.

• 

  Figure 6. Average and standard deviation of peak time differences.

• 

  Figure 7. Split datasets used in the evaluation experiment.

• 

  Figure 8. Illustration of two experimental environments using data consisting of four sessions each from three subjects. The circles represent the peak time differences in each session.

• 

  Figure 9. Examples of final load position estimation by the Mini-method and Vote-method.

• 

  Figure 10. Average F-value when varying the input data length for 12 body parts by using the Mini-method in the user-dependent environment.

• 

  Figure 11. Average F-value when varying the input data length for 12 body parts by using the Vote-method in the user-dependent environment.

• 

  Figure 12. Maximum of the average F-value when varying the number of target body parts with 180 s input data by using the Mini-method in the user-dependent environment.