A novel peak detection algorithm using particle swarm optimization for chew count estimation of a contactless chewing detection
This article uses a proximity sensor to perform noncontact-based (sensing) chewing activity detection, capturing the temporalis muscle movement during food intake. The proposed approach is validated using data from a larger number of participants, 20, and different food types, eight. The proposed ch...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Published: |
Institute of Electrical and Electronics Engineers
2022
|
| Online Access: | http://psasir.upm.edu.my/id/eprint/100267/ https://ieeexplore.ieee.org/document/9942809 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | This article uses a proximity sensor to perform noncontact-based (sensing) chewing activity detection, capturing the temporalis muscle movement during food intake. The proposed approach is validated using data from a larger number of participants, 20, and different food types, eight. The proposed chewing detection classifies the chewing activity with an overall accuracy of 96.4% using a medium Gaussian support vector machine (SVM). In accordance with the result, this article proposes a novel chew count estimation based on particle swarm optimization (PSO). First, the base of the algorithm is developed based on counting the peak of the chewing signal. Next, the insignificant peak is removed by introducing an argument of minimum peak prominence and maximum peak width where the value of the parameters needs to be determined. As the individual chewing pattern varies from person to person, this article uses a novel parameter search using the PSO method to find the multiplier (parameter values) according to the average peak prominence and width value within each chewing episode. The proposed estimation approach simplifies the typical trial-and-error method. During optimization, within 100 iterations, the chewing count is reduced by 12.9% from its first iteration. Overall, the proposed methods achieve a 4.26% mean absolute error of chewing count estimation. |
|---|