.png){kind=icon}
.png){kind=icon}
Return
HSVM-Based Human Activity Recognition Using Smartphones
.png){kind=icon}
PRE
AI
DOI：10.1007/978-3-030-27541-9_19.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Human Activity Recognition (HAR) based on smartphones has a lot of applications to our daily life. Through the years, researchers reached different goals in this topic. Different issues were presented at each approach like accuracy, computational cost, device placement, battery life and limited hardware and software resources. In this paper we developed an HSVM that is capable of classifying 5 types of human activities walking, running, jumping, standing and sitting. 5-fold cross validation was used for the offline training achieving very high accuracy, but in online validation the result was no as accurate as in offline validation. A device placement (orientation) in pocket was considered too, then we present a variety of confusion matrices with four, possible device placement, and their respective accuracy. Wavelet Transform was used to minimize noise into the feature vector.
Keywords:
Human Activity Recognition
HSVM
Sensor fusion
Journal
I
IF:
0
Papers:
7
Citations:
0