Cyber-Physical Systems

Machine Learning in Monitoring Physical Activity  ◾  197

14.4 Discussions

In this chapter, three different machine learning approaches were applied to a CPS problem of

recognition of postures and activities. e SVM and MLP were applied to the raw sensor data. e

DT classification was used on the features extracted from the raw sensor signals. e SmartShoe

system demonstrated high-activity classification accuracy with all of the machine learning meth-

ods used.

It can be observed from the confusion matrices that SVM and MLP achieved high over-

all accuracy, 97.3% and 97.6%, in classifying these six activities. Because both methods pro-

duced compa