Abstract

Human activity recognition systems are designed to capture the state of the user and its environment through the use of heterogeneous sensors.This can be extremely useful in health applications, for automated and smart daily monitoring of activities for the elderly persons.This approach can indeed lead to intelligent and automatic monitors in real time for human activities for electronic health applications.

Key parameters

In this project we use convolution neural network(cnn) with the following parameters:

• kernelSize=10
• numberOfLabels=6
• batch_size = 15
• depth=10
• num_hidden=300
• dropout=0.6

How it works

Here, the data can be collected through accelerometers, gyroscopes, GPS, or the built-in sensor from a smartphone. The data is then saved to the .CSV file. This CSV file can be used to train, test, and verify the neural network. When the network is trained, we can use the confusion matrix to classify the activity.