Deep learning for human activity and emotion detection in IoT

It is important to understand the working principle of an accelerometer- and gyroscope-based human activity detection system, and of a facial expression-based emotion detection system, before discussing the useful deep learning models.

Automatic human activity recognition system

Automatic human activity recognition (HAR) systems detect human activity, based on raw accelerometer and gyroscope signals. The following diagram shows a schematic of a DL-based HAR that consists of three different phases. They are as follows:

• IoT deployment or instrumentation of the subject or person
• Feature extraction and model development
• Activity classification/identification

Generally...