Activity recognition is an underpinning step in behavior analysis, addressing healthy lifestyles, fitness tracking, remote assistance, security applications, elderly care, and so on. Activity recognition transforms low-level sensor data from sensors, such as an accelerometer, gyroscope, pressure sensor, and GPS location, to a higher-level description of behavior primitives.

In most cases, these are basic activities, for example, walking, sitting, lying, jumping, and so on, as shown in the following diagram, or they could be more complex behaviors, such as going to work, preparing breakfast, and shopping:

In this chapter, we will discuss how to add the activity recognition functionality into a mobile application. We will first look at what an activity recognition problem looks like, what kind of data we need to collect, what the main challenges are, and how to address them.

Later, we will follow an example to see how to actually implement activity recognition in an Android application, including data collection, data transformation, and building a classifier.

Let's start!