Sensor Data and Edge AI Preparation

Sensor Data and Edge AI Preparation · TESA Firmware SDK for Edge AI course

Objectives

Get sensor data ready for Edge AI: read at a fixed period, filter, normalise, arrange it into data windows, and pass it on to a host or a Digital Twin.

Lessons

#	Lesson	Content
1	A sensor stream ready for Edge AI	From a sensor pin to a stream ready for a model: a fixed period, filters, normalising, data windows, and passing data on to a host
2	Lab: a sensor stream and a data window ready for AI	Reading two kinds of sensor with a fixed-period task, filtering or normalising, building a data window, then choosing an extension (fusion, host telemetry, or an event)

Approximate time per the original: about 3.5–4 hours (lessons) + a 2.5–3.5 hour lab.

Accompanying sheets and templates (in the resources/ folder of lesson 1):

• sensor-ai-prep.md

The C code in this module uses the API of the TESAIoT Bitstream firmware, which is not yet open source. Read the note at the top of the lesson before you start.

Checkpoint

Before moving to the next module, check that you can do the following:

• Read at least two kinds of sensor in a fixed-period task
• Have at least one filter or normalisation step in the lab code
• A data window produces a summary vector at least once per second
• The table in sensor-ai-prep.md and a short report are both filled in completely

← Module 4 · Course page · Module 6 →