This is the repo for a diy embeded system project.
The system will wake up due to a recorded and detected wake word , load a audio from a storage media and play it. This repository involves two projects. First the ESP32 flash and secondly the wake word detection.
Big shoutout to atomic14 and his project diy-alexa. This is where I took inspiration and guidlines on how to make this project work.
The wake word detection will be done with tensorflow. After a model is trained with the required data it is converted to a tensorflow lite model. Then the tensorflow lite model is transcribed to an applicable format for the use with the ESP32.
The ESP32 will be mostly in doze mode recording surrounding sound and try to detect the wake word with the tensorflow lite model. When it was successful the ESP32 will load an audio sample from a storage media and send it's data via I2C to an audio amplifier.