Hey everyone, I wanted to share a fun project I recently worked on that combines audio detection with machine learning to solve a real-life problem. I’ve always been a bit of a tinkerer, and recently I decided to tackle the issue of detecting when my LG washing machine completes a cycle. It’s a small thing, but being able to get notified when the wash is done without having to constantly check the machine has been a huge convenience!
I started by researching how to detect the specific melody my washer plays when it finishes. After some digging, I came across a project that uses an ESP32 with ESP-DSP to generate spectrograms and then applies machine learning to identify the completion signal. I thought this was perfect because it’s both cost-effective and leverages existing open-source tools.
The setup involved attaching a small microphone to the washer and connecting it to the ESP32. I then trained a simple machine learning model using TensorFlow Lite Micro to recognize the unique sound pattern of the completion melody. Once everything was calibrated, I integrated the system with Home Assistant using MQTT, which allows me to receive notifications on my phone and even trigger other automations, like turning on the dryer.
One challenge I faced was ensuring the microphone picked up the sound reliably without picking up too much ambient noise. I ended up using a small acoustic foam sleeve to help isolate the sound source. It was a bit of trial and error, but it paid off in the end!
I’d love to hear if anyone else has tackled similar projects or has tips for improving the detection accuracy. It’s been a great learning experience, and I’m excited to see how else I can apply these techniques to other appliances around the house! 
