- Terminus FE1.1 USB hub board: the solution to connect four USB devicesPosted 2 weeks ago
- Understanding the Mechanics of 3D PrintingPosted 2 months ago
- SDS011 the Air Quality SensorPosted 3 months ago
- NIXIE STYLE LED DISPLAYPosted 6 months ago
- TOTEM: learning by experimentingPosted 6 months ago
- Google Assistant Voice Controlled Switch – NodeMCU IOT ProjePosted 7 months ago
- Water Softener Salt Level MonitorPosted 7 months ago
- Sparkly Air SensorPosted 7 months ago
- Ultra sonic distance finder with live statusPosted 7 months ago
- Windows interface to have total control over lampsPosted 7 months ago
“Listen to temperatures” with TinyML
Can we “listen” a difference between pouring hot and cold water?
As you see in the video you can do it, but why? It is mentioned that the change is due to complex fluid dynamics reasons. Beyond the scientific investigation, the question we asked ourselves was: is this ability to “listen to temperatures” something that can be replicated using Artificial Neural Networks? We then tried to create an experiment using TinyML (Machine Learning applied to embedded devices).
We used very different water temperatures, with a range of 50 ° C between them. (11 ° C and 61 ° C); for each sample the listening time was the time taken by the glass to be filled (from 3 to 5 seconds). We are interested in capturing the sound of water only during the pouring process. The sound was recorded from the same digital microphone (sampling frequency: 16KHz) and stored as a .wav file in 3 different folders: cold water sound (“Cool”); sound of hot water (“Hot”); no sound of water (“Noise”).
The acquired dataset (via Arduino Nano 33 BLE Sense) was uploaded to Edge Impulse Studio, where it was preprocessed, the neural network (NN) model was trained, tested and deployed on an MCU for a real physical test (an iPhone was also used for live classification).