The team decided to use embedded machine learning to identify the sounds of the surrounding environment, and so they decided to go with the Arduino Nano 33 BLE Sense. After recording numerous examples of various events for example, a conversation making a knock on the doors, watching television, the doorbell and silence and feeding the recordings into an incredibly small model Edge Impulse’s Studio. The model that was created could successfully distinguish between different events for around 90 percent times.
In addition to sending the recognized audio data to an Arduino serial screen, their software also allows the output to be transferred over Bluetooth(r) with low Energy so that a connected phone can receive the information and display the information. The app for mobile devices has three buttons that allow access to an audio list that can be played, as well as certain settings. There is also an additional menu for managing the connection to the Arduino.