• 1
    Interfacing Arduino Nano 33 BLE Sense with Edge Impulse

    Edge Impulse fully supports the Arduino Nano 33 BLE Sense, a compact development board containing a Cortex-M4 microprocessor, motion sensors, a microphone, and BLE. The studio supports sampling raw data, the development of models, and deploying trained machine learning models. It costs roughly $30 and is available from Arduino and a variety of distributors.

    This can be easily achieved by installing all the dependencies on your system from Edge Impulse docs.

    Once you have configured flashed the Nano 33 BLE sense with the edge-impulse framework we can continue with the next step.

  • 2
    Building a machine learning model

    We require a lot of bird data and it is impractical to find data that is high in quantity as well as quality, we took data for specific birds from Xeno-Canto, which is a large database dedicated to sharing bird sounds from all over the world.

    We picked 4 birds that are commonly found in our area.

    • Asian Koel
    • Laughing Kookaburra
    • Square Tailed Drongo
    • Rose-ringed Parakeet
  • 3
    Building a Dataset

    We downloaded around 20-25 audio files for each bird and worked on preprocessing using software called Audacity. And then proceeded to augment the data, while infusing noise. And this helped us generate 200 files of 10 seconds each for all four of the above-mentioned labels as well as a label for the noise alone.

    We now have a balanced dataset with 33 minutes and 20 seconds of data for each label in training and 8 minutes and 20 seconds of test data.