Continuous Audio Recording Wearable

2019•esp32microcontrollercastingcncaudioquantified self
Back

For my final project in How To Make (Almost) Anything (opens in a new tab) was a wearable bracelet that would continously record audio, and, if triggered by a button press, store the last 5 minutes.

This device was intended to augment my memory: to capture fleeting interactions or unintentionally poetic turns of phrase. I wanted to build something that I could wear all the time.

What I Built...

I built a 2-sided ESP-32 board, that interfaces with a microSD card over 4-line SDMMC. The board records audio, sampling at 44k hertz per second, via an analog MEMS microphone and saves it in real-time.

To retrieve the audio files, the user connects to a soft-access point — a wifi network generated by the device — and can download the files over an HTTP server.

The board is 28mm x 44mm x 8mm. It is very small (the same size as Neil's hello world example board). Small enough to fit in on a wrist.

I created a 2-part mold out of machinable wax, with a third registration insert 3D printed on a PrusaMK3. I cast the mold with Mold Star 20T. And then 3D printed a cover for the bottom.

See the more details on my HTM(A)A portfolio (opens in a new tab).

Why I built it

Here are the requirements I created for myself when I started this project in 2019:

The minimum viable product has 4 features.

  1. Continuously recorded audio. The device should have enough battery to continuously record for at least 4 hours — and ideally 10.

  2. Wearable form factor. I want the device to be with me at all times.

  3. Ability to save last 3-5min from audio buffer. The device needs a button to signal when the recording should be saved. The device should have enough storage to save at least 24 hours of audio. Scaling the amount of audio that can be saved should just require buying a more expensive microSD card.

  4. (Easily) export audio files. This is a requirement. Saving audio isn't very useful if I have no way to access it. However, "easily" is in parenthesis because there are a lot of different ways to do the export. The MVP version might be as simple as access to the microSD card.

If I can get this working — and that is a big if — here are 4 of features that would be nice to have.

  1. An LED to signal the device's state. Potentially an RGB diode, so there are more colors to choose from. A rechargeable battery. It would be great if I could engineer the device in a way so that I don't need to take it apart when it runs out of batteries.
  2. Wireless export. This would be huge. Because I use a Mac, exporting files with microUSB or USB is a pain since I would need to keep track of a dongle. Wireless export can conceivably be be done with either Bluetooth Classic or over Wifi. Based on my research, I think it will be easier to create a wifi access point and a custom iOS/Mac app to facilitate the export. Apple's guidelines are quite strict when it comes to Bluetooth on iOS. Any device that doesn't use one of the default Bluetooth profiles must apply for the MFI program. No thank you!
  3. Water Resistant. The best experience with wearables is when you don't have to think about them. I don't want to worry walking in the rain and electrocuting myself — or, more importantly, breaking my bracelet!

If somehow I have managed to get all of the above working, here are some crazy ideas that would bring this device from a clunky, but useful DIY project into the realm of a consumer wearables.

  1. Entirely Wireless.
  2. Never plug a cable into the device. Recharge via induction. Export over wifi.
  3. Custom Charger. All of these stretch goals go together to create a device that is entirely wireless.