I built an AI/IoT prototype that uses a Xiao ESP32 microcontroller and a millimeter-wave radar module to measure heart rate non-contact. It visualizes and sonifies your heartbeat—making your “dokidoki” (heart pounding) visible and audible—through LED blinking synced to your pulse and vibrations from a motor I hacked (repurposed from something like a metal detector’s vibration unit).
It also analyzes time-series heart rate data with AI to generate messages that reflect your emotional state in the moment.
By making your heartbeat visible and tangible, it helps people share feelings with each other and can ease tension in situations like meeting someone new or romantic moments. The AI-based “dokidoki” analysis might even help you notice subtle emotional changes you weren’t aware of.
Demonstrationa) Dokidoki detection and sharing in action
b) Examples of messages sent (AI analysis output: LINE and email notifications)
・LINE Notification
・E-Mail Notification
Press the measurement button, and the AI analyzes changes in your heart rate (“dokidoki”) over 10 seconds. Based on the results, you’ll receive encouraging, push-you-forward messages via email or LINE.
Motivation for the Project(i) People often perceive me as a calm and composed person based on my appearance, but in reality, I frequently feel my heart racing inside. I sometimes sense a gap between how I appear and how I actually feel. It made me wonder if this kind of “hidden dokidoki” isn’t just unique to me, but something others experience too, in varying degrees. I wanted to create a way for people to share these feelings, which inspired me to build this prototype.
(ii) There’s another reason as well. In those moments when I feel nervous and hesitate, I sometimes wish for something that could gently give me a little push forward. At the same time, when someone else is hesitating, I want to be able to offer them that same small bit of courage. I wanted to give form to that “moment that helps you take the first step, ” which led me to create this dokidoki detection device.
System ConfigurationA millimeter-wave radar module measures heartbeats non-contact, and the data is processed by a Xiao ESP32 microcontroller. Synchronized with the heartbeat signal, a vibration motor—hacked and repurposed from a metal detector—is driven in sync to “sonify” the heartbeat, while a connected LED blinks in sync to “visualize” it.
In addition, real-time heart rate data is displayed on an LCD for monitoring. The system is powered by a built-in 9V battery from the metal detector, with a DC-DC converter generating 5V for the microcontroller.
When the button is pressed, the most recent 10 seconds of heart rate data are sent to a generative AI model for “dokidoki” evaluation. The prompt used for the evaluation includes instructions to return messages that gently “give you a little push forward.”
The application is built using SORACOM Flux, enabling a low-code implementation.
Open the metal detector case and solder a harness to the 9V battery terminal, then connect it to a DC-DC converter. Drill mounting holes with an electric drill and secure the converter with screws. Also make a hole to route the harness outside and pull the wiring through.
Next, connect the millimeter-wave radar module (including Xiao ESP32), LCD, switch, and LED according to the wiring diagram, and install them into the metal detector.
Initially, the circuit was built on a breadboard, but repeated attachment and removal during measurements caused the harness to come loose easily. To solve this, a custom PCB was created and the design was changed to a connector-based connection method.
- Magnetic detachable mechanism for flexible LCD orientationTo allow the LCD to be viewed both when measuring your own heartbeat and when measuring someone else’s, a magnetic attachment mechanism was used so the LCD can be easily attached and detached as needed.
- A prototype “dokidoki” detector was developed that visualizes and sonifies heartbeat changes, and uses AI to analyze those changes and send messages via LINE and email.
- A physical AI prototype was created that non-contactly measures biometric data (heart rate) using an mmWave radar sensor, integrates with generative AI through SORACOM Flux in a low-code manner, and feeds the results back through LEDs, a vibration motor, and message notifications.
This project uses Japan’s special provision for the temporary experimental use of non-certified wireless devices (up to 180 days).








Comments