Bicycle safety is always a concern, but is a particularly big problem in cities where cycling is less common.Drivers in cars don’t expect to encounter cyclists and don’t know how to share the road properly.To protect themselves, cyclists need to be as predictable as possible and overt with their intentions.
Signaling is critical, but hand signals may be misinterpreted.That’s why Manivannan developed an AI-enabled smart helmet that leverages Edge Impulse and Arduino for the hands-free voice activation of turn signals.Edge Impulse makes it easy to train machine learning models and deploy them on the edge with embedded devices.
In this case, that device is an Arduino Portenta H7, which has a powerful STM32H7 microcontroller and was designed specifically for applications that rely on computer vision, machine learning, and other resource-intensive tasks. To train the Edge Impulse model, Manivannan only needed three audio sets: one of the word “left” for the left turn signal, one of the word “right” for the right turn signal, and one with examples of ambient background road noise.Road noise samples were important, because they tell the model what sound a voice command and is safe to ignore.The hardware is straightforward, with power coming from a 5V battery pack.
The turn signals are strips of LED “filament” controlled by the Portenta H7 in response to the detected voice commands.The Arduino resides inside of a plastic container for weather protection and that, along with the LED strips and battery pack, strap on to a store-bought bicycle helmet.Now Manivannan can cycle safely, signaling to others on the road without taking his hands off of the handlebars.
