For the student demo session of the Warren B. Nelms Annual IoT Conference 2025, computer engineering student David Bickram set out to design a smart wearable to address gaps in existing assistive hearing technology. The idea was to build a device that could detect and classify environmental sounds and respond instantly, privately, and reliably.
This led him to develop Safe&Sound, a smart wearable system to enhance safety, autonomy, and awareness for individuals with hearing impairments. Sounds like alarms, door knocks, car horns, and sirens are detected, classified, and transmitted to a wearable wristband, alerting the user with haptic and visual feedback.
“The goal of this project was to create an affordable, customizable, and secure assistive technology solution that improves situational awareness without relying solely on smartphones or room-installed systems,” Bickram explained.
Safe&Sound consists of two battery-powered devices. One device listens with a built-in microphone and classifies sounds locally using embedded machine learning. Because it’s not connected to Wi-Fi, no raw audio ever leaves the device and is not stored or streamed. It transmits the sound classification metadata via Bluetooth Low Energy (BLE) to a wearable second device to trigger alerts and handle cloud-based user preferences. The system’s architecture allows for low latency, improved privacy, and flexibility in how different devices respond.
Along with teammates Vivian Rincon and Natalia Rojas, Bickram demonstrated Safe&Sound’s capabilities at the Warren B. Nelms Annual IoT Conference. The demo sparked a deeper conversation about not only what the device could do, but about how it worked.
“At the conference, an industry expert asked: ‘What else could you build with this pattern?’ It turned out that the device architecture was more interesting than the device itself.”
– David Bickram
The pattern is straightforward: (1) sense and process sounds locally, (2) send metadata, and (3) real-time action. As privacy concerns grow and devices shift toward local intelligence, this pattern can be applied to countless applications to help create technology that’s more private and adaptable.
“One idea I imagine is a seamless smart home,” said Bickram. “A small wearable continuously interprets hand motion and context in real time and sends a signal of the intent detected. The environment responds with lights adjusting, music shifting, etc. No app, voice, or cloud required, but all still available if needed, while keeping data local by default.”
Pushing the boundaries of AIoT isn’t just about building new devices. Innovation happens when we rethink how technology is designed and connected, creating smarter, safer, and more adaptable systems.