RC RANDOM CHAOS

Smartwatch Ultrasonic Technology Enables Full-Body Movement Tracking

· via Hacker News

Original source

Researchers track whole body movement with just a smartwatch

Hacker News →

Cornell researchers developed EchoMotion, a system that tracks full-body movement using only a commercial smartwatch’s built-in speaker and microphone. The technology emits ultrasonic sound waves that reflect off a person’s limbs, with a lightweight deep learning model interpreting the echo patterns to determine body joint positions. This represents a significant shift from existing wearable sensors like accelerometers and gyroscopes, which only track the device itself rather than the wearer’s full body pose—a limitation that has historically required complicated multi-device setups including phones, earbuds, or VR headsets.

In testing on an unmodified Samsung Galaxy Watch 6, the system achieved positional accuracy within four inches for new users and roughly 2.4 inches after a brief 12-minute calibration. The same AI model transferred without retraining to Samsung’s Watch 7 and Google’s Pixel Watch 3, demonstrating broad compatibility. The power footprint proves remarkably efficient at around 12 milliwatts—approximately 40 times less than camera-based alternatives—consuming just 5-6% of the Galaxy Watch 6’s battery over a 30-minute session. Field testing across different environments showed the system handles ambient noise and furniture interference well, though crowded spaces with reflections from other people caused slight performance degradation.

The team is directing development toward fitness applications, where the technology could count repetitions and provide real-time form corrections. Future work includes pairing EchoMotion with smartglasses to track objects a person handles during daily tasks. The underlying vision positions this as enabling personal AI agents that could document activities throughout the day—with body tracking serving as a key component alongside other sensor inputs. Researchers anticipate performance improvements with expanded training data as the system scales from dozens to thousands of users.

Read the full article

Continue reading at Hacker News →

This is an AI-generated summary. Read the original for the full story.