GOLDEN REP

Train smarter with Golden Rep. A wearable app for Huawei watches that compares every lift to your "perfect rep" using real-time sensors and AI-assisted performance analysis to improve your workout.

Comment

Inspiration

We’ve all been there: you’re at the gym, trying to push your limits, but as you get tired, your form slips. Bad form doesn't just mean fewer gains; it leads to injuries. Personal trainers are great, but they are expensive and can't be with you for every single rep. We wanted to digitize the experience of having a pro trainer who knows exactly how you move. We were inspired to build Golden Rep, a solution that captures your perfect movement when you are fresh and uses it as the "Gold Standard" to guide you through the rest of your workout.

What it does

Golden Rep is a HarmonyOS wearable application designed for Huawei smartwatches that acts as an intelligent, real-time AI weightlifting coach.

  • The "Golden Rep" Baseline: The user records one perfect repetition (the "Golden Rep"). The app records the precise tempo, rotation, and acceleration of this movement via the sensors of the Huawei watch.
  • Intelligent Monitoring: During your set, the watch compares every subsequent rep against that Golden Rep in real-time using Sensor Fusion (Accelerometer, Rotation Vector, and Heart Rate).
  • Real-Time Feedback: It provides instant auditory motivational feedback if you are moving too fast, too slow, breaking form, or your heart rate is too high. (e.g., "Form is slipping, stay focused" or "Pick up the pace a little").
  • Zero-Touch Interface: During training, your hands are busy holding weights. The app uses motion thresholds to automatically detect when you are holding still to calibrate, when you start moving, and when a repetition is complete, without requiring you to touch the screen.
  • AI Coaching Insights: While the watch locally detects deviations from your "Golden Rep", an integrated LLM analyzes the trend of these failures (e.g., spotting fatigue vs. instability) to provide one actionable, human-like tip for your next set.

How we built it

We built Golden Rep using TypeScript and ArkTS for HarmonyOS, leveraging the Huawei Watch's sensor stack.

  • Sensor Fusion: We utilized the @ohos.sensor API to pull high-frequency data (20ms intervals) from the Accelerometer and Rotation Vector sensors, along with Heart Rate monitoring.
  • Haptics: We used @ohos.vibrator to give physical cues like a short pulse for "Start" and a longer vibration for "Success", so the user doesn't have to look at their wrist while lifting.-
  • Hybrid AI Architecture: We implemented a hybrid approach where critical feedback (rep counting, form checks) happens locally on the watch using ArkTS for zero-latency, while complex pattern recognition is offloaded to an LLM via REST API to generate natural language coaching tips.

Challenges we ran into

  • "Return to Start" Problem: Detecting when a rep is finished is harder than we thought. We had to combine "Motion Energy" (acceleration magnitude) with "Orientation Accuracy" (Quaternion angle error < 25 degrees) to ensure the user had actually returned to the starting position before counting the rep.

Accomplishments that we're proud of

  • The Math: Successfully implementing Quaternion rotation tracking to calculate the precise angular error of a rep compared to the baseline.
  • Feedback Loop: Creating a system that doesn't just record data, but actively "talks back" to the user through haptics and specific coaching advice based on their performance metrics. -Wearable: It was the first time any of us built software for wearable devices, and we were able to get the watch sensors, real-time data streaming, and on-device processing all working smoothly.

What we learned

  • HarmonyOS & ArkTS: We gained deep experience building for wearables using ArkTS, from managing the OpenHarmony sensor stack and raw buffers to designing effective UI layouts for the constraints of a small wearable screen. We also learned how to work with and analyze time series data by applying the Dynamic Time Wrapping (DTW) Algorithm.

What's next for GOLDEN REP

  • Advanced AI Coach: Expanding the post-workout AI summary to provide deeper insights, such as recommending weight adjustments based on, e.g., heart rate stability and tempo consistency.
  • Cloud Sync: Saving the "Golden Rep" profiles to the cloud so users can track their form improvement over months, not just one workout.

Built With

  • arkts
  • arkui
  • devecostudio
  • huaweiwatch5
  • oniro
  • openai
  • openharmonysdk
Share this project:

Updates