Rep Smarter, Not Harder: AI Hypertrophy Coaching with Wearable Sensors and Edge Neural Networks

2025-12-05

AI Summary

Subjective assessment of proximity to muscular failure (i.e., Reps-in-Reserve ≤ 2) during resistance training is unreliable, risking under-stimulation or excessive fatigue. To address this, we propose a real-time, edge-deployable recognition method leveraging a single wrist-worn 6-axis IMU and a lightweight neural network. Our approach introduces a novel two-stage edge architecture: a ResNet backbone performs sliding-window segmentation, while an LSTM integrates temporal dynamics and historical movement features for classification. The model achieves end-to-end inference latencies of 112 ms on Raspberry Pi 5 and 23.5 ms on iPhone 16. Evaluated on 631 real-world bicep curl repetitions terminating at failure, it attains F1-scores of 0.83 for segment-level detection and 0.82 for near-failure classification. This work presents the first single-IMU, low-latency, high-accuracy edge solution for objective RiR ≤ 2 quantification—enabling embedded, real-time intensity feedback for hypertrophy-oriented resistance training.

Technology Category

• Machine Learning: Learning on the Edge & Model Compression
• Computer Vision: Motion & Tracking
• Planning, Routing, and Scheduling: Activity and Plan Recognition

Application Category

• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics

Abstract

Optimizing resistance training for hypertrophy requires balancing proximity to muscular failure, often quantified by Repetitions in Reserve (RiR), with fatigue management. However, subjective RiR assessment is unreliable, leading to suboptimal training stimuli or excessive fatigue. This paper introduces a novel system for real-time feedback on near-failure states (RiR $le$ 2) during resistance exercise using only a single wrist-mounted Inertial Measurement Unit (IMU). We propose a two-stage pipeline suitable for edge deployment: first, a ResNet-based model segments repetitions from the 6-axis IMU data in real-time. Second, features derived from this segmentation, alongside direct convolutional features and historical context captured by an LSTM, are used by a classification model to identify exercise windows corresponding to near-failure states. Using a newly collected dataset from 13 diverse participants performing preacher curls to failure (631 total reps), our segmentation model achieved an F1 score of 0.83, and the near-failure classifier achieved an F1 score of 0.82 under simulated real-time evaluation conditions (1.6 Hz inference rate). Deployment on a Raspberry Pi 5 yielded an average inference latency of 112 ms, and on an iPhone 16 yielded 23.5 ms, confirming the feasibility for edge computation. This work demonstrates a practical approach for objective, real-time training intensity feedback using minimal hardware, paving the way for accessible AI-driven hypertrophy coaching tools that help users manage intensity and fatigue effectively.

Problem

• Develops a real-time system to detect near-failure states during resistance training using a wrist-worn sensor.
• Addresses unreliable subjective assessment of Repetitions in Reserve (RiR) to optimize hypertrophy training.
• Enables edge-deployable AI for objective feedback on training intensity with minimal hardware requirements.

Innovation

• Real-time repetition segmentation using ResNet on IMU data
• Near-failure classification with LSTM and convolutional features
• Edge deployment on Raspberry Pi and iPhone for low latency