Author(s): JIAN GUANG, X., ZHANG, Q.S., Institution: SHANGHAI UNIVERSITY OF SPORT, Country: CHINA, Abstract-ID: 541

INTRODUCTION:
Velocity-Based Training (VBT) facilitates objective load prescription in resistance training. While Linear Position Transducers (LPTs) remain the "gold standard" for velocity monitoring, their high cost hinders widespread practical adoption. Low-cost Inertial Measurement Units (IMUs) offer a wireless alternative; however, they are often prone to signal noise and integration drift, particularly during ballistic movements such as Olympic weightlifting derivatives. This study aimed to validate a custom signal processing algorithm designed for a cost-effective IMU to accurately monitor Hang Power Clean (HPC) velocity.
METHODS:
Twelve resistance-trained athletes participated in a progressive HPC loading protocol (20–100% 1RM). Vertical barbell velocity was monitored concurrently by a criterion LPT (GymAware, 50 Hz) and a barbell-mounted IMU (Movella Xsens DOT, 120 Hz). Data were matched post-hoc using a magnitude-based ranking and temporal sequencing algorithm to align valid LPT entries with IMU-detected peaks. A custom MATLAB script utilized quaternions to isolate global vertical acceleration, which was processed via a 4th-order zero-phase Butterworth low-pass filter (cutoff: 12 Hz) to mitigate noise and converted to velocity using trapezoidal integration with linear detrending. Validity and reliability were assessed on a pooled dataset of 174 repetitions using Pearson’s r, Intraclass Correlation Coefficient (ICC), bias, Coefficient of Variation (CV%), and Standard Error of Measurement (SEM).
RESULTS:
Both Peak Velocity (PV) and Mean Velocity (MV) demonstrated strong linearity (r > 0.80) and acceptable agreement (ICC > 0.745) between devices. Regarding measurement error, systematic biases of -0.091 m/s and -0.070 m/s were observed for PV and MV respectively, with Standard Error of Measurement (SEM) values of 0.142 m/s and 0.087 m/s. Furthermore, the system exhibited high intra-trial reliability across the load spectrum, with low coefficients of variation (CV%) for both PV (5.8%) and MV (5.7%).
CONCLUSION:
The custom 120 Hz signal processing algorithm effectively tracks ballistic weightlifting movements using consumer-grade IMUs. Although the raw algorithm results in consistent velocity underestimation, it maintains high linearity and low variability. Therefore, this system provides a valid, cost-effective solution for monitoring general HPC training loads and velocity trends, offering a practical alternative where "gold standard" precision is inaccessible.