Extended Abstract
1.    Introduction
In recent years, velocity-based training (VBT) has emerged as an alternative to traditional percentage-based loading, using barbell velocity as a real-time indicator of training intensity and neuromuscular fatigue (3). A nearly linear relationship between load and movement velocity allows for strength estimation without maximal testing (4). Metrics such as mean concentric velocity, peak velocity, and velocity loss are commonly used to monitor fatigue and regulate training volume (2,5).
Linear position transducers (LPTs) provide high accuracy but are expensive and less practical in field settings due to their cable-based design (8). In contrast, inertial measurement units (IMUs) are lightweight, affordable, and portable, capable of capturing acceleration, angular velocity, and orientation, making them a promising alternative (9,10). Previous research has supported the validity of wearable sensors in movements like the squat (11,12), but evidence for more complex free-weight exercises such as the deadlift is limited and inconsistent (13,14). Reduced accuracy, particularly for peak velocity under high loads, has been reported, along with significant differences compared to motion capture systems (14–16). Given their potential for practical field use (17), this study aimed to examine the validity of a low-cost IMU for measuring barbell velocity during the concentric phase of the deadlift, using motion capture as the reference standard.
2.    Methods
This study involved 16 healthy young men (mean age: 21.7 ± 3.4 years; height: 170.7 ± 7.3 cm; body mass: 69.3 ± 11.1 kg), each with at least one year of resistance training experience. Sample size was determined a priori using G*Power (version 3.1.9.6) with α = 0.05, power = 0.95, and an expected ICC of 0.75. All participants were free of musculoskeletal or neuromuscular disorders and provided written informed consent. The study was approved by the Ethics Committee of Yazd University (IR.YAZD.REC.1403.078) and conducted in accordance with the Declaration of Helsinki.
All procedures were completed in a single laboratory session, including familiarization, warm-up, and testing. The warm-up consisted of five minutes of dynamic exercises, followed by two sets of deadlifts (10 reps at 30% body weight and 5 reps at 50%). The main protocol involved three sets of five deadlifts at 50% body weight, performed at a self-selected pace. Test order was counterbalanced, with three minutes rest between sets.
Barbell velocity was simultaneously recorded using a triaxial IMU (BSNlab, Iran) and a 3D motion capture system (OptiTrack Dou120, USA) at 120 Hz. The IMU’s default 250 Hz sampling rate was reconfigured to 120 Hz to match the motion capture system. The sensor was mounted at the barbell’s center, with a reflective marker attached for synchronization.
Motion capture data were reconstructed (if needed) and filtered (6 Hz Butterworth). IMU accelerometer data were low-pass filtered (25 Hz), fused using the Madgwick algorithm to remove gravity, and integrated after applying Zero Velocity Update at the start of each concentric phase to reduce drift. Instantaneous vertical velocity was derived by trapezoidal integration.
The concentric phase was segmented from displacement data, and the third and fourth repetitions of each set were analyzed. Mean and peak velocities were extracted. Normality was assessed using the Shapiro–Wilk test, and differences between tools and conditions (raw vs filtered) were examined using repeated-measures ANOVA with Bonferroni correction (p < 0.05). Agreement was evaluated using mean bias, CV, RMSE, correlation coefficients, and Bland–Altman analysis. All processing and statistics were performed in MATLAB R2021b.
3.    Results
The two-way repeated-measures ANOVA revealed no significant main effects of measurement tool (IMU vs. MoCap), filtering (raw vs. filtered), or their interaction for mean velocity, indicating that both systems provided comparable results across processing conditions. 
In contrast, for peak velocity, significant main effects of both tool and filtering, as well as a significant interaction (p < 0.05), were observed. This suggests that peak velocity values were influenced by both the measurement system and the application of filtering.
Table 1 presents the Bland–Altman and ICC analyses. For mean velocity, the mean bias between IMU and MoCap was small (≈ ±0.02 m/s), with narrow 95% confidence intervals, high ICC values (0.86–0.87), and low error indices (RMSE ≈ 0.16 m/s; CV ≈ 14%). These findings indicate excellent agreement between the two measurement systems for this variable.

For peak velocity, the mean bias was larger (−0.18 to −0.07 m/s), ICC values were moderate (0.59–0.67), and error indices were higher (RMSE ≈ 0.47–0.51 m/s; CV > 22%). These results reflect greater variability and a systematic underestimation of peak velocity by the IMU compared to MoCap. 
4.    Discussion
This study examined barbell velocity during the concentric phase of the deadlift. The IMU demonstrated a strong correlation with the motion capture system for mean velocity and a moderate correlation for peak velocity, indicating good validity for mean values but reduced accuracy for peak values. Bland–Altman analysis revealed minimal bias and narrow limits of agreement for mean velocity, whereas larger bias and wider limits were observed for peak velocity. Reliability indices (rpc and SEM) further indicated greater stability for mean compared with peak measurements. With a relative error of approximately 14%, the IMU appears sufficiently accurate for detecting meaningful changes in mean velocity but is less suitable for precise assessments of peak velocity.
These findings are consistent with previous studies. McGrath et al. (13) confirmed the validity of accelerometer-based devices during the deadlift, and Cuverel-Ibáñez et al. (19) also reported high correlations for mean velocity under moderate loads. Conversely, measurement accuracy tends to decrease for peak velocity during rapid or high-amplitude movements (9, 10, 14), which aligns with the present results. Several methodological factors—such as sensor placement, sampling rate, and filtering approach—can also influence measurement accuracy (10, 11, 22). In this study, using raw data, precise bar-mounted placement, and a 120 Hz sampling rate, we observed small biases but relatively wide limits of agreement in certain trials, similar to those reported by Sato et al. (12). Overall, IMUs provide a valid and practical alternative to motion capture systems for field-based assessment of barbell velocity (17, 18, 20). Nevertheless, caution should be exercised when interpreting peak velocity values due to their higher variability and lower agreement with reference systems (21, 23).

Ethical Considerations
Compliance with ethical guidelines
All ethical considerations were fully observed in this study. Participants were informed about the research procedures in detail and provided written consent prior to participation. They were also assured of their right to withdraw from the study at any time without penalty. Furthermore, all personal information was kept strictly confidential and used solely for research purposes.
Funding
This research did not receive any grants from funding agencies in the public, commercial, or non-profit sectors. 
Authors' contributions
The author was involved in all stages of the study, including design, implementation, and manuscript preparation.
Conflicts of interest
The author declares that there is no conflict of interest related to this study.