IMU Jump Height Accuracy vs Force Plate: Research Review

Jump height is one of the most widely used performance metrics in athlete monitoring, physical testing, and training programme evaluation. For decades, force plates have served as the reference standard, calculating jump height from the impulse-momentum theorem (vertical impulse divided by body mass yields take-off velocity, from which peak height is derived). However, force plates cost USD 3,000–25,000, require laboratory installation, and cannot be used in field or game environments.

Inertial measurement unit (IMU) sensors — small devices combining accelerometers and gyroscopes — have emerged as a practical field alternative. Their adoption has been rapid, with the number of peer-reviewed validation studies increasing from 4 in 2015 to more than 40 by 2024. But not all IMU sensors perform equally. Sampling rate, sensor placement, signal processing algorithms, and calculation method all affect accuracy. This review addresses the central question coaches and sports scientists ask: how closely does an IMU measure jump height compared to a simultaneous force plate, and under what conditions is the agreement clinically meaningful?

Force plates measure ground reaction force (GRF) at approximately 1000–2000 Hz and derive jump height via one of three methods:

1. Impulse-momentum method — integrates the GRF-body-weight curve over the propulsive phase to obtain vertical impulse, then divides by body mass to obtain take-off velocity. This is considered the most accurate approach (bias ±0.2–0.5 cm vs optical motion capture).
2. Flight-time method — calculates jump height from the duration of the airborne phase using h = g × t²/8. Simple and practical but slightly underestimates true height by 0.5–1.5 cm due to body configuration differences at take-off and landing (Moir, 2008).
3. Kinematic method — tracks centre of mass displacement directly via the double-integration of acceleration. Most sensitive to integration drift over long time windows.

The impulse-momentum method is the reference against which IMU sensors are typically validated. Studies report force plate within-session reliability for CMJ height of ICC = 0.97–0.99, CV = 1.2–2.8% (Cormack et al., 2008). This exceptional reliability makes force plates the appropriate criterion measure for IMU validation work.

IMU sensors calculate jump height using one of two primary methods:

• Double integration of vertical acceleration — integrates the vertical acceleration signal twice to obtain displacement. This method mirrors the kinematic approach used in force plates. Its limitation is drift accumulation — small errors in the acceleration signal amplify during integration over the flight phase. Higher sampling rates reduce drift by capturing the acceleration signal more precisely between time steps.
• Flight-time calculation — detects the moment of take-off (acceleration drops to near zero) and landing (large acceleration spike), then applies the projectile equation h = g × t²/8. This method is less affected by drift but is sensitive to accurate event detection, which can be confounded by pre-landing knee flexion or sensor noise.

Camargo-Junior et al. (2021) compared both IMU calculation methods against simultaneous force plate measurement in 40 trained athletes performing 3 CMJ trials each. The double-integration method showed a mean bias of −0.8 cm (95% limits of agreement: −3.1 to 1.5 cm); the flight-time method showed a mean bias of +0.6 cm (95% LoA: −2.2 to 3.4 cm). Both methods were within clinically acceptable agreement for athlete monitoring purposes, defined by the authors as ±2 cm at 95% LoA.

Importantly, the choice of method interacts with sampling rate. At 200 Hz, double-integration bias increased to −2.1 cm (LoA: −5.4 to 1.2 cm), crossing the threshold of acceptable error. At 800 Hz, bias was −0.4 cm (LoA: −1.8 to 1.0 cm), meeting the ±2 cm criterion comfortably (Camargo-Junior et al., 2021).

Sampling rate is perhaps the most consequential hardware parameter for IMU jump height accuracy. The Nyquist theorem requires that the sampling rate be at least twice the highest frequency component in the signal; vertical jump acceleration signals contain meaningful components up to approximately 30–50 Hz, meaning a minimum of 100 Hz is theoretically sufficient for signal reconstruction.

In practice, however, the double-integration accumulates error at each time step, so higher sampling rates reduce the absolute size of each integration step and therefore reduce drift accumulation. The empirical evidence is consistent:

• 200 Hz IMUs: mean jump height bias −1.5 to −2.5 cm vs force plate impulse-momentum method (Picot et al., 2021).
• 400 Hz IMUs: mean bias −0.7 to −1.2 cm, LoA approximately ±2.5 cm (Picot et al., 2021).
• 800–1000 Hz IMUs: mean bias −0.3 to −0.6 cm, LoA approximately ±1.5–2.0 cm — within the smallest detectable change (SDC) for CMJ monitoring (Picot et al., 2021; Claudino et al., 2022).

A key practical implication: the smallest detectable change for CMJ height in trained athletes is approximately 1.7–2.5 cm (Cormack et al., 2008; Taylor et al., 2022). An IMU with a ±2 cm 95% LoA is therefore at the boundary of detecting true performance changes versus measurement noise. Sensors operating at 800 Hz push the LoA well below the SDC threshold, meaning real performance changes can be distinguished from measurement error.

Placement site has a meaningful effect on IMU jump height accuracy because the sensor must faithfully track the centre of mass (COM) trajectory. The further the sensor is from the COM, the more its signal is contaminated by rotational accelerations around the COM during flight.

Evidence from placement comparison studies:

• Lower back (L4–L5, sacrum): consistently shows the lowest bias against force plate (mean −0.5 to −0.8 cm). This placement is closest to the body's anatomical COM location and experiences minimal rotational displacement during a standardised jump (Picot et al., 2021).
• Hip/pelvis: mean bias −0.8 to −1.6 cm; greater inter-individual variability due to pelvic tilt differences (Camargo-Junior et al., 2021).
• Sternum/chest: mean bias −1.1 to −2.0 cm; affected by trunk flexion during the countermovement phase.
• Wrist or ankle: mean bias −2.5 to −4.0 cm; not recommended for jump height measurement (Claudino et al., 2022).

For standardized athlete monitoring, the lower back placement at L4–L5 (approximately 2 cm below the posterior iliac crest, midline) with a firm elastic strap is the evidence-based recommendation across all jump types.

The majority of IMU validation studies focus on the countermovement jump (CMJ), but evidence also exists for squat jump (SJ) and drop jump (DJ):

Countermovement Jump (CMJ)

The most-studied task. Across 23 studies reviewed by Claudino et al. (2022), IMU jump height (400–1000 Hz sensors at L4–L5) showed:

• Mean bias: −0.7 cm (range: −0.2 to −1.8 cm across studies)
• ICC against force plate: 0.95 (95% CI: 0.91–0.97)
• Within-session IMU reliability: ICC = 0.96, CV = 2.1%

Squat Jump (SJ)

SJ validation is more challenging because the absence of a countermovement reduces the range of motion, compressing the integration window. Picot et al. (2021) found mean bias −1.2 cm for SJ vs −0.5 cm for CMJ in the same sample using the same 800 Hz sensor. ICC for SJ was 0.93 — lower than CMJ but still within acceptable range for athlete monitoring.

Drop Jump (DJ)

Drop jump presents the greatest challenge for IMU validation because the brief ground contact phase (100–250 ms) requires precise event detection. Taylor et al. (2022) found that 800 Hz IMU sensors at L4–L5 showed DJ height bias of −1.1 cm (LoA: ±2.9 cm) versus force plate — acceptable for monitoring but with wider limits than CMJ. Reactive strength index (RSI) calculated from IMU flight time and contact time showed ICC = 0.94 vs force plate RSI, supporting its use in field monitoring.

IMU accuracy varies somewhat across athlete populations, primarily because differences in body composition, jump technique, and absolute jump height affect the signal-to-noise characteristics:

• Elite power athletes (rugby, basketball, volleyball): mean bias −0.4 to −0.9 cm at 800 Hz. These athletes tend to have more stereotyped jump technique, reducing movement artifact.
• Recreational athletes: mean bias −0.8 to −1.5 cm. Greater intra-individual technique variability widens the LoA.
• Youth athletes: limited data; one study (Roberts et al., 2023) found bias −1.2 cm in youth soccer players (age 14–16) — within acceptable range but with wider LoA (±3.4 cm) than adults.
• Female athletes: Balsalobre-Fernandez et al. (2019) found no significant sex difference in IMU vs force plate agreement when body mass was controlled, suggesting the validation data generalizes across sexes.

Higher absolute jump heights (>45 cm) showed slightly greater IMU bias (−1.0 cm) than lower jump heights (<30 cm, −0.4 cm), likely because greater flight duration amplifies integration drift. This finding has minimal practical consequence for athlete monitoring but should be noted in research contexts.

Based on the validation literature, practitioners using IMU sensors for jump height measurement should observe the following protocol standards to maximise agreement with force plate reference values:

1. Use a sensor sampling at 400 Hz minimum; 800 Hz preferred — this is the single most impactful hardware decision for accuracy.
2. Place the sensor at L4–L5, midline, with a firm non-elastic strap — avoid waistbands that allow vertical sensor movement during the jump.
3. Standardize the jump protocol — arm swing vs no arm swing, squat depth, and pause duration between countermovement and concentric phase all affect jump height. Pick one standard and apply it consistently within your monitoring system.
4. Average 3 trials, discard the lowest — within-session CV for CMJ at 800 Hz is ~2%, so 3 trials is sufficient to achieve a reliable mean (SEM < 1 cm).
5. Interpret change scores, not absolute values — even a well-validated IMU may have a systematic bias of ±1 cm vs force plate, but that bias is consistent over time and does not contaminate week-to-week change scores.
6. Report both jump height and IMU model/version in research settings — the validation literature is device-specific; generalizing accuracy data from one sensor to another is inappropriate.