05.05.2026

Invisible monitoring is used to collect and assess athlete-related data with minimal disruption to practice. This report categorizes technologies used for invisible monitoring into four broad categories: (1) positional, (2) kinetic, (3) kinematic, and (4) physiological tracking systems. It summarizes data-collection requirements, typical outputs, validity and reliability, athlete burden, and practical relevance for workflows involving invisible monitoring.

This report is an output of Invisible Monitoring in the Development of Well-being and Performance project – a collaborative research project coordinated by Turku University of Applied Sciences and funded by Business Finland (7507/31/2024). The project scrutinizes both technology and practices of invisible monitoring primarily in the context of sports and physical activity and interprets and translates the findings also for other areas of application where human well-being and performance are important.

List of abbreviations

2D – Two-dimensional
3D – Three-dimensional
CGM – Continuous glucose monitor
COD – Change of direction
COP – Center of pressure
CV – Coefficient of variation
ECG – Electrocardiography
EMG – Electromyography
GNSS – Global navigation satellite system
GPS – Global positioning system
GRF – Ground reaction force
HR – Heart rate
HRV – Heart rate variability
Hz – Hertz
ICC – Intraclass correlation coefficient
IMU – Inertial measurement unit
IMTP – Isometric mid-thigh pull
LPS – Local positioning system(s)
LPT – Linear position transducer
MAPE – Mean absolute percentage error
MPV – Mean propulsive velocity
NIRS – Near-infrared spectroscopy
pH – Potential of hydrogen
PPG – Photoplethysmography
RFD – Rate of force development
RMS – Root mean square
RMSE – Root mean square error
RMSSD – Root mean square of successive differences
SmO₂ – Muscle oxygen saturation
TRIMP – Training impulse
USG – Urine specific gravity
VBT – Velocity-based training
vGRF – Vertical ground reaction force

 

0 Introduction

Invisible monitoring, defined as collecting and assessing athlete status-related data with minimal disruption to practice, has become a key concept in contemporary athlete monitoring (Leduc & Weaving 2025). Monitoring in field environments has significantly advanced through improvements in wearable and tracking technologies (Seçkin et al., 2023). Compared to traditional laboratory systems, modern devices are small, lightweight, and designed for use in-play, resulting in minimal interruption to training (Adesida et al., 2019; Camomilla et al., 2018). Such systems permit objective quantification of information relevant to the decision domains of monitoring and generate an evidence base to guide sports practitioners’ decision-making (Seshadri et al., 2019).

A framework introduced by Leduc and Weaving (2025) conceptualizes invisible monitoring along continuums of measurement burden (obtrusiveness and frequency) and construct coverage (the range of constructs a tool captures). Invisible monitoring thus includes both passive and in-play data collection (Leduc & Weaving 2025), where devices and systems gather data without additional actions taken by athletes (e.g., global positioning systems; GPS), and low touch scheduled tests integrated into existing routines (e.g., countermovement jump with force plates) that minimally disrupt practice. This approach contrasts with higher burden methods (Camomilla et al., 2018) (e.g., frequent athlete self-report questionnaires or standalone laboratory tests) and emphasizes the practical benefit of reducing athlete and staff load (Seshadri et al., 2019) while increasing the continuity and ecological validity of measurement.

The utility of invisible monitoring relative to traditional higher burden methods arises from two main features. Firstly, it enables continuous observation during training sessions and matches, rather than only during separate testing (Camomilla et al., 2018; Seshadri et al., 2019). This enhances ecological validity and data-collection efficiency, while also improving athlete compliance and ensuring more complete data throughout the season (Camomilla et al., 2018; Seshadri et al., 2019). Critically, the lower measurement burden enables more frequent observations, supporting stronger intrapersonal athlete monitoring and inference, in addition to aligning more closely with decision making timelines in applied sport (Leduc & Weaving, 2025). For example, richer data allows coaches to evaluate trends over time to assess recovery trajectories and load-response patterns rather than relying on isolated snapshots (Windt et al., 2020).

Secondly, because lower burden capture can be conducted concurrently across multiple domains (McLean et al., 2025; Cossich et al., 2023) (e.g., training and competition load, athlete status, and performance), it supports a multisource perspective. This aligns with systems and complexity thinking in sport (McLean et al., 2025), which cautions against reductionist interpretations of single metrics in isolation. Instead, it encourages the integration of diverse interacting data streams to reflect how performance, fatigue, and injury risk emerge from multiple determinants (McLean et al., 2025; Cossich et al., 2023).

At the same time, before integrating new technologies into the training system, measurement quality and practical limitations of the technologies need to be assessed (Camomilla et al., 2018; Leduc & Weaving, 2025). Aspects of scrutiny include measurement validity and reliability, relevance of the information obtained, and the burden of data collection (Camomilla et al., 2018; Leduc & Weaving 2025). Ultimately, the benefits of the information gained and enhanced decision-making capabilities must outweigh the costs of data collection and management to justify the use of the technology (Camomilla et al., 2018; Seshadri et al., 2019; Leduc et al., 2025; Exel & Dabnichki, 2024).

The purpose of this report is to map current technologies suitable for invisible or low-burden monitoring within the categories of positional, kinetic, kinematic, and physiological tracking systems, and to evaluate them with respect to measurement burden, informational yield, and measurement quality.

 

1 Concept and taxonomy

Invisible monitoring refers to the collection of data with minimal disruption to athletes’ daily practices (Leduc & Weaving, 2025). Technologies that may be used for invisible monitoring have been grouped into the following broad categories adapted from existing athlete monitoring frameworks (e.g., Halson 2014; Bourdon et al., 2017).

• Positional tracking systems: time-resolved coordinates and derived locomotor metrics.
• Kinetic tracking systems: Forces measured from implements or the athlete’s ground contacts.
• Kinematic tracking systems: Motion measured from implements or the athlete’s body segments.
• Physiological tracking systems: Measurements related to the athletes’ internal systems or processes.

Each section details:(i) data collection and athlete burden, (ii) measurements and metrics, (iii) validity, (iv) reliability, and (v) relevance to invisible monitoring.

 

2 Positional tracking systems

Athlete position tracking systems are employed to measure athletes’ location and movement trajectories during training and competition with relatively low added burden, thus forming a central component of external load monitoring in many professional environments (Hoppe et al., 2018; Linke et al., 2018). These systems provide time-resolved coordinates for each athlete, including wearable global positioning system (GPS) or global navigation satellite system (GNSS) units, local positioning systems (LPS), optical multicamera video tracking systems, and computer-vision–based markerless systems (Felipe et al., 2019; Müller et al., 2022; Hoppe et al., 2018; Linke et al., 2018). Inertial-based devices may estimate short-term trajectories via dead-reckoning (Müller et al., 2022) or have the potential to be integrated within GPS units to refine velocity estimates (Gómez-Carmona et al., 2020).

2.0 GPS and GNSS Wearable Units

Relevance to invisible monitoring. Commonly used for external load monitoring in various field sports with outdoor and open-air venues.

Burden and context. Minimal but nonzero burden due to wearing small receivers between the scapulae or other unobtrusive locations.

Measurements and metrics. Timestamped positions, velocities, distances by speed zones, peak velocities, acceleration and deceleration counts. Acceleration-speed profiles derived from game and practice session motion data (Calderón-Pellegrino et al., 2022; Linke et al., 2018; Sánchez-Sáez et al., 2021; Clavel et al., 2023; Cormier et al., 2023).

Validity and reliability. Functions optimally in outdoor open-air conditions, as validity depends on satellite visibility. Signal quality decreases in covered stadiums or urban canyons. In comparative field testing against motion capture and optical or LPS systems, GPS often exhibited larger positional and highspeed running errors than LPS and engineered optical tracking, particularly during rapid changes of direction and sprints (Linke et al., 2018). League validations have shown acceptable agreement for distance-based metrics under favorable environmental conditions. Reliability in modern high-sampling units shows good repeatability in standard metrics across sessions when environmental conditions remain stable. Variability between units is typically higher than in venue fixed LPS systems, so cross-device comparisons should be avoided unless units are carefully batch-matched and protocols are standardized (Hoppe et al., 2018; Clavel et al., 2023; Sánchez-Sáez et al., 2021; Linke et al., 2018).

Implementation notes. Prefer open-air environments; avoid covered stadiums and urban canyons. Do not mix units across athletes or sessions; between-unit variability can exceed within-unit error. Standardize firmware/sampling and maintain antenna orientation and vest placement to improve repeatability (Linke et al., 2018).

2.1 Local Positioning Systems

Relevance to invisible monitoring. External load and tactical monitoring in indoor venues without reliance on satellite signals (Hoppe et al., 2018; Linke et al., 2018).

Burden and context. Small tags are worn by athletes. Requires instrumented venues with anchors or beacons mounted around the field of play. Athlete burden is similar to GPS vests, while organizational burden is higher due to installation, calibration, and maintenance (Hoppe et al., 2018; Müller et al., 2022).

Measurements and metrics. Timestamped positions, velocities, distances by speed zones, peak velocities, acceleration and deceleration counts (i.e. same as with GPS/GNSS) (Linke et al., 2018; Müller et al., 2022; Hoppe et al., 2018).

Validity and reliability. A 20 Hz ultra-wideband LPS has shown lower typical error than 10–18 Hz GPS for distance and sprint related mechanical variables, particularly in multidirectional drills. Compared with motion capture, LPS often exhibits the smallest positional errors when evaluated against GPS and optical video during high-speed running (Linke et al., 2018). Between device coefficient of variation (CV) for distance and sprint variables are generally low, indicating high between unit reliability. Whereas cross venue comparability depends on antenna layouts and calibration (Linke et al., 2018; Müller et al., 2022; Hoppe et al., 2018).

2.2 Optical Multicamera Video Systems

Relevance to invisible monitoring. Permits external load and tactical monitoring without wearable devices. The systems’ complexities stem from the necessary infrastructure and processing pipelines.

Burden and context. Athlete burden is negligible as data collection occurs via stadium-wide camera arrays. However, the initial infrastructure and staffing demands are substantial initially. Once the systems are in place, automated data collection and processing substantially reduce the burden on practitioners (Felipe, 2019; Linke et al., 2018; Linke et al., 2020).

Measurements and metrics. Movement trajectories, distances, speeds, and accelerations are similar to those recorded by GPS/GNSS and LPS systems. Inter-player distances, which can be synchronized with event tagging, provide physical–technical information. Team level spatiotemporal variables, such as centroids, surface area, synchrony, and pitch control, are derivable and linked to match outcomes (Beato et al., 2018; Ju et al., 2022; Duarte et al., 2013; Frencken et al., 2011; Martens et al., 2021).

Validity and reliability. Errors below 1% have been reported in controlled validations. Football-specific validations indicate small biases and acceptable agreement for position, speed, and distance. A video-based performance analysis system versus GPS showed trivial to slight differences in distance metrics in soccer. For acceleration and deceleration counts, optical systems demonstrate strong agreement with GPS and inertial measurement unit (IMU) integrated systems for most thresholds but tend to underestimate the highest intensity events. Very high to near-perfect agreement for inter- and intra-observer reliability for coding passes, shots, and other technical-tactical events have been observed with video-based systems. Team tactics-related data, such as team centroids, team surface area, inter-team synchrony, and pitch-control metrics, have been accurately computed (Duarte et al., 2013; Frencken et al., 2011; Mara et al., 2017; Felipe et al., 2019; Beato 2018; Ju 2022; Linke et al., 2018; Linke et al., 2020; González-Rodenas 2025; Pons 2021).

 

3 Kinetic tracking systems

Kinetic tracking systems quantify the forces, impulses, and neuromuscular characteristics underlying athletic performance. These technologies extend force assessment beyond laboratory environments, enabling routine assessment of strength, asymmetries, and mechanical readiness with minimal intrusion. Core modalities include portable force plates, eccentric hamstring testing devices, pressure- and force-estimating insoles, and a range of instrumented sport-specific systems, each offering different trade-offs in accuracy, ecological validity, and operational burden (Collings et al., 2024; Mylonas et al., 2023; Wang et al., 2025; Bishop et al., 2022; Renner et al., 2019; Davidson et al., 2025; Bonito et al., 2022; Thorwartl et al., 2022).

3.0 Portable Force Plates

Relevance to invisible monitoring. Enables regular monitoring of physical performance through integration into training environments and routines.

Burden and context. Lightweight and wireless force plates enable rapid setup. Athletes can perform assessment tasks as part of warm-up and pre-training routines (Collings et al., 2024).

Measurements and metrics. Force and impulse outputs are represented as time-series data. Derived measures from force, velocity and displacement allow for the calculation of performance-related metrics such as jump height, reactive strength index, rate of force development, and concentric and eccentric peak power (Collings et al., 2024; Mylonas et al., 2023; Bishop et al., 2022; Wang et al., 2025).

Validity and reliability. Strong validity has been reported against mounted laboratory force plates for jumps, drop landings, and isometric mid-thigh pull peak force with small biases and strong correlations. Reliable measures for jump peak force, jump height, and force–time and isometric mid-thigh pull (IMTP) peak force and impulse metrics have been reported. The between day reliability for IMTP peak force and impulse was also found to be excellent (Collings et al., 2024; Mylonas et al., 2023; Wang et al., 2025).

3.1 Force Estimating and Pressure Insoles

Relevance to invisible monitoring. Hidden in the shoe, pressure insoles enable near invisible kinetic monitoring during walking, running, and functional tasks. Extension to high-intensity cutting and jumping shows promise but is less well documented (Cudejko et al., 2023; Davidson et al., 2025; Hajizadeh et al., 2023; Renner et al., 2019).

Burden and context. Thin insoles in standard footwear cause minimal perceptual burden. Most insole systems incorporate arrays of capacitive or resistive pressure sensors embedded in a thin insole that can be positioned inside standard shoes (Cudejko et al., 2023; Davidson et al., 2025; Hajizadeh et al., 2023; Renner et al., 2019).

Measurements and metrics. Plantar pressure maps, vertical ground reaction force (GRF) estimates, center of pressure (COP) trajectories, and ground contact times.

Validity and reliability. Insoles evaluated in literature have demonstrated good to excellent agreement for vertical GRF and COP during walking and running. Better agreement with a force plate instrumented treadmill impulse loading rate during walking and running has been reported with higher sampling frequencies (100 vs. 200 Hz). Wireless multi-sensor insoles can estimate COP and GRF with minimal error during functional tasks. Three-segment force-estimating insoles with drift correction reduce the root mean square error (RMSE) of GRF and enhance COP agreement. Pressure-only insoles may estimate absolute force magnitudes with greater error but are valid for timing and relative change metrics. Reliability is good for COP and timing. Force metrics reliable with drift correction and standardized placement/footwear; reported CVs for peak force are approximately 5% in controlled gait. Fit, footwear model, sampling rate, and calibration procedures have been shown to impact data quality (Davidson et al., 2025; Cudejko 2023; Hajizadeh et al., 2023; Renner et al., 2019).  

3.2 Instrumented Sports and Training Equipment

Relevance to invisible monitoring. Ecologically valid, offering sport and training specific insights. Technologies are highly variable, and integration is specific to the technology used (Bonito et al., 2022; Thorwartl et al., 2022; Bouillod et al., 2017; Bishop et al., 2022).

Burden and context. Sensors integrated into implements or power meters, which preserve natural movement. For physical training, specific technologies and exercises used for measurement can be integrated into strength and conditioning plans to minimize the testing burden.

Measurements and metrics. Metrics are specific to the implements and technologies, but may include force-time profiles, implement deformation and external mechanical work, such as power and work (Bonito et al., 2022; Thorwartl et al., 2022; Bouillod et al., 2017; Bishop et al., 2022).

Validity and reliability. Accurate force profiling and deformation sensing are demonstrated in kayaking and skiing prototypes. Commercial cycling power meters exhibit device-dependent biases but maintain acceptable validity for measuring power output. For sports implements, reliability is generally high within devices across repeated trials and days. For training equipment measurements, between-session reliability has been reported as moderate with low typical error (coefficient of variation (CV) typically <10%), though it requires familiarization sessions prior to measurements for monitoring purposes (Bonito et al., 2022; Thorwartl et al., 2022; Bouillod et al., 2017; Bishop et al., 2022).

 

4 Kinematic tracking systems

Optical kinematic tracking systems use camera-based technologies to quantify full-body or segmental movement during sport, rehabilitation, and performance assessment (Bonnechère et al., 2014). These systems traditionally rely on marker-based stereophotogrammetry, but recent advances in computer vision and deep learning have accelerated the development and validation of marker-less optical systems capable of producing biomechanical estimates (Nakano et al., 2020; Turner et al., 2024; Fukushima et al., 2024). Although optical solutions vary considerably in camera configuration, feature extraction, and biomechanical modelling, they share a common objective: to enable accurate, non-contact motion tracking for real-world sport environments (Cronin et al., 2024; Federolf et al., 2025).

Wearable kinematic tracking systems rely on body-mounted sensors, most commonly inertial measurement units (IMUs), linear position transducers (LPTs), accelerometers, gyroscopes, magnetometers, and specialized embedded inertial technologies to quantify joint kinematics, segmental motion, movement velocities, and skill-specific performance metrics (Teufl et al., 2019; Dahl et al., 2020). These technologies are increasingly employed for in-field biomechanical monitoring due to their portability, affordability, and operability in environments where optical systems are impractical (James et al., 2011; Born et al., 2024). Although the degree of athlete burden varies between devices, wearable systems generally require minimal setup and allow continuous monitoring during training and competition.

4.0 Segment Mounted IMUs

Relevance to invisible monitoring. Imparts a low-burden, low-cost, easy-to-use device that is flexible across tasks and has been shown to be useful in velocity and displacement measurements.

Burden and context. Small sensors on trunk/limbs are used for assessment. Brief calibration is required, and sampling typically occurs between 100–1000 Hz, with streaming/logging to mobile or base stations (Teufl et al., 2019).

Measurements and metrics. 3D acceleration and angular velocity, orientation, joint angles with multi IMU setups, jump counts and heights, event timing, center of mass surrogates, and postural adjustment indices (Teufl et al., 2019).

Validity and reliability. Segment-mounted IMUs demonstrate acceptable agreement with optical motion capture for several lower-limb joint kinematics during both static and dynamic tasks, with greater accuracy typically observed in the sagittal plane and larger errors in the frontal and transverse planes, particularly as movement speed and task complexity increase (Teufl et al., 2019). Trunk-mounted inertial sensors provide valid estimates of specific temporal and kinematic outcomes during landing and stabilization tasks when compared with laboratory reference measures (Gallina et al., 2025). Reliability is generally good for joint angle measures and trunk-based metrics under controlled protocols, but declines with movement complexity, multiplanar movements and is sensitive to sensor placement and calibration procedures (Teufl et al., 2019; Gallina et al., 2025).

Implementation notes. Accuracy depends strongly on placement, calibration, and fusion algorithms (Horsak et al., 2023; Teufl et al., 2019).

4.1 Linear Position Transducers (LPTs)

Relevance to invisible monitoring. Under standard conditions, LPTs provide low-burden, high-fidelity, velocity-based training (VBT) assessment data.

Burden and context. It is attached with a cable tether to a bar/belt, allowing for a quick gym setup providing high-frequency displacement signals with clear mechanical interpretation (Tomasevicz et al., 2020).

Measurements and metrics. Bar displacement/velocity, mean propulsive velocity (MPV), peak velocity, velocity loss, and jump height proxies derived from displacement time can be provided (Tomasevicz et al., 2020).

Validity and reliability. The GymAware device overestimates jump height by several centimeters compared to the criterion (Pueo et al., 2020), whereas Vitruve shows good validity for bar velocity across loads; indicating that validity is device and exercise dependent with varying biases/limits of agreement (Tomasevicz et al., 2020). Within device reliability is high, supporting longitudinal tracking when protocols are consistent (Pueo et al., 2020).

Implementation notes. Suitable for structured profiling and real-time autoregulation via velocity feedback.

4.2 Smartphone Inertial Sensing

Relevance to invisible monitoring. Smartphones offer a pragmatic entry point where IMUs are unavailable (Gallina et al., 2025).

Burden and context. A smartphone positioned on or near participants’ chests, meaning minimal new hardware is needed (Gallina et al., 2025).

Measurements and metrics. Time to stabilization, flight time, trunk orientation, cadence, and ground contact can be assessed (Gallina et al., 2025).

Validity and reliability. For single-leg landing stabilization, smartphones align well with reference devices for temporal outcomes such as time-to-stabilization and flight time, although acceleration-derived measurements tend to produce larger discrepancies. Within-session repeatability is acceptable, but multiday test–retest reliability has not been fully established. Consistent device placement and orientation are crucial for obtaining reliable and accurate measurements. (Gallina et al., 2025)

4.3 Instrumented Mouthguards

Relevance to invisible monitoring. Aligns with invisible monitoring for head impact exposure; clinical translation remains under active research (Kuo et al., 2018).

Burden and context. Used in collision sports where mouthguards are standard, thus introducing minimal extra burden (Jones et al., 2022).

Measurements and metrics. Head linear and angular acceleration, impact counts, and derived brain strain indices (Liu et al., 2022).

Validity and reliability. Laboratory accuracy can be high, although field performance depends on fit and algorithms. Repeated impact precision is considered strong within models, but video review remains essential for context (Kuo et al., 2018; Liu et al., 2022).

Implementation notes. Prioritize fit and retention; different instrumented mouthguard models and algorithms produce different outputs. Pair with synchronized video for event verification. Establishing acceptance thresholds for spurious impacts should be done before reporting aggregates (Jones et al., 2022).

4.4 Video-Based Barbell Velocity

Relevance to invisible monitoring. Low burden VBT when hardware attachment is impractical.

Burden and context.  Uses camera-based tracking without physical attachment to the bar or athlete, requiring controlled camera placement and system-specific software (Pueo et al., 2021; Tomasevicz et al., 2020).

Measurements and metrics. Repetition velocity (Pueo et al., 2021; Tomasevicz et al., 2020)

Validity and reliability. Video-based barbell velocity tracking demonstrates good to excellent concurrent validity for mean and peak barbell velocity when compared with linear position transducers, particularly under standardized camera placement and controlled lifting conditions. However, systematic bias and exercise-specific error are evident, limiting interchangeability with criterion devices, particularly for displacement-derived outcomes. Reliability is high within systems when protocols are held constant, supporting longitudinal tracking. However, between-device reliability is insufficient to justify mixing video-based and LPT systems within the same monitoring program (Pueo et al., 2021; Tomasevicz et al., 2020).

Implementation notes. Calibrate camera position and lens parameters, ensure standardized camera setup (Pueo et al., 2021; Tomasevicz et al., 2020).

4.5 Sport or Implement Mounted Sensors

Relevance to invisible monitoring. Promising, low burden complements to standard monitoring with strong ecological validity (Bortolotti et al., 2023).

Burden and context. Miniaturized sensors on skis, skates, rackets, and gloves; focus on athlete-implement interaction (Russo et al., 2022)

Measurements and metrics. Sport-specific indicators (e.g., ski turn metrics, figure skating jump detection, boxing punch classification, racket stroke characteristics) (Panfili et al., 2022; Worsey et al., 2020).

Validity and reliability. Pilot studies demonstrate feasibility and accurate event detection, but broader field deployment and longitudinal robustness remain to be documented (Bortolotti et al., 2023; Russo et al., 2022).

Implementation notes. Define attachment points and alignment relative to the implement and document environmental conditions (ice, snow, surface). Combine signals with positional, kinetic, or video data for context and revalidate when changing equipment models (Bortolotti et al., 2023).

4.6 Smart Garments (Kinematics and Physiology)

Relevance to invisible monitoring. Attractive for reducing visible device burden, but careful selection, sizing, and periodic validation are essential (Navalta et al., 2020).

Burden and context. Sensors embedded in shirts/vests/bras with textile or printed electrocardiogram (ECG) electrodes and a small recording/transmission module. The burden relates to sizing, laundering, and periodic signal checks of the garments (Navalta et al., 2020; Villar et al., 2015).

Measurements and metrics. Heart rate, sometimes respiratory parameters via strain bands, and basic movement data from embedded accelerometers (Navalta et al., 2020).

Validity and reliability. The best performing garments approach chest-strap accuracy during steady activity, although performance depends on fit and electrode-skin contact. Reliability declines with garment shift or intense movements (Navalta et al., 2020; Parak et al., 2021), though matchplay/high-impact contexts remain under validated (Villar et al., 2015).

Implementation notes. Once integrated as standard kit, garments can support near continuous internal load capture across sessions, but empirical evidence on real-world burden/adherence is limited in current literature.

4.7 Ear-worn Inertial Sensors

Relevance to invisible monitoring. Emerging near invisible option for monitoring; currently a promising complement (Mekruksavanich & Jitpattanakul, 2023).

Burden and context. In-ear devices with IMUs; leverages widespread earbud acceptance (Mekruksavanich & Jitpattanakul, 2023).

Measurements and metrics. Exercise classification, repetition counting, and potential execution quality markers (Mekruksavanich & Jitpattanakul, 2023).

Validity and reliability. High classification accuracy in controlled studies. Gold-standard kinematic validity and multiday reliability are less developed in the current synthesis. Protocol specific validation is recommended prior to implementation (Mekruksavanich & Jitpattanakul, 2023).

Implementation notes. Ensure a secure in-ear fit. Use as a complement to reference IMUs until kinematic validity and multiday reliability are established (Mekruksavanich & Jitpattanakul, 2023).

 

5 Physiological tracking systems

Heart rate–based physiological tracking systems are among the most widely adopted methods for monitoring internal load, autonomic balance, and recovery status in sport (Lacome et al., 2018). These technologies range from chest straps and optical heart rate monitors to advanced wearables capable of estimating heart rate variability (HRV), nocturnal physiology, and cardiorespiratory dynamics (Wang et al., 2017; Gillinov et al., 2017; Dial et al., 2025). Because many devices can passively collect data during sleep or regular daily activities, heart rate–based systems are central to the “invisible” monitoring frameworks that minimize athlete burden (Dial et al., 2025; Cao et al., 2022).

Other physiological tracking systems, as well as hybrid ones, encompass a diverse set of technologies designed to monitor biochemical, neural, thermal, and metabolic states using non-invasive or minimally invasive sensors (Imani et al., 2016; Currano et al., 2018). These tools extend beyond traditional heart-rate-based monitoring by capturing physiological parameters such as muscle activation, hydration, lactate concentration, sleep staging, brain activity, tissue oxygenation, and metabolic substrate consumption (Cretot-Richert et al., 2023; Currano et al., 2018). Their level of “invisibility” varies significantly: with some systems integrated into clothing or worn passively, while others require contact sensors or small wearable modules (Navalta et al., 2020).

5.0 Chest Strap ECG Monitors

Relevance to invisible monitoring. A practical reference device for field heart rate (HR) monitoring; not entirely “invisible” but imposes minimal burden (Parak et al., 2021).

Burden and context. An adjustable strap fitted with chest electrodes that provides low interference but is not usually worn 24/7 (Speer et al., 2020, Wang et al., 2017).

Measurements and metrics. Instant HR, time in zones, R–R intervals for HRV (e.g., root mean square of successive difference; RMSSD), training impulse (TRIMP), excess post-exercise oxygen consumption, and calorie proxies (Parak et al., 2021).

Validity and reliability. Validity is near gold standard accuracy represented by a concordance correlation coefficient of ≈ 0.99 during exercise when compared to the gold standard ECG with minimal bias (Wang et al., 2017). Reliability is very high when worn properly (ICC ~0.8–0.9 for HR and HRV) but drops with poor skin contact (Speer et al., 2020; Wang et al., 2017).

Implementation notes. Wet electrodes if needed; ensure strap tension and placement are consistent to obtain accurate readings (Speer et al., 2020).

5.1 Wrist-Worn Optical Photoplethysmogram (PPG) devices

Relevance to invisible monitoring. Strong for longitudinal resting trends and routine session HR where intensity/motion allows (Wang et al., 2017).

Burden and context. Everyday watch/band that can be worn 24/7 (Navalta et al., 2020).

Measurements and metrics. Provides HR during exercise as well as resting HR trends.  Some devices can also provide resting state HRV scores (Sarhaddi et al., 2022).

Validity and reliability. Validity is good at rest and during moderate exercise (ICC > 0.90), but becomes more variable at high intensity. In some newer devices MAPE was reported as <2% in graded exercise, though motion artefact, skin tone, and ambient light can affect accuracy (Wang et al., 2017; Kim et al., 2023; Navalta et al., 2020; Sarhaddi et al., 2022). Reliability is high when stationery/sleep and day-to-day resting HR are consistent, but greater variability is present during dynamic exercise due to rapid transients (Sarhaddi et al., 2022).

Implementation notes. Prefer resting and steady-state use for HRV. Minimize artefact movement and light leakage under the watch. T-shirt/strap placement to stabilize the device during running can be done if necessary (Kim et al., 2023; Navalta et al., 2020).

5.2 Finger PPG (Smart Rings)

Relevance to invisible monitoring. Highly aligned with invisible recovery/autonomic balance monitoring (Cao et al., 2022).

Burden and context. Continuous wear; excellent acceptance; primarily used at rest and sleep (Cao et al., 2022).

Measurements and metrics. Nocturnal HR and HRV (e.g., RMSSD), recovery/readiness composites; lowest nightly HR; average sleep HR (Cao et al., 2022).

Validity and reliability. In nocturnal recordings, HR and RMSSD have shown high concordance with ECG in several validation settings, particularly when data-quality filtering is applied (r ≈ 0.97–0.99). The level of agreement can vary with device model, algorithms, and the proportion of valid interbeat intervals. Frequency-domain HRV is typically more sensitive to signal quality. Consequently, frequency-domain HRV may show lower agreement than HR and RMSSD. Trend-based interpretation is generally the most defensible under stable conditions (Cao et al., 2022). Night-to-night reliability is high when sleep timing and conditions are consistent; however, daytime spot checks and high-intensity exercise are less well supported by current evidence (Cao et al., 2022; Dial et al., 2025).

5.3 In-Ear HR Sensors (PPG)

Relevance to invisible monitoring. Useful for low burden HR capture during steady aerobic work or recovery blocks (Navalta et al., 2020).

Burden and context. Earbud/ear clip; minimal burden for individual exercise/rest (Ellebrecht et al., 2022).

Measurements and metrics. HR, resting-state HRV, and sometimes temperature (Navalta et al., 2020; Ellebrecht et al., 2022).

Validity and reliability. High accuracy at low/moderate intensities; reliability comparable to other PPG when the earbud fit is stable (Navalta et al., 2020; Ellebrecht et al., 2022).

Implementation notes. Optimize earbud fit and monitor for slippage. Use steady-state segments for HR and HRV capture and verify against a chest strap during onboarding (Navalta et al., 2020; Ellebrecht et al., 2022).

5.4 Capacitive (Contactless) ECG

Relevance to invisible monitoring. Close to ideal for “invisible” resting/sleep monitoring in recovery suites, athlete accommodation or vehicles (Xiao et al., 2025; Škorić 2023).

Burden and context. Sensors embedded in chair/bed surfaces; no wearables; completely passive during rest (Xiao et al., 2025; Škorić 2023).

Measurements and metrics. Resting HR, full HRV features, stress classification, and sleep analysis (Xiao et al., 2025; Škorić 2023).

Validity and reliability. Strong correlation with contact ECG and high night-to-night reliability, but signal quality drops with motion (Xiao et al., 2025; Škorić 2023).

Implementation notes. Deploy in static contexts for best signal quality; avoid interpreting during movement. Maintain consistent posture and textile thickness; position in recovery areas or athlete accommodation where possible (Xiao et al., 2025).

5.5 Textile multi-sensor garments

Relevance to invisible monitoring. Garments worn beneath the standard kit enable low-interaction internal load capture across sessions. The evidence is strongest in controlled treadmill/lab research and should be further validated in sport-specific settings (Navalta et al., 2020; Villar et al., 2015).

Burden and context. Tight-fitting shirts, vests, or bras with textile or printed ECG electrodes in addition to some garments including dual respiratory bands and small logging/transmit modules. Proper sizing and stable electrode–skin contact are crucial for accurate readings (Navalta et al., 2020; Villar et al., 2015).

Measurements and metrics. Heart rate (beat-to-beat), basic HRV at rest/steady state, respiratory rate/ventilation, and activity counts from embedded accelerometers (Navalta et al., 2020; Villar et al., 2015).

Validity and reliability. Validity is near chest-strap ECG at rest and during submaximal exercise, with degradation at higher intensities and with garment shift. Reliability is good under standardized conditions but sensitive to fit and motion. Validity and reliability can vary between products and manufacturers (Navalta et al., 2020; Villar et al., 2015).

Implementation notes. Size carefully (Navalta et al., 2020; Villar et al., 2015).

5.6 Sleep and nocturnal physiology trackers

Relevance to invisible monitoring. Overnight, passive capture of HR/HRV and movement facilitates recovery monitoring with minimal disruption. The strongest evidence exists for nocturnal HR and RMSSD trends (Dial et al., 2025).

Burden and context. Finger rings or bedside/under-mattress sensors operate unattended after setup; thus, adherence is high because the wear/installation burden is low (Dial et al., 2025; Edouard et al., 2021).

Measurements and metrics. Nocturnal HR and HRV (e.g., RMSSD), sleep duration/efficiency, device-specific readiness/recovery composites (Dial et al., 2025).

Validity and reliability. Validity is very high for nightly HR and RMSSD versus ECG. Reliability is robust for night-to-night resting indices, whereas stage classification is more device-dependent (Dial et al., 2025; Edouard et al., 2021).

5.7 Hybrid chemico-physiological patches and composites

Relevance to invisible monitoring. Combined sensing (electrical + optical + inertial ± respiratory bands) offers a richer automated context (load + autonomic + movement) with modest additional burden (Imani et al., 2016).

Burden and context. Small chest/upper-arm patches or integrated garments. They require a brief setup followed by background logging. Firmware and algorithm versions affect the produced outputs (Imani et al., 2016).

Measurements and metrics. Composite HR/HRV with motion and breathing features, or composite physiological and motion features. Sometimes such features are summarized into device-specific composite indices, providing recovery/readiness scores derived from multiple channels (Imani et al., 2016).

Validity and reliability. Validity relies on constituent modalities. In higher-intensity and motion-affected conditions, ECG-based measurements usually display greater accuracy than PPG-based estimates. Performance depends on device design, sensor placement, and signal processing. Respiratory bands perform best when stable. Reliability is generally good in steady conditions, but integration can introduce device-specific biases that require local verification (Imani et al., 2016).

Implementation notes. Lock the hardware/firmware for the monitoring block, document the algorithm versions, and verify the composite outputs against the reference sensors used in your program (Imani et al., 2016).

5.8 Muscle activation and local oxygenation

Relevance to invisible monitoring. Adds modality-specific physiological context (activation, local oxygenation) to internal-load tracking (Bawa et al., 2022; van Vlerken et al., 2025).

Burden and context. Adhesive electromyography (EMG) electrodes or near‑infrared spectroscopy (NIRS) sensors on target muscles; placement and motion control are important for signal quality (Bawa et al., 2022; van Vlerken et al., 2025).

Measurements and metrics. EMG amplitude/frequency features (e.g., root mean square, mean frequency) and muscle oxygen saturation time-courses (Bawa et al., 2022; van Vlerken et al., 2025).

Validity and reliability. Validity is good for canonical activation/oxygenation responses; reliability is acceptable under standardized placement yet sensitive to electrode/optode shift and movement artefact (Bawa et al., 2022; van Vlerken et al., 2025).

Implementation notes. Standardize the placement and calibration; pair with load and technique data; prioritize within-athlete trends before drawing between-athlete comparisons (Bawa et al., 2022; van Vlerken et al., 2025).

5.9 In-ear physiological sensing (PPG ± temperature) as a hybrid alternative

Relevance to invisible monitoring. Ear-canal PPG (± temperature) supports low-burden internal-load capture as the form factor aligns with typical earbud use (Ellebrecht et al., 2022).

Burden and context. Earbud/ear-clip with optical sensor; a stable fit is critical for signal quality (Ellebrecht et al., 2022).

Measurements and metrics. Heart rate and resting-state HRV; some devices also include temperature sensing (Ellebrecht et al., 2022).

Validity and reliability. Validity is high at low-to-moderate intensities, whereas reliability is fit-dependent and may decline with vigorous movement (Ellebrecht et al., 2022).

Implementation notes. Optimize retention, use steady-state segments for HRV, and verify against a chest-strap ECG during onboarding (Ellebrecht et al., 2022).

 

6. Discussion

With the increasing variety of invisible monitoring technologies now available, selecting appropriate tools for training and competition settings remains a constant challenge for coaches, practitioners, and sport scientists. Effective implementation begins by clearly defining the practical questions the monitoring system aims to address. Identifying the sport’s demands and the athlete’s strengths, weaknesses, and development priorities helps determine the most appropriate technologies while minimizing the risk of unnecessary data collection that could increase the burden on athletes and staff, contribute to burnout, or undermine long‑term compliance.

The context in which technology is used is just as important. Several monitoring tools such as GNSS and local positioning systems for external load monitoring (Hoppe et al., 2018), engineered optical tracking systems in stadium environments (Linke et al., 2018), portable force plates for jump and strength assessments (Collings et al., 2024; Mylonas et al., 2023), chest-strap ECG monitors for heart rate and heart rate variability (Parak et al., 2021), cycling power meters for mechanical output in submaximal conditions (Bouillod et al., 2017), and finger-based photoplethysmography rings for nocturnal recovery monitoring (Cao et al., 2022) have reached a level of maturity where their outputs are generally valid and reliable when used within their intended contexts. In contrast, emerging technologies such as muscle oxygenation sensors (van Vlerken et al., 2025), instrument-mounted inertial sensors (Teufl et al., 2019), and multi-sensor composite systems (Navalta et al., 2020) are often more context-sensitive and may require greater practitioner expertise to integrate, interpret, and communicate effectively. Consequently, these technologies may increase the overall implementation burden despite offering richer data streams.

Beyond data collection, the ability to analyze and communicate monitoring outputs to coaches and athletes is a critical determinant of practical value. Some invisible monitoring technologies, such as nocturnal HR and HRV metrics from finger-based PPG devices, produce relatively intuitive outputs that are easily understood by end users (Dial et al., 2025). Others, including electromyography- or near-infrared spectroscopy–based systems, generate complex physiological data that require extensive education and contextualization to support decision-making (Bawa et al., 2022). Therefore, staff expertise, available time, and analytical resources should be considered alongside measurement validity when selecting monitoring tools.

Table 1 summarizes the current landscape of invisible monitoring technologies used in sport, including their relevance, measurement burden, key metrics, and evidence for validity and reliability. Ultimately, successful implementation depends on ensuring that the information gained meaningfully enhances decision-making and provides greater value than the financial, organizational, and cognitive costs associated with data collection and analysis (Camomilla et al., 2018; Seshadri et al., 2019; Leduc & Weaving, 2025; Exel & Dabnichki, 2024). When guided by clearly defined questions and practical constraints, invisible monitoring technologies can provide valuable insights while minimizing disruption to the training process.

Table 1. Summary of existing technologies for invisible monitoring in sports.

Technology	Relevance to invisible monitoring	Burden and context	Measurements and metrics	Validity and reliability
GPS and GNSS wearable units	Field‑sport external load; simple, widely used.	Small back-worn unit; best in open-air.	Distance, speed/accel, High-speed running, peaks.	Good repeatability; accuracy drops with occlusion/covered venues.
Local Positioning Systems	Low-wear, high‑fidelity in equipped venues.	Tag + venue anchors; indoor/arena suited.	Positions, speed/acceleration, COD; IMU fusion possible.	Lower error than GPS for sprints/COD; strong between‑unit reliability.
Engineered optical multi‑camera video	Zero on‑body hardware; match/tactical + load.	Stadium camera array; ops/processing needed.	Trajectories, distance, speed, and inter‑player spacing.	Small bias for distances; undercounts extreme accel events.
Computer‑vision markerless systems	Leverages existing video; minimal athlete burden.	Broadcast/in‑house multi‑camera; calibration heavy.	2D/3D kinematics; trajectories; screening angles.	Good test–retest; absolute agreement < mocap, fine for trends.
Portable force plates	Frequent field testing with lab‑like outputs.	Light, wireless plates; rapid setup.	GRF, impulse, RFD, jump height, COP, asymmetry.	Intraclass Correlation Coefficient often >0.90; small bias vs lab plates.
Force‑estimating and pressure insoles	Hidden in the shoe; near‑invisible gait/impact data.	Thin insoles; placement and footwear matter.	Plantar pressure, COP, contact times; vGRF estimates.	Good for COP/timing; force magnitude needs calibration/drift correction.
Instrumented sports equipment (embedded)	Ecological, sport‑specific mechanics with low burden.	Sensors in paddles/skis; cycling power meters.	Force‑time, deformation, external work (power).	Valid with device‑specific biases; reliable within device.
Segment‑mounted IMUs	Flexible monitoring of technique/events with low burden.	1–5 small sensors; short calibration.	Accel/gyro, orientation; joint angles; jump/event counts.	Acceptable vs mocap (best sagittal); placement affects reliability.
Linear position transducers (LPTs)	Backbone for velocity‑based training.	Cable to bar/belt; quick gym setup.	Bar displacement/velocity, MPV, velocity‑loss, jump proxies.	Device‑dependent biases; good between‑session reliability.
Smartphone inertial sensing	Pragmatic when IMUs unavailable; minimal hardware.	Phone secured to pelvis/trunk; orientation critical.	Time-to‑stabilization, flight time, cadence proxies.	Good for temporal features; multiday reliability less established.
Instrumented mouthguards	Head‑impact exposure with minimal added burden.	Mouthguard with IMU; fit critical.	Head linear/angular accel; impact counts.	Lab accuracy high; field depends on fit/algorithms; verify with video.
Video‑based barbell velocity	VBT without attachments; low-cost entry.	Existing cameras/apps; no bar hardware.	Rep velocity; velocity‑zones; fatigue markers.	Encouraging in controlled gyms; validate locally.
Sport or implement‑mounted sensors	Low‑burden, sport‑specific event detection.	Mini‑sensors on skis/skates/rackets/gloves.	Turn metrics, jump detection, punch/racket classification.	Pilot studies show feasible accuracy; broader field data pending.
Smart garments	Reduces visible devices; continuous internal load.	Sensor embedded shirts/bras; fit and contact of key importance.	ECG HR (± respiration), basic motion.	Approaches chest‑strap at steady state; motion/fit sensitive.
Ear-worn inertial sensors	Near‑invisible gym monitoring; emerging.	In‑ear IMU; relies on a stable fit.	Exercise classification, rep counting; quality markers.	High classification accuracy; kinematic validity still limited.
Chest‑strap ECG monitors	Field reference for HR/HRV with low burden.	Elastic strap electrodes; not 24/7.	HR, R–R, HRV; time‑in‑zones, TRIMP.	Near gold‑standard; very high reliability if contact is good.
Wrist‑worn optical PPG	Everyday wear; great for rest trends.	Watch/band; continuous wear.	HR during exercise; resting HR/HRV (some).	Good at rest/moderate work; motion artefact at high intensity.
Finger PPG (smart rings)	Highly aligned with recovery/autonomic trends.	Ring worn 24/7; best at rest/sleep.	Night HR and HRV (RMSSD); readiness metrics.	Very high concordance with ECG at night; robust night‑to‑night.
In‑ear HR sensors (PPG)	Low‑burden HR/HRV during steady aerobic work.	Earbud/clip; not for contact play.	HR; resting HRV; sometimes temperature.	Accurate at low/moderate work; fit‑dependent reliability.
Capacitive (contactless) ECG	Ideal for passive rest/sleep monitoring.	Sensor embeded chair/bed; through clothing.	Resting HR and HRV; sleep/stress features.	Strong vs contact ECG when static; motion degrades the signal.
Textile multi-sensor garments	Low-interaction internal-load capture across sessions.	Tight shirt/vest with textile ECG and dual respiratory bands; small logger; sizing and contact matter.	HR (beat-to-beat), resting/steady HRV, respiratory rate/ventilation surrogates, posture/activity counts.	Near chest-strap at rest/submax; degrades at high intensity or garment shift; reliability good when fit is stable.
Sleep and nocturnal physiology trackers	Passive overnight recovery tracking with minimal disruption.	Finger ring or under-mattress/bedside sensor; unattended after setup.	Nightly HR and HRV (e.g., RMSSD), sleep duration/efficiency, device readiness/recovery scores.	Very high concordance for nightly HR and RMSSD; staging is more device-dependent; robust night-to-night for resting indices.
Hydration and fluid-balance sensing	Low-interaction hydration status checks.	Small electrodermal wearable or automated urine specific gravity (USG) in a restroom.	Hydration category and/or USG value with alerts.	Accurate classification in controlled protocols; automated USG agrees with benchtop; day-to-day biological reliability is less characterized.
Biochemical and metabolic sensing	Metabolic load/substrate insights with modest user burden.	Skin patch/armband; standardized breath sampling; Continuous Glucose Monitors (CGM) with setup/calibration.	Sweat lactate (± electrolytes, pH), breath acetone (substrate/energy expenditure proxy), glucose (mean/variability/time-in-range).	Tracks trends with lags/individual variability; CGM good at rest/steady but larger errors during rapid changes.
Muscle activation and oxygenation	Physiological context on activation and local oxygenation.	Adhesive EMG or NIRS optodes on target muscles; placement expertise needed.	EMG amplitude/frequency (RMS, mean frequency), muscle oxygen saturation time-courses.	Good validity for canonical responses; reliability acceptable with standardized placement; motion/shift sensitive.

2D = two-dimensional, 3D = three-dimensional, CGM = continuous glucose monitor, COD = change of direction, COP = center of pressure, ECG = electrocardiography, EMG = electromyography, GNSS = global navigation satellite system, GPS = global positioning system, GRF = ground reaction force(s), HR = heart rate, HRV = heart rate variability, ICC = intraclass correlation coefficient, IMU = inertial measurement unit, LPT = linear position transducer, MoCap = motion capture, MPV = mean propulsive velocity, PPG = photoplethysmography, R–R = R–R intervals (beat-to-beat), RFD = rate of force development, RMSSD = root mean square of successive differences, TRIMP = training impulse, USG = urine specific gravity, VBT = velocity-based training, vGRF = vertical ground reaction force.

 

 

Images: Adobe Stock (Education License)

This report publication is part of the activities of the Knowledge Management for Human Performance research group.