Application of Wireless EMG Sensors for Assessing Agonist–Antagonist Muscle Activity During 50-m Sprinting in Athletes

round:Wireless surface electromyography (sEMG) enables the investigation of neu- romuscular control in realistic sports settings; however, ensuring reliable signal acquisition during sprinting remains challenging. This study examined the feasibility of continuous wireless EMG recording in sprinting athletes and evaluated their agonist–antagonist coor- dination patterns.Methods:Ten trained sprinters performed four maximal 50-m sprints on a force plate–equipped track. sEMG was recorded from the biceps femoris (BF), rectus femoris (RF), soleus (Sol), and tibialis anterior (TA) under two receiver configurations: fixed-receiver condition (FRC) and mobile-receiver condition (MRC). Integrated EMG, kinematics, and cross-correlation analyses were performed on a stride-by-stride basis. Results:Continuous high-quality EMG was feasible under MRC, highlighting the practical importance of maintaining receiver proximity in sprint experiments. BF activity during the late swing phase correlated positively with sprint velocity, supporting the performance relevance of pawing. BF/RF interactions varied substantially across individuals, whereas Sol/TA were consistently coactivated, indicating ankle stabilization.Conclusions:Wireless EMG enables reliable in-field monitoring of sprinting athletes, revealing both individu- alized and shared coordination strategies relevant to performance and injury prevention in athletes. Keywords:electromyography; muscle activity; ground reaction force; sprint running; cross-correlations; lissajous figure 1. Introduction Recent advances in wireless sensor technology, including surface electromyography (sEMG), have significantly expanded the possibilities for recording physiological and biomechanical data in naturalistic and high-performance sports

in-field monitoring of sprinting athletes, revealing both individu- alized and shared coordination strategies relevant to performance and injury prevention in athletes. Keywords:electromyography; muscle activity; ground reaction force; sprint running; cross-correlations; lissajous figure 1. Introduction Recent advances in wireless sensor technology, including surface electromyography (sEMG), have significantly expanded the possibilities for recording physiological and biomechanical data in naturalistic and high-performance sports settings [1]. Traditionally, sEMG systems were wired, limiting their application to laboratory-based experiments because of restrictions on movement and measurement range. However, the development of miniaturized and wireless systems has overcome many of these limitations, making it possible to record muscle activity during full-body dynamic tasks, such as sprinting, in real athletic environments. These wireless technologies have led to increased research on sprinting biomechanics, not only in terms of kinematics and kinetics but also in terms of Sensors2025,25, 6395 https://doi.org/10.3390/s25206395

Sensors2025,25, 6395 2 of 19 neuromuscular control [2]. Earlier studies often employed treadmills because of the physi- cal constraints of wired systems [3], but recent studies have demonstrated the feasibility of field-based EMG recordings during sprinting on official 400-m tracks using wireless systems [4–6]. Among the various aspects of muscle function during sprinting, the interaction be- tween agonist and antagonist muscles has drawn increasing attention. This interest stems from both performance and injury prevention perspectives [7–9]. From a performance standpoint, efficient coordination between these muscle groups optimizes force transmis- sion and minimizes unnecessary co-contraction, thereby enhancing the sprinting speed and movement economy [10]. From an injury prevention perspective, appropriate levels of co-contraction can stabilize joints under high loads, but excessive antagonist activation may elevate the strain on muscle–tendon units. This is particularly relevant for the hamstrings during the late swing phase when they undergo rapid eccentric loading and when most sprint-related hamstring injuries occur. During typical locomotion, the alternating activation of agonist and antagonist mus- cles supports efficient and smooth movement. This coordination is mediated by reciprocal inhibition, a neural mechanism in which the activation of one muscle suppresses the ac- tivity of its antagonist. However, spinal reciprocal inhibition decreases significantly as the movement speed increases [11]. Such a reduction in inhibition may contribute to a shift toward co-contraction under high-speed or high-load conditions, where agonist and antagonist muscles are simultaneously activated [12–15]. While co-contraction enhances joint stability and may help prevent injury, it can also diminish agonist force output, thereby reducing movement efficiency [16]. This trade-off is particularly critical in sprinting, where extremely high velocities and impact forces occur within very short timeframes. Despite its relevance to performance optimization and injury prevention, the neuromuscular co- ordination between agonist and antagonist muscles during maximal sprinting, especially over longer distances such as 50 m, is still not fully understood. This knowledge gap arises not only from the technical challenges of collecting precise, continuous physiological and biomechanical data under genuine sprint conditions but also from the complex and highly individual-specific nature of neural control strategies during maximal effort. The present study aimed

antagonist muscles during maximal sprinting, especially over longer distances such as 50 m, is still not fully understood. This knowledge gap arises not only from the technical challenges of collecting precise, continuous physiological and biomechanical data under genuine sprint conditions but also from the complex and highly individual-specific nature of neural control strategies during maximal effort. The present study aimed to characterize lower limb neuromuscular coordination during maximal sprinting in a field-based environment, with a particular focus on the inter- action between agonist and antagonist muscles. A wireless EMG sensor was used to capture muscle activation patterns under realistic sprint conditions. In addition, we sought to ex- amine the individual-specific characteristics of these coordination patterns, acknowledging the highly individual nature of neural control strategies at maximal effort. Furthermore, we addressed a methodological issue concerning the feasibility of continuously recording EMG signals over the full sprint distance in a field-based environment. 2. Materials and Methods 2.1. Participants Ten sprinters participated in this study (nine males: 21.8±3.1 years, 1.72±0.06 m, 67.7±3.0 kg, one female: 22 years, 1.63 m, 59.3 kg). The participants specialized in sprint running event (100-m, 200-m, 110-m Hurdle, 400-m Hurdle). This study was approved by the research ethics committee of the National Institute of Fitness and Sports in Kanoya (23-1-6). All participants were free of any history of lower extremity injury or neurological disorders at the time of the test.

Sensors2025,25, 6395 3 of 19 2.2. Experimental Protocol The sprint trials were conducted indoors at the SPORTEC Sports Performance Research Center (NIFS in Kanoya, Japan) on a 50-m straight tartan track equipped with an array of 54 force plates embedded along the entire length of the track. After completing a standard- ized warm-up, the participants performed four 50-m maximal sprints from a crouched position using starting blocks. Each sprint trial was separated by a rest interval of at least five minutes to minimize fatigue and ensure consistent maximal performance. During each sprint, surface electromyographic (sEMG) signals from the lower limb muscles and ground reaction forces were recorded simultaneously. All signals were time-synchronized to ensure the precise alignment of neuromuscular and biomechanical data for subsequent analysis. To evaluate the optimal signal acquisition conditions, two sprint trials were conducted under a fixed-receiver condition (FRC), where the wireless receiver unit (Wave Plus, Cometa, Italy) was placed at the 25-m midpoint of the 50-m track. In the other two trials, measurements were performed under a mobile-receiver condition (MRC), with the receiver unit dynamically following the runner at an approximate distance of 5–7 m to ensure continuous data transmission throughout the entire sprint distance. 2.3. Instrumentation and Data Collection Surface electromyographic (sEMG) signals were recorded from the tibialis anterior (TA) and soleus (Sol), rectus femoris (RF) and biceps femoris (BF) muscles of the right leg using wireless Cometa Pico EMG sensors (Cometa Systems, Milan, Italy). These muscles were selected because they represent functionally antagonistic pairs at the thigh (BF–RF) and shank (Sol–TA), allowing the analysis of reciprocal muscle activation patterns between agonist and antagonist groups during sprint running. Prior to electrode placement, the skin was shaved if necessary, lightly abraded, and cleansed with alcohol to ensure low impedance (<5 kΩ). Each sensor unit incorporated its own integrated transceiver and transmitted data via Bluetooth transmission to a common receiver. The hardware weight of each sensor unit was approximately 7 g, and the device provided continuous operation for approximately 12 h on a full charge. No participant reported disturbance or performance impairment caused by the sensor units.

ensure low impedance (<5 kΩ). Each sensor unit incorporated its own integrated transceiver and transmitted data via Bluetooth transmission to a common receiver. The hardware weight of each sensor unit was approximately 7 g, and the device provided continuous operation for approximately 12 h on a full charge. No participant reported disturbance or performance impairment caused by the sensor units. Conductive-gel-based disposable Ag/AgCl surface electrodes with a 8-mm diameter (Cardinal Health, Dublin, OH, USA) were applied in a bipolar configuration over the muscle bellies, aligned parallel to the muscle fiber direction, in accordance with the SENIAM guidelines [17]. The interelectrode distance was maintained at approximately 20 mm. The EMG signals were amplified and band-pass filtered (10–500 Hz). EMG signals were sampled at 2000 Hz to prevent aliasing and to capture high-frequency components during rapid, ballistic contractions. Although the main frequency content of surface EMG lies below ~500 Hz, oversampling at 2000 Hz provides adequate resolution for detecting rapid fluctuations and minimizes reconstruction errors. According to the Nyquist theorem, the sampling frequency should exceed at least twice the maximum frequency component of interest; in practice, 2–3 times higher frequencies are recommended to ensure adequate temporal resolution and to reduce aliasing risk [18]. This rationale is also in line with previous sprint EMG studies that adopted the same sampling rate [4–6]. GRF data were collected from a 50-m long force plate system (TF-90100, TF-3055, TF-32120, Tec Gihan, Uji, Japan) at a sampling rate of 2000 Hz. Data were collected using the “All-Plate Mode”, in which 50 force plates installed along the 50-m runway are CPU-integrated to function as one continuous force plate (1 m×50 m). The remaining four plates at the starting line were used independently for the starting blocks (left and right foot) and hand placements, and were not included in the CPU integration.

Sensors2025,25, 6395 4 of 19 sEMG and GRF signals were synchronized using a transistor–transistor logic (TTL) trigger signal transmitted via a Bayonet Neill–Concelman (BNC) cable. The sEMG signal was collected using a wireless receiver unit (Wave Plus, Cometa, Milan, Italy) and transmit- ted to a PC via USB using proprietary acquisition software (EMG and Motion Tools 8.6.2.0, Cometa, Milan, Italy). To accommodate receiver mobility during overground sprinting, the Wave Plus receiver was carried by the examiner, following the participant. Because the receiver was bus-powered via USB, a 30-m active USB extension cable with a built-in signal repeater (KB-USB-R230; Sanwa Supply, Okayama, Japan) was used to prevent signal degradation. The Wave Plus also supports external synchronization via a 2.5 mm audio jack. For precise temporal alignment, a TTL pulse was generated using a starting pistol and transmitted simultaneously to both the Wave Plus and GRF systems via a 2.5 mm jack–to–BNC cable (Figure). Figure 1.A series of methods for recording sEMG and GRF during 50-m sprint. 2.4. Data Analysis The sEMG and GRF data were analyzed using MATLAB (The Math Works, Natick, MA, USA). The sEMG signals were processed as follows: (1) bandpass filtering with a band- width of 20–500 Hz, (2) rectification, (3) smoothing with a time constant of 0.03, and (4) normalization based on the maximum amplitude during maximal voluntary contraction (MVC) (%EMG@MVC, i.e., EMG amplitudes expressed as a percentage of MVC). Following normalization, the EMG signals were time-integrated within each stride cycle to obtain the integrated EMG (iEMG), expressed in %MVC·s. Agonist–antagonist muscle activity patterns were visualized using x–y plots based on the smoothed sEMG signals. The x- and y-axes of the plot represent the posterior (BF, Sol) and anterior (RF, TA) muscles, respectively, based on their anatomical locations. To assess the temporal relationship between agonist and antagonist muscle activity patterns, cross- correlation analysis was performed between the normalized iEMG signals of biarticular thigh and lower leg muscle pairs using the normalized cross-correlation function (xcorr, coeff) in MATLAB. For each stride cycle, the BF and Sol signals were designated as the reference signals (first muscle), and the RF

locations. To assess the temporal relationship between agonist and antagonist muscle activity patterns, cross- correlation analysis was performed between the normalized iEMG signals of biarticular thigh and lower leg muscle pairs using the normalized cross-correlation function (xcorr, coeff) in MATLAB. For each stride cycle, the BF and Sol signals were designated as the reference signals (first muscle), and the RF and TA signals were designated as the second muscle in each pair. The maximum cross-correlation coefficient (r) and its corresponding lag time (in seconds) were extracted, in addition to the coefficient at a lag of zero. A positive lag indicated that

Sensors2025,25, 6395 5 of 19 the activity of the second muscle (RF or TA) followed that of the first muscle (BF or Sol), whereas a negative lag indicated that the second muscle preceded the first muscle. The coefficient at lag zero was used as an indicator of the coactivation. GRF data were processed using a low-pass filter (The Math Works, Natick, MA, USA) with a cutoff frequency of 50 Hz, and a threshold of 20 N was applied to the vertical GRF component (Fz) to distinguish the foot strike, following previous studies [19,20]. In the present study, the foot strike was defined as the point at which Fz exceeded this threshold for 20 consecutive samples, and the foot-off was identified as the point at which this continuity ceased. The period between the foot strike and subsequent foot-off was defined as the ground contact phase, and its duration was defined as the contact time (CT). One lower limb stride cycle was defined as the interval from the foot strike of one leg to the subsequent foot strike of the same leg. The duration of this cycle was termed cycle time (1CTime). Each cycle was divided into a ground contact phase and a swing phase (Swing), with the swing phase further subdivided into early swing (ESwing), mid-swing (MSwing), and late swing (LSwing), resulting in four distinct phases. ESwing was defined as the period from foot-off of the ipsilateral leg to foot strike of the contralateral leg, MSwing, from contralateral foot strike to its foot-off, and LSwing, from contralateral foot-off to ipsilateral foot strike (Figure). Figure with specific stride cycle phases. This figure depicts the period from one right foot contact to the next occurrence of right foot contact two cycles later, thus covering two complete stride cycles. The duration of each was defined as the phase time. The horizontal component of the center of pressure (COP) was used to calculate the stride length (SL) and cycle time. For each cycle, the midpoint of the COP during the ground contact phase of the right leg was identified, and the average of 10 samples

covering two complete stride cycles. The duration of each was defined as the phase time. The horizontal component of the center of pressure (COP) was used to calculate the stride length (SL) and cycle time. For each cycle, the midpoint of the COP during the ground contact phase of the right leg was identified, and the average of 10 samples (five before and five after the midpoint) was computed [20]. SL was defined as the distance between the average COP position at the right foot strike and the average COP position at the subsequent right foot strike. The cycle time was defined as the elapsed time between two successive right-foot strikes. The cycle frequency (CF) was calculated as the reciprocal of 1Ctime, and the running speed (RS) was obtained as the product of SL and CF. The mean anteroposterior force (mAP), which has been reported in previous studies to be associated with sprint acceleration performance, was calculated [19]. For each step, the anteroposterior GRF component was averaged over the ground-contact phase and normalized to body mass (N/kg). Figure 2.Each right-leg stride cycle was divided into four phases. The contact phase was defined as the duration from the foot strike to the foot-off. The swing phase was further divided into three subphases: early swing phase (from foot-off to contralateral foot strike), mid-swing phase (from contralateral foot strike to contralateral foot-off), and late swing phase (from contralateral foot-off to ipsilateral foot strike). 2.5. Statistical Analysis Statistical analyses were performed using MATLAB (MATLAB ver. R2023a; The MathWorks, USA). Data are expressed as mean±standard deviation (SD). To assess the within-rater reliability of the EMG measurements, intraclass correlation coefficients

Sensors2025,25, 6395 6 of 19 (ICC) were calculated based on a two-way mixed-effects model [ICC (3,1)], reflecting the consistency of single measurements across three repeated trials for each participant. ICC (3,1) was selected because it is appropriate when the same rater evaluates all measure- ments under fixed conditions, and when absolute agreement for single measurements is of interest. Following Carius et al. (2015) [21], reliability was interpreted as good (ICC≥0.80), fair (ICC 0.60–0.79), or poor (ICC < 0.60). The ICC (3,1) for TA EMG amplitude at MVC was 0.964, with a 95% confidence interval ranging from−0.845 to 1.00, indicating good within-rater reliability. Figure 3.Representative recordings of sEMG (BF, RF, Sol, and TA) and ground reaction force (Fy: anterior–posterior; Fz: vertical) for participant #03. The gray dashed lines indicate the boundaries between the movement phases, and the blue-shaded region marks the stride cycle analyzed in this study. Pearson’s correlation coefficients were calculated to examine the relationships between the integrated EMG (iEMG) activity of each muscle (BF, RF, Sol, and TA) and sprint performance variables, running speed (RS) and mean anteroposterior force (mAP). The relationships between running speed and relative iEMG (%MVC·s) of each muscle (BF, RF, Sol, and TA) were examined using regression analyses. Both linear (Pearson’s correlation) and nonlinear (exponential and quadratic) models were tested. For linear models, Pearson’s correlation coefficients (r) and their significance levels were reported, whereas for nonlinear models the goodness of fit was evaluated by the coefficient of determination (R 2 ). Model selection was based on both the magnitude of R 2 and the inspection of residual plots. The model that achieved the highest R 2 without systematic patterns in the residuals was considered the best fit for each muscle group. Cross-correlation analyses were performed for each muscle pair (RF–BF and TA–Sol). EMG signals were segmented stride-by-stride throughout the 50-m sprint, and cross- correlations were computed for each stride. Both the peak correlation coefficient and the corresponding time lag were extracted as primary outcome measures, characterizing the strength and temporal characteristics of the agonist–antagonist muscle activation patterns. Statistical significance was set atp< 0.05.

for each muscle pair (RF–BF and TA–Sol). EMG signals were segmented stride-by-stride throughout the 50-m sprint, and cross- correlations were computed for each stride. Both the peak correlation coefficient and the corresponding time lag were extracted as primary outcome measures, characterizing the strength and temporal characteristics of the agonist–antagonist muscle activation patterns. Statistical significance was set atp< 0.05.

This study examined the feasibility of continuous wireless EMG recording in sprinting athletes.