Self-powered swimming motion monitoring using a single-wire structure triboelectric nanogenerator

Summary

Wearable sensors for swimming safety monitoring have attracted growing interest due to the need for real-time and reliable tracking in aquatic environments. Here, we present a single-wire triboelectric nanogenerator (SW-TENG) integrated into a life jacket for self-powered swimming safety monitoring. The SW-TENG is composed of a large silicone tube coated with copper paint and a smaller silicone tube wrapped around a single wire, forming a flexible structure suitable for underwater operation. The device delivers a peak voltage of 57.2 V, transferred charge of 30.1 nC, and current of 3.5 μA under 25 N load at 2 Hz. When embedded into a swimsuit, it enabled sensitive multi-joint motion tracking and distinguished among various swimming behaviors. With wireless data transmission, the system provides real-time, self-sustained monitoring crucial for swimmer safety and performance evaluation. This work demonstrates the potential of SW-TENG technology for intelligent aquatic safety systems and rapid rescue applications.

Graphical abstract

Highlights

• •A single-wire TENG enables self-powered swimming motion monitoring
• •The wearable device captures multi-joint motions with high sensitivity
• •Deep learning identifies normal, atypical, and emergency swimming states
• •The system supports real-time aquatic safety monitoring and warning

Electrochemistry; applied sciences

Introduction

The rapid and sustained expansion of the maritime industry, driven by significant advancements of society and economy, has greatly increased the demand for marine infrastructure and transportation services worldwide.^1,2 Despite these developments, the unpredictable and harsh marine environment continues to pose substantial risks, particularly for individuals who accidentally fall into the water. These risks are further intensified by extreme conditions such as low water temperatures and limited rescue resources in remote areas.^3,4,5 Prolonged exposure to cold seawater can lead to rapid heat loss and a sharp decline in physical capability, significantly reducing survival probability during maritime accidents.⁶ In addition to environmental challenges, rescue operations are often hindered by the lack of timely and reliable information about the motion states and behaviors of individuals in distress, especially in complex aquatic environments with poor visibility and accessibility.⁷ Wearable sensing technologies capable of monitoring human motion and activity states during swimming can provide critical information for assessing distress conditions and improving rescue efficiency.^8,9,10 Continuous motion-based monitoring enables rapid identification of abnormal behaviors, such as loss of coordinated movement or emergency gestures, thereby supporting faster response and more informed rescue decisions.¹¹ Nevertheless, the practical deployment of wearable sensing systems in aquatic environments remains challenging due to constrained power supply, delayed response, and insufficient adaptability to water-induced mechanical and environmental disturbances.^12,13 These limitations highlight the need for robust, self-sustained sensing approaches specifically tailored to reliable motion monitoring in harsh marine conditions.

Triboelectric nanogenerator (TENG) technology converts mechanical energy into electrical energy through the triboelectric effect, providing a self-powering capability that makes it particularly suitable for real-time monitoring in complex marine environments without reliance on external power sources.^{14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37} TENGs have been widely employed in the design of self-powered sensors, where their ability to convert mechanical stimuli into electrical signals enables reliable and sustainable sensing applications.^{38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57} By integrating TENGs, smart life jackets can continuously monitor the wearer’s physiological data, such as heart rate, body temperature, and physical activity, while utilizing ambient mechanical vibrations for energy generation and continuous power supply, overcoming the limitations of traditional battery-powered sensors. Traditional life jackets, though providing buoyancy and insulation, often lack integrated monitoring capabilities, complicating real-time rescue decisions. For instance, Liu et al. developed a self-heating life jacket to improve insulation, and Han et al. created an inflatable version powered by chemical energy for rapid inflation, but both designs lack real-time physiological monitoring capability.^58,59 Researchers have attempted to address this by integrating sensors for monitoring underwater physiological states, such as Zou et al.’s biomimetic stretchable nanogenerator for energy harvesting and posture detection.⁶⁰ While effective, such sensors require attachment to multiple body parts, making them susceptible to interference and reducing comfort, whereas wireless body area network (WBAN) technology overcomes these limitations by enabling wireless sensor deployment across the body, thereby enhancing comfort and flexibility.^61,62 For example, Rasyide designed a WBAN-based smart life jacket for mountaineers to monitor physiological status.⁶³ However, while these systems have not addressed waterproofing challenges for underwater use, integrating TENG with WBAN can significantly enhance the stability and functionality of smart life jackets in aquatic environments.^61,62 Despite these advancements, challenges remain, particularly concerning waterproofing, sensor sensitivity, and signal stability. This study aims to develop a smart life jacket combining TENG and WBAN technologies to provide flexibility, real-time monitoring, and robust waterproof performance, enhancing water rescue efficiency and improving wearers’ survival rates.

In recent years, wearable sensors for swimming safety monitoring have gained significant attention, driven by the need for real-time, reliable tracking of swimmer conditions in aquatic environments. This study introduces a single-wire structure TENG (SW-TENG) integrated into a wearable system for real-time monitoring of swimming motions and swimmer safety. The term “single-wire structure” refers to the structural integration and simplified wiring configuration of the device, where a single flexible wire is used for external electrical connection. It should be noted that the SW-TENG operates based on a dual-electrode contact-separation triboelectric mechanism, rather than a conventional single-electrode mode, using the ground as a reference. This unique feature makes it particularly suitable for continuous monitoring in dynamic and challenging aquatic environments. The SW-TENG is constructed with a flexible, wire-based structure, composed of a large silicone tube coated with conductive copper paint and a smaller silicone tube wrapped around a single wire. The device achieves a peak open-circuit voltage (V_OC) of 57.2 V, a maximum transferred charge (Q_SC) of 30.1 nC, and a peak short-circuit current (I_SC) of 3.5 μA under a maximum load of 25 N and a working frequency of 2 Hz. These outstanding electrical outputs validate its potential as a robust, self-powered energy harvester for wearable monitoring systems. Furthermore, by integrating deep learning techniques for real-time motion classification, the SW-TENG can accurately distinguish between various swimming states, such as normal swimming, help-seeking gestures, and tremor-like movements. This capability, combined with its flexibility and self-powered operation, positions the SW-TENG as a highly sensitive motion sensor, capable of identifying emergency situations and enhancing rescue strategies. This work highlights the innovative combination of energy harvesting and motion recognition for intelligent swimmer safety monitoring, providing an advanced solution for improving real-time safety evaluation and performance analysis in aquatic environments.

Results and discussion

Design and application of SW-TENG for intelligent life-saving systems

Figure 1A illustrates the overall concept of integrating the SW-TENG into water rescue scenarios. When a swimmer falls into distress, the self-powered device continuously acquires motion and posture signals, which are wirelessly transmitted to the monitoring station. Rescue ships and helicopters can promptly locate the individual, while an artificial intelligence-assisted decision-making system analyzes the received attitude data to guide efficient rescue operations. This integrated system ensures real-time monitoring, automatic distress alerts, and adaptive rescue responses. Figure 1B presents the configuration of the self-powered life jacket embedded with multiple SW-TENG units at strategic body locations. These distributed sensors detect limb motions and torso posture, converting biomechanical stimuli into electrical signals. The generated signals are transmitted wirelessly to an external processing module, where embedded algorithms analyze the motion patterns and provide feedback for recognition. This closed loop—comprising signal acquisition, algorithmic analysis, and action feedback—enables continuous and autonomous monitoring of the swimmer’ activity states. Figure 1C shows the structural principle of the SW-TENG device. The device consists of an internal silicone tube encapsulating a single conductive wire electrode, surrounded by a slightly larger external silicone tube. The inner surface of the external tube is coated with a thin layer of conductive copper paint serving as the counter electrode. During deformation or bending, periodic contact-separation between the silicone tube and electrodes induces charge transfer through triboelectric coupling and electrostatic induction, thereby producing stable electrical output. The cross-sectional schematic highlights the coaxial configuration, which ensures uniform distribution of the triboelectric interface and reliable signal collection. Figure 1D displays photographs of the fabrication process and mechanical adaptability of the SW-TENG device. In Figure 1D1, a silicone tube containing the internal wire electrode is shown. Figure 1D2 displays the external silicone tube coated with copper paint as the outer electrode. In Figure 1D3, the integrated structure of the assembled SW-TENG device is demonstrated. Finally, Figure 1D4 shows the device in a knotted state, verifying its excellent flexibility and mechanical robustness under complex deformations. Hence, the SW-TENG design combines structural simplicity, mechanical compliance, and effective triboelectric performance. Its integration into wearable life-saving equipment provides a promising approach for self-powered motion monitoring, real-time swimmer safety assessment, and intelligent water rescue systems.

Figure 1.

Schematic illustration and structural design of the SW-TENG for swimmer monitoring and rescue applications

(A) Conceptual framework showing integration of the SW-TENG system into life-saving operations.

(B) Configuration of a self-powered life jacket with distributed SW-TENG units for continuous biomechanical monitoring and feedback.

(C) Structural design of the SW-TENG device consisting of an internal silicone tube with a wire electrode and an external silicone tube coated with copper as the counter electrode.

(D) Photographs of device fabrication and flexibility tests: (D1) internal silicone tube with wire electrode, (D2) external tube coated with conductive copper paint, (D3) assembled SW-TENG device, and (D4) device in a knotted state demonstrating high mechanical compliance.

The working principle of SW-TENG device

The SW-TENG device functions through a cyclic contact-separation process between the silicone tube and the electrodes, where triboelectric charging and electrostatic induction jointly contribute to electrical output. The operation sequence, presented in Figures 2A1–A4, can be divided into four successive stages. Figure 2A1 depicts the first pressing step. When an external load is applied, the coaxial silicone tube is deformed, forcing the wire electrode layer and the dielectric interface into close contact with the copper-coated outer electrode. The difference in triboelectric polarity results in the accumulation of opposite charges at the interface. During the release stage, as illustrated in Figure 2A2, the applied pressure is removed. The elastic recovery of the silicone tube increases the separation between the two electrodes, generating an electrostatic potential difference. This potential drives electrons through the external circuit, giving rise to an output signal. The process then enters the re-pressing state, shown in Figure 2A3. As the tube is compressed again, the gap diminishes and the interfacial contact is restored. The change in potential forces the electrons to flow in the reverse direction compared with the previous stage, resulting in alternating current output. Finally, Figure 2A4 corresponds to the recovery state, where the silicone tube returns to its original shape and the system approaches equilibrium. The charge distribution stabilizes at the interface, completing a full working cycle. By repeating this sequence continuously, the SW-TENG can produce stable alternating electrical signals. The coaxial geometry ensures uniform charge transfer and reliable output, while the elasticity of the silicone tube provides excellent mechanical adaptability, making the device highly promising for energy harvesting and motion monitoring in aquatic environments.

Figure 2.

Working mechanism and output performance of the SW-TENG device

(A) Working mechanism of SW-TENG device.

(B–D) V_OC (B) Q_SC (C), and I_SC (D) at different operating frequencies.

(E–G) Influence of applied force on V_OC (G) Q_SC (H), and I_SC (I), all of which increased with stronger compression.

(H–J) Effect of compressive strain on V_OC (H), Q_SC (I), and I_SC (J), with outputs enhanced as strain increased from 100% to 180%.

(K) Durability test over 15,000 cycles demonstrating stable current output without signal degradation.

(L) Capacitor charging performance of SW-TENG with different capacitances (1, 2, and 3 μF).

Effect of mechanical and environmental conditions on the output performance of SW-TENG device

As shown in Figure 2B, the V_OC was measured at frequencies ranging from 2 to 6 Hz. Remarkably, the V_OC remained almost unchanged across the entire frequency window, suggesting that the generation of surface potential is dominated by the triboelectric charge density and effective contact area rather than the actuation rate. This phenomenon demonstrates that even when the SW-TENG is subjected to motions of different speeds, the voltage output can remain stable, which is particularly beneficial for wearable or aquatic applications where excitation frequency is not constant. The Q_SC, as displayed in Figure 2C, exhibited a similar trend, showing negligible fluctuation with increasing frequency. This observation further confirmed that the amount of charge generated per cycle is a relatively intrinsic property of the device, primarily determined by the material combination and contact-separation mechanism, rather than the external operation speed. Such frequency-independent stability in V_OC and Q_SC ensures consistent energy conversion efficiency across diverse working scenarios. In contrast, the I_SC, as shown in Figure 2D, increased notably as the frequency rose. As current is defined as the rate of charge transfer per unit time, faster mechanical actuation naturally accelerates electron migration in the external circuit, thereby resulting in higher peak currents. This frequency-dependent enhancement of current output highlights the ability of the SW-TENG to respond dynamically to motion speed, which is highly advantageous for sensing applications that rely on real-time current signals. The influence of applied force was also systematically examined. As shown in Figure 2E, the V_OC steadily increased with the applied force, reaching a peak of 57.2 V at the maximum load of 25 N, under a working frequency of 2 Hz. This increase in V_OC is attributed to higher pressure, which enhances the interfacial contact between the electrodes and the dielectric surface, thereby elevating the triboelectric potential. A similar trend was observed in Q_SC (Figure 2F), which also rose with increasing force, reaching a maximum value of 30.1 nC, confirming that larger compressive forces facilitate greater charge generation and transfer. As shown in Figure 2G, the trend in I_SC further supported this observation, which peaked at 3.5 μA under the maximum load of 25 N, demonstrating that stronger loads not only improve contact electrification but also shorten the charge transfer time, leading to a significant enhancement in current output. As shown in Figures 2H–2J, when the compressive strain increased from 100% to 180%, the V_OC, Q_SC, and I_SC all exhibited pronounced enhancements. This behavior can be attributed to the fact that greater deformation induces larger effective contact areas and higher interfacial stress, thereby amplifying triboelectric charge production. The proportional relationship between the output and strain demonstrated the suitability of the SW-TENG for biomechanical monitoring, where human motion often involves variable deformation levels. Device reliability was further assessed through cyclic tests, with the results shown in Figure 2K. The SW-TENG maintained highly stable current output for nearly 15,000 cycles at 10 Hz, and no visible signal degradation was observed throughout the test. This excellent long-term durability highlights the robustness of the coaxial silicone structure, ensuring consistent functionality under repeated deformation and continuous operation. Finally, the capability of the device to store harvested energy, demonstrated in Figure 2L, was evaluated by charging capacitors of different capacitances (1, 2, and 3 μF). The charging curves revealed that the capacitor voltage increased progressively with time, with smaller capacitors achieving higher voltages more quickly, whereas larger capacitors required longer durations to reach comparable levels. These results clearly indicated that the SW-TENG can efficiently convert mechanical energy into storable electrical energy, which can then be used to power small-scale electronics or wireless transmission modules. Collectively, these systematic investigations revealed that while V_OC and Q_SC are largely unaffected by frequency variations, the applied force and compressive strain strongly enhance the overall electrical output. Coupled with excellent durability and effective energy storage performance, the SW-TENG demonstrates significant promise for integration into self-powered sensing and monitoring systems, particularly in aquatic environments where stable and reliable energy harvesting is indispensable.

The capacitor charging results are presented as a proof-of-concept demonstration of mechanical energy generation by the SW-TENG, rather than as an indication of sufficient power supply for continuous system operation. Due to the intrinsically high internal impedance and intermittent output characteristics of TENGs, the harvested energy is primarily suitable for illustrating the feasibility of energy conversion from swimming motions. It should be noted that the charging of small capacitors does not imply the capability to continuously power electronic components such as microcontrollers, power management circuits, or wireless transmission modules, which typically require significantly higher power levels. Therefore, the capacitor charging behavior is intended to demonstrate the existence of usable electrical energy generated from human swimming motions at a conceptual level.

Integration of SW-TENG into wearable swimsuits for real-time swimming posture monitoring

To clearly illustrate the working principle of the proposed self-powered swimming motion monitoring system, the overall system architecture is described here. The system consists of distributed SW-TENG units, a signal acquisition module, an energy management unit, a microcontroller unit (MCU), and a wireless transmission module. When swimming motions occur, mechanical deformation of different body joints induces electrical outputs from the corresponding SW-TENGs. These outputs serve two roles in the system: (1) providing electrical signals that directly reflect joint motions, and (2) supplying electrical energy derived from human motion. It should be emphasized that the SW-TENGs are not required to perform energy harvesting and signal sensing simultaneously at the same electrical port. In the proposed system concept, the SW-TENG outputs are functionally decoupled in time and circuit pathways. The electrical energy generated by repeated swimming motions can be accumulated through a rectification and energy storage module, which supports intermittent system operation, such as signal acquisition and wireless data transmission. Meanwhile, the sensing signals are obtained through high-impedance electrical measurement, which does not impose a significant load on the SW-TENG outputs. Therefore, energy harvesting and sensing are not physically conflicting processes in this system. Instead, the SW-TENGs enable self-powered sensing and event-driven system operation, where motion-induced electrical signals are directly used for motion monitoring, while the harvested energy supports auxiliary electronic functions in a duty-cycled manner. This architecture allows the swimming motion monitoring system to operate without reliance on external power sources for signal generation, demonstrating a practical and conceptually sound self-powered sensing system for wearable aquatic applications. The electrical outputs of SW-TENG represent motion sensing signals acquired under high-impedance measurement conditions, which are intentionally optimized for reflecting joint movements and swimming states, rather than energy harvesting. As a result, the voltage amplitudes in these figures are relatively low and are not intended to charge energy storage devices or directly power electronic modules. These motion-induced electrical signals encode rich temporal and amplitude information related to swimming behaviors, enabling effective motion monitoring and state recognition, while remaining electrically decoupled from the energy harvesting pathway.

Figure 3A shows the electrical response of the SW-TENG installed at the elbow joint during bending motions of 30°, 60°, and 90°. The generated voltage signals displayed clear distinctions in amplitude and frequency, corresponding to different bending angles, thereby verifying the sensitivity of the sensor to elbow flexion. Similarly, Figure 3B presents the voltage outputs when the device was attached to the knee joint. Consistent with the elbow test, increasing bending angle led to stronger voltage outputs, demonstrating effective detection of lower-limb movements. Figure 3C further illustrates the monitoring of shoulder abduction at different angles, with voltage signals exhibiting comparable trends, confirming the versatility of the SW-TENG in detecting upper-limb dynamics. To establish a complete self-powered monitoring system, a signal processing and transmission framework was designed, as illustrated in Figure 3D. The TENG sensors generate electrical signals that are processed by a microcontroller through an ADC acquisition module, followed by regulation and voltage boosting to ensure stable DC supply for the electronics. The processed signals are then transmitted via a UART-based wireless communication module to a terminal device for visualization and data storage. This modular configuration highlights the capability of the SW-TENG to function not only as a power source but also as a real-time motion sensor. The wearable prototype is displayed in Figure 3E, where multiple SW-TENG units were strategically distributed across the front and back of the swimsuit. Sensors were mounted on the shoulders (RS and LS), arms (RA and LA), and knees (RN and LN), covering key motion joints to ensure comprehensive motion detection during swimming activities. The distributed configuration allows simultaneous monitoring of multi-joint dynamics, enabling full-body motion recognition. The corresponding electrical outputs under representative swimming actions are shown in Figures 3F–3H. During tremor-like movements (Figure 3F), all sensors generated synchronous voltage fluctuations, reflecting the uniform microvibrations transmitted throughout the body. For swimming strokes (Figure 3G), the sensors recorded periodic voltage signals with distinct amplitudes at different joints, illustrating that the system can effectively distinguish swimming-related limb dynamics. Under help-seeking or distress gestures (Figure 3H), the device produced unique signal patterns across multiple channels, providing critical information for identifying emergency states. Hence, the results confirmed that the SW-TENG can be seamlessly integrated into swimsuits for distributed motion sensing across multiple joints. The sensors demonstrated high sensitivity to angular variations, robust adaptability to both upper- and lower-limb monitoring, and reliable recognition of characteristic swimming behaviors. Coupled with the wireless acquisition and transmission system, this approach establishes a self-powered, real-time monitoring framework that holds strong potential for swimmer safety evaluation, performance analysis, and intelligent rescue operations in aquatic environments.

Figure 3.

Demonstration of SW-TENG integrated into a wearable swimsuit for motion and posture monitoring

(A–C) Voltage outputs of SW-TENG sensors attached to the elbow (A), knee (B), and shoulder (C) under bending angles of 30°, 60°, and 90°, showing clear signal distinctions corresponding to different joint movements.

(D) Schematic of the self-powered sensing and transmission system.

(E) Photographs of the wearable prototype with multiple SW-TENG units distributed across the front (RS, LS, RA, and LA) and back (RN and LN) of the swimsuit.

(F–H) Representative voltage signals from multi-sensor arrays during different swimming-related motions. (F) Tremor-like movements. (G) Swimming strokes. (H) Help-seeking gestures.

Self-powered motion sensing and swimming posture analysis based on SW-TENG device

The V_OC signals generated from different body parts, including the shoulder, leg, and arm, are displayed in Figure 4A. Distinct voltage waveforms were observed for each joint, confirming the high sensitivity of the SW-TENG to localized biomechanical deformation. The corresponding I_SC, shown in Figure 4B, revealed clear pulse outputs with magnitudes that varied across different joints, further validating the capability of the device to distinguish between multiple motion sources. Figure 4C illustrates the Q_SC associated with these movements, which closely matched the voltage and current patterns, indicating consistent energy conversion performance and stable triboelectric charge transfer during repeated actuation. Collectively, these data confirmed that the SW-TENG exhibits reproducible voltage, current, and charge outputs when integrated onto different body parts, thereby laying the foundation for wearable motion sensing applications. As shown in Figure 4D, the output voltage statistics of sensors placed on the arm, shoulder, and leg were analyzed across different motion states, including swimming strokes, help-seeking gestures, and tremor-like micromovements. The results revealed that swimming strokes produced the highest output signals, while help gestures and tremors generated relatively weak responses, suggesting that the device can not only capture regular swimming movements but also identify atypical or emergency states. The sensitivity of the sensor to bending deformation was quantified (Figure 4E). The V_OC exhibited a nearly linear increase with the increase in bending angle from 70° to 140°, demonstrating excellent angle-resolving capability. This strong correlation highlights the feasibility of using SW-TENG for quantitative posture tracking during swimming. Figure 4F further illustrates the dynamic evolution of voltage signals synchronized with bending and recovering processes, where the voltage waveform corresponded precisely to the angular variation curve. Such agreement confirms the accuracy of the sensor in capturing real-time joint kinematics. The monitoring of complete swimming strokes is schematically illustrated in Figure 4G. The process can be divided into three characteristic phases: the initial phase, where the arm and leg prepare for propulsion; the propulsion phase, where both limbs generate thrust through synchronous movements; and the recovery phase, where the limbs return to their original positions. Each phase is accompanied by distinct SW-TENG signals, demonstrating the capacity of the device to resolve complex sequential motions. Finally, Figure 4H presents representative voltage and bending angle signals during arm, shoulder, and leg movements across different phases of swimming. For the arm, the signal exhibited clear bending, keeping, and recovering stages, with well-defined voltage peaks corresponding to dynamic angular changes. For the shoulder, voltage responses reflected flapping and release actions, capturing both power strokes and relaxation intervals. For the leg, the voltage signal synchronized with bending and recovery, allowing precise identification of the kicking motions. The consistent correspondence between measured voltage signals and actual motion phases underscores the reliability of SW-TENG for comprehensive swimming posture analysis. The results validated the dual functionality of SW-TENG as both a robust energy harvester and a self-powered motion sensor. The direct outputs (voltage, current, and charge) establish its fundamental triboelectric characteristics, while the signals acquired through the data acquisition module demonstrate its capability to monitor, classify, and analyze complex swimming postures. By distinguishing between swimming strokes, tremors, and help-seeking gestures and by resolving detailed joint kinematics, the SW-TENG proves to be highly promising for applications in intelligent aquatic monitoring, safety assessment, and real-time swimmer rescue systems.

Figure 4.

The output performance and swimming posture monitoring of SW-TENG device

(A–C) Direct electrical outputs of sensors attached to different body parts, showing voltage (A), current (B), and charge transfer (C) from the shoulder, leg, and arm.

(D) Statistical voltage outputs of SW-TENG units placed on the arm, shoulder, and leg under swimming strokes, help-seeking gestures, and tremor-like motions.

(E) Voltage response as a function of bending angle.

(F) Dynamic correlation between voltage signals and bending-recovery cycles.

(G) Schematic of three swimming phases, initial (G1), propulsion (G2), and recovery (G3), with corresponding sensor placements on arms, shoulders, and legs.

(H) Representative voltage and angle signals for arm, shoulder, and leg movements, clearly resolving bending, keeping, flapping, releasing, and recovering stages during swimming actions.

Real-time swimming state detection and classification with SW-TENG and deep learning

The potential of SW-TENG for monitoring swimming motions and recognizing different swimming states was thoroughly explored through a series of experiments designed to assess its performance and classification accuracy in real-world scenarios. The study focuses on detecting distinct swimming states, including normal swimming, help-seeking gestures, and tremor-like motions, with SW-TENG integrated into wearable systems for real-time posture and motion recognition. The data collected from different body parts, including the arm, shoulder, and leg, were analyzed for pattern recognition and classification using deep learning techniques. During the swimming motion experiments, the electrical signals from the SW-TENG sensors were recorded using an external data acquisition system via wired connections under high-impedance measurement conditions. This configuration was adopted to ensure reliable and high-fidelity signal collection for subsequent signal analysis and classification. The wireless transmission module shown in the system schematic is intended to illustrate a potential integrated system architecture and was not used for real-time data acquisition in the experiments. Figure 5A shows the visual representation of the three states being monitored. “Swimming state” (Figure 5A1) involves standard swimming strokes, “help state” (Figure 5A2) corresponds to distress or emergency situations where a swimmer raises a hand or signals for assistance, and “tremor state” (Figure 5A3) refers to minor muscle tremors or vibrations that could indicate exhaustion or disorientation. These states are captured through the wearable SW-TENG sensors positioned on key joints such as the arm, shoulder, and leg. “Three strokes” denotes a representative swimming segment consisting of three consecutive stroke cycles, which is used as a temporal motion interval for signal comparison, rather than a specific swimming stroke type. For wearable swimming motion monitoring, the sensing device is required to operate reliably under continuous water immersion, complex body deformation, and dynamic fluid disturbance. Under such conditions, structural adaptability and deformation-dominated signal generation are more critical than maximizing peak electrical output. The coaxial tubular configuration adopted in this work enables the triboelectric contact-separation behavior to be primarily governed by body motion-induced deformation, rather than external water flow or pressure fluctuations. The concentric geometry provides mechanical symmetry and uniform stress distribution during bending and stretching, contributing to stable and repeatable signal patterns in an aquatic environment. Moreover, the compact tubular structure facilitates conformal integration with the human body and minimizes interference from surrounding water, making it particularly suitable for swimming motion monitoring where reliable motion-induced signals are required. Figure 5B shows the voltage signals from different body parts during swimming, where the voltage outputs were relatively stable across all testers, indicating that the device reliably captures normal swimming motions. Figure 5C displays the voltage outputs during the help state, where the generated voltage was slightly lower than that during swimming, suggesting that the device can detect deviations in motion characteristic of a distress signal. Figure 5D presents the voltage data from the tremor state, where the voltage signal was noticeably reduced, which is indicative of the device’s sensitivity to weaker or irregular movements associated with fatigue or trembling. The consistent voltage outputs across testers confirm the robustness of the SW-TENG in capturing different motion states reliably. Figure 5E illustrates the methodology employed to process and classify the data. Initially, the sensor data are transmitted to a processing module, where feature extraction and selection are performed. These features are then fed into a deep learning model based on convolutional neural networks (CNNs) for classification. The training process utilizes pre-processed data from known motion patterns (swimming, help, and tremor states), and the model is subjected to multiple training cycles to optimize its performance. Once trained, the model can predict unknown motion states based on the collected sensor data, enabling real-time monitoring and detection of swimming gestures and emergency states. Figure 5F shows the voltage responses from different body parts (LA, RA, LS, RS, LN, and RN) during the three states. The voltage data revealed that certain body parts, such as the arms and shoulders, generate more pronounced signals during swimming and help-seeking motions, while the leg sensors showed less variation across states. This information highlights the importance of sensor placement in optimizing system performance for detecting specific motion states. The body parts associated with larger motion amplitudes (e.g., arms and shoulders) tend to provide stronger and more distinguishable signals, which is crucial for accurate state recognition. Figure 5G presents the architecture of the deep learning model used for motion recognition. The input layer receives sensor data from different body parts, which are then processed through hidden layers to extract relevant features. The model classifies the motion states into three categories: swim, help, and tremor. This architecture leverages the strong signal variations associated with each state to enable robust classification. The output layer provides a prediction of the swimmer’s current state, facilitating real-time feedback and potential integration into wearable safety systems for monitoring and assistance. Finally, the results are presented in the form of a confusion matrix, which provides a comprehensive summary of the model’s prediction accuracy, highlighting the classification performance and potential areas for improvement in distinguishing between different motion states. The CNN-based classification results in Figure 5H are presented as a proof-of-concept demonstration of swimming state recognition using self-powered sensing signals. The high classification accuracy observed for the swimming state was achieved under controlled experimental conditions with a limited number of participants, where the signal patterns were relatively distinct. It should be noted that the dataset size was modest, and the analysis was performed without subject-independent cross-validation. Therefore, the current results mainly demonstrate the feasibility of combining self-powered motion sensing with data-driven classification, and further validation using larger and more diverse datasets will be required to assess robustness and generalization in future studies. This demonstrates the effectiveness of the deep learning approach in distinguishing between different motion states based on the data collected from the SW-TENG sensors. The results further emphasize the capability of the SW-TENG for precise and real-time recognition of swimming-related motions and emergency states. The swimming motion monitoring experiments in this study were performed under shallow water conditions, representative of typical swimming activities near water surface. Under such conditions, the variation in hydrostatic pressure is minimal, and the gap between the silicone tubes is dominated by motion-induced mechanical deformation rather than external water pressure. The elastic silicone materials and coaxial tubular configuration provide sufficient mechanical resilience to maintain stable contact-separation behavior during swimming motions. At greater depths, hydrostatic pressure may influence the device geometry and electrical output, which will be investigated in future work aimed at deep water applications. The deep learning-based signal classification presented in this work was performed as an offline data analysis approach, aiming to demonstrate the potential of combining self-powered sensing signals with intelligent motion recognition algorithms. The classification and signal processing were not implemented as on-board real-time operations powered solely by the SW-TENGs in the current system demonstration. This strategy allows the feasibility of intelligent swimming motion recognition to be evaluated without imposing unrealistic power requirements on the energy harvesting capability of the SW-TENGs.

Figure 5.

Application of SW-TENG in swimming motion monitoring and state recognition

(A) Illustration of the three motion states being monitored: swimming, help-seeking gestures, and tremor-like motions.

(B–D) Voltage outputs from SW-TENG sensors placed on different body parts (shoulder, arm, and leg) during the three states, showing distinct signal characteristics: stable output during swimming (B), reduced output during help-seeking (C), and minimal output during tremor states (D).

(E) Overview of the data processing pipeline.

(F) Voltage outputs from different body parts (LA, RA, LS, RS, LN, and RN) during each motion state, highlighting differences in signal amplitudes for accurate state recognition.

(G) Deep learning model architecture for motion state classification, with sensor data processed through CNNs to predict swimming, help, and tremor states.

(H) Confusion matrix showing the classification accuracy of the deep learning model.

In summary, we present a “single-wire structure” SW-TENG integrated into a life jacket for real-time swimming safety monitoring. The SW-TENG, composed of two silicone tubes with conductive copper coating and a single wire, demonstrated effective energy harvesting and reliable performance in aquatic environments. The device achieved V_OC of 57.2 V, Q_SC of 30.1 nC, and I_SC of 3.5 μA under 25 N load and a 2 Hz working frequency, which highlight its ability to function efficiently in real-world applications. Integrated into a swimsuit, the SW-TENG could successfully monitor multi-joint motion, with high sensitivity to angular variations, and distinguish between different swimming behaviors. The SW-TENG device provides real-time, self-powered monitoring through wireless data transmission, ensuring continuous tracking of swimmer safety and performance. Furthermore, by employing deep learning algorithms for real-time classification, the SW-TENG accurately recognized normal swimming, help-seeking gestures, and tremor-like movements, which demonstrate its potential for emergency state detection and intelligent motion recognition in aquatic environments. The results validate the SW-TENG’s effectiveness as both a robust energy harvester and a self-powered motion sensor, making it a promising solution for wearable aquatic safety systems. Future developments will focus on enhancing waterproofing, sensor sensitivity, and signal stability to improve the device’s overall performance and expand its applications in diverse aquatic environments.

Limitations of the study

This study provides a proof-of-concept demonstration of SW-TENG-based swimming motion sensing and motion-state classification. However, several limitations remain. The human experiments were performed with a limited number of participants and under shallow water conditions, which restrict evaluation of inter-subject variability and applicability to deep water conditions. In addition, the CNN-based classification was conducted offline on a modest dataset without subject-independent cross-validation, and, thus, further validation is needed to assess robustness and generalization. Moreover, the current experiments relied on external wired signal acquisition, and the system has not yet achieved fully integrated on-board real-time processing powered solely by the SW-TENG. Future studies will address these limitations through larger-scale testing and deeper system integration.

Resource availability

Lead contact

Further information and requests for resources should be directed to and will be fulfilled by the lead Contact, Wei Sun (sunweity@zzu.edu.cn).

Materials availability

This study did not generate new unique reagents.

Data and code availability

• •Data: The raw and processed data supporting the findings of this study are available from the lead contact upon reasonable request.
• •Code: Custom code used for data processing and CNN-based offline classification is available from the lead contact upon reasonable request.
• •Other items: Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon reasonable request.

Acknowledgments

This work was supported by Postgraduate Education Research Project of Zhengzhou University (YJSJY2025030) and Zhengzhou University General Education Core Curriculum Project (2025ZZUTSKC43).

Author contributions

Conceptualization, W.S.; methodology, W.S.; investigation, data curation, formal analysis, visualization, and writing – original draft, L.Z.; writing – review and editing, supervision, funding acquisition, and project administration, W.S.

Declaration of interests

The authors declare no competing interests.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Chemicals, peptides, and recombinant proteins
Silicone tube (inner tube, OD 2 mm, ID 1 mm)	Commercial supplier	N/A
Silicone tube (outer tube, OD 5 mm, ID 4 mm)	Commercial supplier	N/A
Copper wire	Commercial supplier	N/A
Conductive copper paint	Commercial supplier	N/A
Silicone adhesive	Commercial supplier	N/A
Critical commercial assays
Raw and processed swimming-motion voltage data	This paper	Available from the Lead Contact upon reasonable request; N/A
Software and algorithms
Custom code used for data processing and CNN-based offline classification	This paper	Available from the Lead Contact upon reasonable request; N/A
Other
High-precision electrometer (Keithley 6514)	Keithley Instruments	Model 6514
Programmable linear motor	Commercial supplier	N/A
Digital force transducer	Commercial supplier	N/A
Environmental chamber with integrated temperature and humidity regulation	Commercial supplier	N/A
Arduino Pro Mini (ATmega328P, 3.3 V, 8 MHz)	Commercial supplier	Arduino Pro Mini
STM32F103C8T6 development board	STMicroelectronics	STM32F103C8T6
nRF24L01+ wireless transceiver module	Nordic Semiconductor	nRF24L01+
ESP32-C3 module	Espressif Systems	ESP32-C3
Schottky diode SS14	Commercial supplier	Cat#SS14
Low-dropout regulator AMS1117–3.3	Commercial supplier	Cat#AMS1117–3.3

Experimental model and subject participant details

Swimming motion experiments were conducted with healthy adult volunteers (n = 1). All participants were adults with no reported musculoskeletal disorders that could affect swimming motion. The experiments were performed in shallow-water conditions representative of near-surface swimming activities. No personal identifying information was collected during the study. The experiments involved non-invasive motion monitoring using wearable sensors and did not include any medical intervention. According to local institutional guidelines, formal ethical approval was not required for this type of non-invasive motion study.

Method details

The preparation process of SW-TENG device

The SW-TENG device was fabricated using a coaxial configuration consisting of two silicone tubes and conductive electrodes. A small silicone tube (Length: 8 cm) with an outer diameter of 2 mm and an inner diameter of 1 mm was first prepared to host the inner electrode. A copper wire was inserted along the central axis of the tube, ensuring straight alignment and reliable electrical contact at both ends. The exposed wire tips were reserved for subsequent electrical measurements and external circuit connections. An outer silicone tube with an outer diameter of 5 mm and an inner diameter of 4 mm was then selected to serve as the encapsulation and triboelectric layer. The inner surface of this tube was uniformly coated with a thin layer of conductive copper paint, which acted as the external electrode. The coating was applied through rotary brushing and allowed to dry naturally at room temperature for 12 h, forming a continuous and adherent conductive layer. Subsequently, the small silicone tube containing the wire electrode was coaxially inserted into the larger copper-coated silicone tube to form the integrated SW-TENG structure. The ends of the device were sealed with silicone adhesive to prevent liquid infiltration and to maintain structural stability under aquatic conditions. The resulting device demonstrated excellent flexibility, capable of bending, twisting, and knotting without damage to its functional components. This simple fabrication process ensures reproducibility and scalability while maintaining stable triboelectric output performance. In the experimental demonstration, the SW-TENG sensors are integrated into a tight-fitting swimsuit-type wearable garment to enable stable attachment to body joints during swimming motion tests.

Characterization and measurements

The output characteristics of the SW-TENG, including V_OC, I_SC, and Q_SC, were characterized using a high-precision electrometer (Keithley 6514). To generate controllable and reproducible mechanical stimuli, a programmable linear motor was employed to drive periodic contact–separation motions with tunable frequency and applied force. To obtain real-time feedback on the loading process, a digital force transducer was incorporated to record instantaneous contact forces, thereby ensuring accurate correlation between mechanical excitation and electrical response. All experiments were carried out inside a specially designed environmental chamber equipped with integrated temperature and humidity regulation. The chamber allowed precise adjustment of ambient humidity to stabilize surface charge retention, while the temperature was held constant to eliminate thermal disturbances. This experimental platform provided a reliable means to assess triboelectric outputs under well-defined mechanical and environmental conditions, enabling comprehensive evaluation of device performance in various operational scenarios. The system was implemented using commercially available electronic modules to ensure reproducibility. A low-power microcontroller (Arduino Pro Mini, ATmega328P, 3.3 V, 8 MHz) or an STM32F103C8T6 development board was used as the main controller. Wireless data transmission was achieved using a nRF24L01+ (2.4 GHz) transceiver or an ESP32-C3 module. The AC output of the SW-TENG was rectified by a full-wave bridge based on Schottky diodes (SS14 or SK14). Voltage regulation was provided by a 3.3 V low-dropout regulator (AMS1117–3.3 or MIC5205–3.3). All components are standard, low-cost modules readily available from common electronics marketplaces (e.g., Taobao), enabling straightforward reproduction of the system architecture.

CNN-based offline signal classification method

The convolutional neural network (CNN) used in this work is designed for one-dimensional time-series signal classification. The input to the CNN consists of raw voltage time-series signals generated by the SW-TENGs, without manual feature extraction. Each input sample corresponds to a fixed-length signal segment obtained from a single swimming motion event. A total dataset comprising several hundred signal samples was constructed from repeated swimming motions. The dataset was randomly divided into a training set (approximately 80%) and a testing set (approximately 20%) for offline analysis. No online or on-board training was performed. The CNN architecture consists of three one-dimensional convolutional layers, each followed by a rectified linear unit (ReLU) activation and max-pooling operation, enabling automatic extraction of temporal features from the raw signals. The extracted features are then flattened and passed to two fully connected layers, with the final layer employing a softmax function for multi-class classification. This lightweight CNN architecture provides sufficient representation capability for swimming motion classification while avoiding excessive model complexity, which is suitable for demonstrating the feasibility of intelligent analysis based on self-powered sensing signals.

Quantification and statistical analysis

All experiments were repeated independently multiple times to ensure reproducibility. Electrical output signals (V_OC, I_SC, and Q_SC) were analyzed to compare relative trends under different motion states and mechanical conditions. Due to the proof-of-concept nature of this study and the limited number of participants, no formal statistical hypothesis testing was performed. Signal processing and data visualization were conducted using standard scientific software. Details of data representation, including definitions of signal amplitudes and measurement conditions, are provided in the corresponding figure legends.

Published: April 11, 2026