Abstract
With the increasing demand for daily gait monitoring devices, this study presents an intelligent insole equipped with an array of elastomeric optical fibers that serves as an affordable, flexible, and accurate foot pressure sensor. As the weight-bearing foot compresses the soft optical fibers connected to light sources, the amount of transmitted light decreases proportionally to the load. Fiber optic-embedded insole prototyped using 3D printed mold and polyurethane cast foam in various foot sizes. During walking, the raw signal from our optimized prototypes showed a satisfactory correlation with the signal trend of a clinical-grade commercial sensor (r2mean = 0.84). The correlation was significantly improved (r2mean = 0.945) by implementing a simple machine learning model. The capabilities of the sensor extended to predicting the center of force of plantar pressure and tracking the toe-walking pattern, one of the most common abnormal gaits in children with autism spectrum disorder.
The demand for sensors to monitor daily gait patterns is growing rapidly driven by increasing awareness of fitness and home healthcare through wearable technologies. Consumers can also take advantage of gait monitoring devices to improve physical activity performance, posture, or virtual/augmented reality experiences (Stoltenberg et al., 2019). Monitoring dynamic foot biomechanics such as pronation and supination can help correct the wearer's walking/running/loading posture and prevent ankle injuries (Cain et al., 2007). In addition, gait-monitoring devices can improve the safety and quality of life for individuals with gait abnormalities, such as children with developmental disorders like cerebral palsy (Armand et al., 2016), one-third of people over the age of 70 (Verghese et al., 2006), as well as stroke survivors. For example, approximately 20% of children with autism spectrum disorder (ASD) exhibit toe walking, which is bilateral walking using only the forefoot or toes without heel strike at the beginning of a stance (Ming et al., 2007). Persistent toe walking can tighten the heel cords, limit ankle dorsiflexion, and lead to compensatory abnormal gaits (Barrow et al., 2011). Furthermore, tightened calf can develop into ankle equinus, which is one of the most prevalent symptoms (∼70%) among the clinical adult population, limiting ankle range of motion (DeHeer et al., 2021). Among the elderly, abnormalities in postural balance and gait are significantly associated with serious injury and even death (Verghese et al., 2006).
Gait analysis is essential for effective treatment and evaluation (Armand et al., 2016). However, it is mostly available in laboratory settings with trained clinicians and expensive medical-grade equipment. A sensory system becomes much less meaningful if it cannot track the gait pattern of a sports player during a game, detect the rapidly declining mobility of an elderly person, or monitor the impact of daily practice of a child with a developmental disorder who has limited time to receive effective intervention. Inertial measurement unit (IMU) sensors have been widely explored for wearable gait tracking systems owing to their compact size and light weight (Zhang et al., 2019). Though IMUs can serve as reliable measurement tools for foot progression angles, they cannot capture detailed pressure distribution that the sole experiences. Also, they often need to be placed on multiple body locations such as ankle, knee, thigh, and torso, and are susceptible to noise caused by foot strikes (Shull et al., 2014). The plantar pressure sensor, on the other hand, is relatively more useful and comfortable for everyday gait tracking when it is part of a shoe, insole, or sock. Most plantar pressure sensors with high resolution and accuracy are not affordable for the general public ($5,000 - $15,000) and are cumbersome due to heavy and rigid components and cords (F-Scan System, n.d.). In addition, most foot pressure sensors introduced in academia and the marketplace use resistive (Ershadi et al., 2021) or capacitive (Chen et al., 2021) systems that must be insulated to prevent short circuits or false readings caused by sweat or body heat (Subramaniam et al., 2022). Thin dielectric films that provide insulation not only lack resilience to various external forces such as repeated compression, folding, and shear stress, but also create a slippery, non-breathable, and rigid interface that is not conducive to user comfort and safety (Ahroni et al., 1998). There are affordable products (∼$300) on the market, but they provide limited information based on a few sensing nodes with repeatability issues (Karki & Lekkala, 2006).
Fiber optics have been introduced as an alternative wearable sensor due to the immunity of light to environmental changes in humidity, temperature, chemical conditions, and electromagnetic interference (Lee, 2003). While PMMA (polymethyl methacrylate) optical fibers have been adopted for wearable applications instead of the traditionally popular but highly fragile silica fibers, they are still rigid and not ideal for the dynamic movements (Zha et al., 2024). Despite relatively lower light transmittance, elastomeric optical fibers can be more suitable for on-body sensing systems because the human body requires mechanical resilience to external forces such as strain, rather than long-distance light coverage (Ma et al., 2002). Furthermore, unlike the non-breathable films used in flexible printed circuit boards (PCBs), fiber optics do not cover the entire surface, which allows greater breathability when they are integrated into textiles, foams, or lattices. The thin and lightweight form factors are suitable for integration into textiles for wearables, such as weaving, knitting, embroidering, and/or threading (Gong et al., 2019). Proper cladding with a low refractive index can increase light transmission and provide mechanical protection to the soft core. Methods for measuring different aspects of light can track changes around the human body including foot pressure, and techniques such as Fiber Bragg Gratings (FBGs) have demonstrated advantages in accuracy and convenience when using a single fiber, but they are associated with large and bulky electronic units (Domingues et al., 2017). In contrast, collecting only one or a few measurements such as light intensity or color keeps the system lightweight and small, while remaining effective for monitoring strain, pressure, and bending in the body and textiles (Bai et al., 2020; Harnett et al., 2017; Jo & Park, 2024).
The present study aimed to explore the potential of elastomeric optical fibers for a low-cost plantar pressure sensor for moderately active individuals in the general population who require regular monitoring for possible gait abnormalities. Elastomeric optical fiber offers advantages in mechanical resilience in various scenarios compared to the commercial film-type devices, provides immunity to heat and humidity, and demonstrates expandability for embedding in various soft materials, including foams and textiles. The current study investigated the feasibility of elastomeric fiber optics for gait monitoring devices through laboratory experiments and human participant testing. Specifically, the experiments aimed to determine if the sensor could measure plantar pressure and center of force of the foot at the level of clinical commercial sensors. The study demonstrated that the elastomeric fiber optic-based insole device was capable of detecting toe walking, a hallmark of abnormal gait patterns. Because the fiber optic sensor is durable and cost effective, this device shows promise for benefiting individuals with gait abnormalities as well as those interested in monitoring lower-body mobility. Additional applications leveraging its pressure-sensing capabilities were also discussed.
Methods
Design and Prototype
Sensor Structure and Sensing Principle
As the main sensing material, a soft, elastic, lightweight, thin, and transparent polyurethane cord (Stretch Magic, Pepperell Braiding Co., Pepperell, MA) was used as an elastomeric fiber optic. The fiber exhibited sufficient mechanical properties to be part of a wearable application, including extensibility (> 500% ultimate strain, Harnett et al., 2017), manageable hysteresis to repetitive pressure (Jo et al., 2022), and durability after machine washing (Jo & Park, 2024). A red LED (light emitting diode; produced by Kingbright, Walnet, CA) provided the light with a wavelength of 625 nm at one end of the fiber, considering the low absorption of near infrared light (Harnett et al., 2017). At the other end, a photodiode (SFH 229, ams OSRAM, Austria) was connected to the fiber according to a previous practice (Xu et al., 2019). The fiber (refractive index = 1.52; Harnett et al., 2017) was not clad but surrounded by the EVA foam (refractive index is typically less than 1.5 at 625 nm; Shamim et al., 2018), which may reduce light leakage.
When the foot pressure during the stance phase of the gait cycle compresses not only the insoles but also the embedded soft fibers, the transmittance of the light is reduced because the cross-section of the fiber is deformed from circular to elliptical (Figure 1). While the cross-sectional deformation was the main source of the signal changes, the change in curvature of the fiber from straight to curved by the deformed insole was the secondary cause of the light intensity responses.
Elastomeric Fiber Optic-Based Pressure Sensing Insole. (a) Design of the Device, (b) Principle of Sensing.
Design Factor Exploration
A number of design factors, from the fiber diameter to the thickness of the foam, were investigated to achieve the best end performance. Commercial elastomeric fibers (The diameter was 1.5 mm unless otherwise noted. All fibers (produced by Pepperell Braiding Co., Pepperell, MA as Stretch Magic) were embedded in a 2*2*2 cm3 EVA foam cube for lab testing. The sensor-embedded cube was compressed with loads of 0, 5, 10, 15, and 20 kg for two seconds, representing common in-shoe plantar pressure peaks of up to 2–3 kg/cm² (Owings et al., 2009).
Fiber diameter affects light transmission up to ∼30 cm and embeddability in ∼2 cm thick insoles, suggesting that fibers that are too narrow or too thick are not feasible as part of the gait sensor. The results showed that when the fiber diameter was between 1–1.8 mm, there was no noticeable difference in pressure sensing capability between the fibers with different diameters except for the 1.2 mm fiber (Figure 2a). Though the 1.2 mm diameter fiber displayed the highest sensitivity, the thickest 1.8 mm diameter fiber, which showed fair sensitivity compared to 1.2 mm and others, was selected for the final prototype to ensure sufficient light transmission on the photo-detecting side.
Normalized Light Signal Changes by Design Factors and Cyclic Compressions. (a) Fiber Diameter, (b) Number of Fibers Installed into a Foam Cubea, (c) Foam Thickness, and (d) Cyclic Compressionb. Note. All Samples, otherwise indicated, used a fiber optic with 1.5 mm diameter inserted into a 2*2*2 cm foam cube that was compressed by 62% when 10 kg load was applied. The red line shows the materials used in the final prototype. a For 2 fiber samples, adjacent fibers were perpendicular with the even spacing from the edge. For example, the 2 fiber samples had 0.5 cm spacing from the edge to the top fiber (“2-Top”), 0.5 cm between the top fiber (“2-Top”), and the bottom fiber (“2-Bottom”), and 0.5 cm between the bottom fiber and the bottom edge. b The samples were compressed at 5, 10, and 15 kg / 2 s for 100 cycles respectively, not sequentially.
Plantar pressure is two-dimensional: during stance, the center of pressure moves from the rear of the foot to the forefoot, while it also moves between the lateral and medial sides of the foot depending on the shape of the arch and the gait style. Therefore, a set of parallel fibers may not be able to monitor the two-dimensional movements in the plantar pressure plane, and a fiber array with two directions that track both front-back and medial-lateral movements is inevitable. Figure 2b shows the signal change when two 1.5 mm-diameter fibers are installed in a 2*2*2 cm3 foam cube. For example, 2-Top, and 2-Bottom indicate the two fibers embedded in a foam as shown in the image in Figure 2b. The two-fiber samples exhibited monotonic signal trends that can correspond to a pressure level like the single-fiber sample. Meanwhile, they exhibited greater signal changes than the single-fiber samples, which implies that the fibers, less compressible than the foam, may compress each other and cause light leakage at the intersection (Jo et al., 2022). The different signal trends between the upper and lower fibers indicate that the fibers need to be treated individually to be directly translated into plantar pressure.
To prevent excessive denting when the intersecting fibers are too close together, a buffer layer between the fibers is necessary. However, the buffer layer will increase the overall thickness of the insole, which may affect user comfort. As an exploration to decide the foam thickness for the prototype, Figure 2c shows that if the foam is too thin, the 1.5 mm-diameter fiber can directly bear the load from the foot and the light loss will increase significantly due to the accelerated fiber deformations. It is possible to calibrate such signal changes, but it is still unfavorable because the user may feel the fiber on the bottom of the foot. On the other hand, increasing the foam thickness did not affect the signal change trend much. It is assumed that the light attenuation by the compressed foam has reached a saturation point where the compressed foam no longer deforms the fiber, which needs to be verified by more detailed investigation of the mechanical interaction.
To verify the signal recovery of the elastomeric fiber optic under repeated compression, 100 cycles of 5, 10, and 15 kg load were applied for 2 s to the single-fiber foam sample (Figure 2d). The result showed that the peak signal experienced variations (standard deviation: 1.74, 1.72, and 2.22% for 5, 10, and 15 kg, respectively) and the signal at rest was slightly shifted (2.80, 0.55, and 5.02%, respectively). Though no signal variation or drift would be ideal for reliability, this still falls within the usable range, compared to the 19% creep and 21% hysteresis observed in one of the most widely used clinical plantar pressure devices (Woodburn & Helliwell, 1996). Nevertheless, sensor readings need to be monitored during daily use and recalibrated as needed.
Insole Prototyping
Based on the exploration of the design factors, the gait monitoring insole was designed and prototyped using elastomeric fiber optics (Figure 3).
Prototyping Process.
The mold for the ergonomic casting of the insole was 3D printed in three different sizes based on the foot length. After connecting the LED and photodiode to each end of the fiber optic, the eight fibers (5 for anterior/posterior motion and 3 for right/left) were placed in the mold at critical points on the foot pressure plane, which can be used to derive a foot pressure map with higher resolution (Mun & Choi, 2022). The two fiber arrays were spaced so that they did not touch or create indentations. A flexible polyurethane foam (FlexFoam-iT!™ IV Tuff Stuff, Smooth-On, Inc., Macungie, PA) was poured into the mold and cured. After curing, the light source/detector electrodes were soldered to the custom-designed flexible PCB. The PCB included a 330-ohm resistor to protect each LED, and a 500-ohm potentiometer connected to each LED to adjust the initial amount of light to prevent complete loss or plateauing of light intensity. The flexible PCB was connected directly to a ribbon cable connector installed on a hard PCB containing a microcontroller with wireless communication capability (Arduino Nano 33 BLE). The hard PCB also had I-V converters (Xu et al., 2019), which converted the currents generated by the photodiodes upon light input into voltage changes that could be sensed at the analog pins of the microcontroller. The whole system was powered by a 3.7 V lithium battery, which can be charged by a circuit embedded in the hard PCB.
User Evaluation
Experiment Setting
In order to obtain ground truth foot pressures, a widely-used commercial thin-film clinical plantar pressure sensor (F-Scan System, Tekscan, Norwood, MA) was utilized in this study, along with our sensor. Though the prototype described in Figure 3 is ideal for creating an ergonomic 3D shape with arch support and installing electronics, it was not compatible with the F-Scan system, so the same fiber optics were directly embedded in a flat EVA foam sole of a sandal. Right-sided sandals in three different longitudinal sizes were prototyped to accommodate the different foot lengths of the human participants. The commercial sensor was placed between the foot and the sensor-embedded sole.
A total of 11 healthy adults (gender: 8 females and 3 males, age: 36.5 ± 12.1 years, height: 168.9 ± 9.0 cm, weight: 66.4 ± 21.2 kg, BMI: 23.4 ± 7.9, foot length: 248.2 ± 16.6 mm) participated in the study after agreeing to the protocol approved by the Institutional Review Board (IRB). Although the present study aimed to develop an affordable sensor for populations with gait abnormalities, the prototype was tested only on healthy adults due to safety concerns. Participants wore the given prototype according to their foot length size along with the commercial sensor. To minimize random error and increase the reliability of the data in the performance evaluation, the 11 participants were asked to perform each movement three times, resulting in 33 data sets. Participants were asked to walk (60 s * 3 sessions, Appendix A), run (60 s * 3 sessions), stand still with eyes closed (30 s * 3 sessions), squat (5 times * 3 sessions), and tiptoe (60 s * 3 sessions) on the treadmill in an indoor laboratory environment. All participants were able to adjust the walking/running speed as they wished. For the tiptoe walk, they randomly mixed normal walking (heel to toe) and tiptoe walking (forefoot only) for 60 s. The task of standing still with eyes closed was included as one of the simplest balance tests, as well as for the purpose of seeing the residual force when the person is not moving.
Data Acquisition and Processing
Prior to each human participant test, the commercial sensor was calibrated according to the manufacturer's instructions. The commercial sensor collected two-dimensional foot pressure data with a resolution of 100 Hz, which is the required sampling frequency for accurate running mechanism measurements (Mann et al., 2016). Since the proposed sensor has a two-dimensional array of fibers, it is possible to generate a two-dimensional pressure map based on the intersection points. However, the common practice in gait analysis is to eventually group the individual point pressures in the longitudinal/transverse direction to identify gait abnormalities associated with the axial movements. Therefore, from the commercial sensor, the sum of the point pressures on the linear area where each fiber optic is located was collected as the ground truth to be compared with the signal from the fiber optic sensor. The microcontroller collected the light intensity through the analog pin at 100 Hz. Before starting the tasks, the participants were asked to lift their legs to measure the residual force not generated by their body weight. The sensor readings from each fiber optic sensor were normalized respectively. After joining the two datasets (the commercial sensor and the prototype) in terms of time, the 30% of the dataset (approximately 3,000 data points per training) of each movement of each participant was used to train a neural network MLP (multilayer perceptron) Regressor model of the Scikit-learn library (Sklearn.Neural_network.MLPRegressor, n.d.). The MLP model was used to reduce the computational load instead of more advanced but heavier models, such as the LSTM (long short-term memory) model specialized for time series datasets. The input was the signals from the eight fiber optic sensors, and the expected output was the ground truth value of the target location from the commercial sensor. The model was trained with the Adam optimizer to convergence or a maximum 1,000 iterations. The hidden layer size was 64*64 and the activation function was ReLU. Though ReLU is piecewise linear, it has been used in nonlinear applications including gait patterns (Horst et al., 2019). Its computational cost is low, which makes it suitable for low-budget computing. The test result using the other 70% of the dataset was described in this paper, with the label of “Prediction” and there was no validation set.
Results
Foot Pressure Monitoring
Walk
The raw light intensity signals and the MLP prediction matched the ground truth foot pressure trends. Figure 4 shows the foot pressure changes at each sensor location of a stance. Five sensors monitored the anterior/posterior motion (H1–5), while three sensors monitored the right/left movements (V1–3). While the rise, peak, and fall of the raw signals (blue dotted lines) agree well with the plantar pressure trend from the commercial sensor, the MLP model trained by all eight sensors satisfactorily predicted the characteristic changes in trend at each location.
Sensor Performance. (a) Local Foot Pressure Measurement During Stance. (b) Ground Truth from the Commercial Sensor, Predicted Foot Pressure from the MLP, and the raw Signal from the Sensor of a Walking Stance. (c) Correlation Between the Results of the Fiber Optic Sensor and the Commercial Sensora. a The bar graph represents the root mean square error (RMSE) of the MLP prediction with respect to the ground truth at each location. Error bars are standard errors.
Overall, the prediction from the MLP model showed a high correlation with the foot pressure values from the commercial sensor (Figure 4c). The correlation coefficient (r2) of the eight locations for the walking tasks was high (walk: r2 = 0.945 ± 0.019). The correlation coefficients (r2) of the raw signal with the commercial sensor data were also satisfactory (Meanwalk = 0.84, SDwalk = 0.06). The foot pressure prediction showed an average root mean square error (RMSE) of 14.48 KPa (SD = 10.92 KPa).
Other Lower Body Movements
In addition to walking, the sensor with MLP model could accurately predict other lower body movements such as running (r2 = 0.919 ± 0.024) and squatting (r2 = 0.709 ± 0.069) as shown in Figure 5. While the H4 sensor showed the lowest correlation during running (r2 = 0.892), the H5 showed the lowest correlation (r2 = 0.819) because some participants preferred to run with their forefoot. The RMSE of the MLP model for running and squatting was 15.70 ± 10.96 KPa and 10.53 ± 8.61 KPa, respectively.
Foot Pressure Signal Change in Other Activities. (a–b) Ground Truth and MLP Model Prediction for Running and Squat. (c) Ground Truth and Raw Signals for Locations H1 and H2 for Static Standing.
Standing still with eyes closed is a widely used diagnostic tool called the Romberg test, which detects neurological impairment (Forbes et al., 2023). The result for this task showed the sensitivity of the fiber optic-based insole to static posture. The commercial sensor used in the current study filtered out the forces less than 10% of the expected maximum value to remove the residual force unrelated to the gait pattern. As a result, it failed to capture subtle changes in foot pressure, especially when the user was static, unlike the newly developed sensor of this study (Figure 5c). The filtering policy of the commercial sensor partially influenced the lower correlation between the fiber optic insole and the ground truth in both stance (r2 = 0.467 ± 0.116) and squat.
Demographic Factors
The correlation coefficient between the two sensors at H1, H3 and V3 showed a significant relationship with weight (H1: r2 = 0.44, p = 0.01, H3: r2 = 0.45, p < 0.01, V3: r2 = 0.47, p < 0.01). Foot length was also related to sensor performance at H4 (r2 = -0.54, p < 0.01) and V3 (r2 = 0.45, p < 0.01). However, the scatter plots of the coefficients by weight or foot length suggest that the significance may be due to some of the outliers (Figure 6a and 6b), so the causal effect of user's weight or foot length on sensor accuracy should be investigated by testing with a larger scale.
Correlation Coefficients Between Ground Truth and raw Signal by (a) Weight, and (b) Foot Length.
Center of Force
The center of force (COF), one of the most important measures for gait analysis, is the centroid of all forces applied to the plantar surface of the foot (Lugade & Kaufman, 2014). The COF and its trajectory provide valuable information about gait balance and speed, as well as foot morphology (Mizelle et al., 2006). Following the formulas used in the commercial sensor (FAQs | Tekscan, n.d.), the signals less than 10% of the maximum value from the fiber optic sensors were filtered out as residual pressure and processed to obtain
Center of Force (COF) Prediction. (a) Ground Truth from the Commercial Sensor, the MLP Prediction, and the Calculated COF Using the Raw Signals, (b) Correlation Coefficients between the Fiber Optic Sensor (Raw Signal and MLP Prediction) and the Commercial Sensor, and the RMSE of the MLP Prediction Results compared to the Ground Truth.
Case Study: Toe Walking
One of the main symptoms of children with ASD is toe walking, where only the forefoot area is used when walking. To meet their needs for daily gait monitoring and to verify the ability of the sensor to detect abnormal gait patterns, this study asked the recruited healthy adults to randomly mix normal walking and toe walking as one of the tasks. The result showed that when the participant toe walked, there was no response from the horizontal sensors around the rear foot (Figure 8). The three vertical sensors covering the entire foot also showed different pressure trends with a single peak compared to normal walking with double peaks generated at the heel and ball of the foot. The COFy clearly stayed in the forefoot area during toe walking. The result also showed a high correlation between the two sensors (r2pressure = 0.933 ± 0.02, r2COF = 0.944 ± 0.009).
Toe Walking. (Top) Anterior/Posterior Motion, (Middle) Right/Left Motion, (Bottom) COF Changes for Each Stance.
Discussion
In this study, elastomeric fiber optic arrays were embedded in flexible foam-based footwear to track the wearer's plantar pressure. As the foot pressed down on the insole, the elastomeric optical fibers compressed, reducing light transmission. From these signals, the sensor inferred the trajectory of the center of force (COF) of each stance, enabling advanced gait analysis to evaluate dynamic foot functions such as pronation and supination. The sensor also demonstrated the ability to detect tow walking, a common abnormal gait pattern in children with autism spectrum disorder (ASD), by capturing foot pressure trends and COF trajectories.
Implications
The low-cost manufacturing of this sensor (approximately $0.11 per sole based on current fiber retail pricing) suggests potential for an affordable and convenient plantar pressure monitoring device. Such technology could assist younger populations with developmental disorders experiencing gait abnormalities by allowing daily home monitoring, offering immediate and continuous feedback for better correction.
Beyond children with ASD, the device could benefit populations such as stroke patients, the elderly, and those in rehabilitation by supporting fall risk management and lower-body performance improvements. Patients with diabetes who often face serious foot problems due to ill-fitting footwear may also benefit if sensors are embedded not only in soles but also around the upper shoe surface to detect excessive pressure. Furthermore, the sensor could enhance safety of those working in harsh environments, such as firefighting, underwater, or in high electromagnetic fields, by assisting with localization (e.g., the location of the firefighters within the fire site), fall detection, or fatigue tracking. The possibility of incorporating infrared (IR) light LEDs may provide added therapeutic benefits such as muscle relief and recovery acceleration (Selm et al., 2007). Although not investigated in the current study, the user of IR light LED as a light source may be able to provide muscle relief functionality in addition to foot pressure measurement.
The fibrous form factor of the optical fiber opens opportunities for embedding the sensor into other materials. For example, fiber can be woven, knitted, or embroidered into textiles to create gait monitoring socks, pressure sensing garments, or gloves. Applications can extend beyond humans to animals and robots, where mobility and dexterity should be monitored. Moreover, due to its stability against environmental changes and stretchability, this sensor could be adapted for unconventional, uneven, or dynamic soft surfaces to maximize its pressure sensing capabilities.
Limitations
The current study presented a manufacturing process for the ideal ergonomic insole prototype, but conducted user evaluation with a sandal-like prototype to ensure surface compatibility with the commercial sensor. Although ergonomic shape may slight alter results, the sensor behavior is expected to remain consistent. A practical limitation lies in the flexible circuit board housing LEDs and photodiodes because unexpected foot movement may bend or damage the parts and the electrical connections on the PCB. Future designs should incorporate protection against this risk.
Another limitation is the sampling rate of 100 Hz. While sufficient for normal walking and higher than the commercial device used for comparison, a higher rate would be necessary for capturing fast sporting activities or detecting slips and falls (Mann et al., 2016). Additionally, the optical fiber material used in this study was a commercial elastic cord not originally designed for optical sensing. Its manufacturing quality, including uniformity of diameter and the coefficient of variation of transmission, was not evaluated. These factors must be verified to ensure reliability.
Lastly, demographic effects were only briefly examined. Although body weight and foot length were included, other foot morphological variables were not analyzed due to the small sample size. Future studies with larger and more diverse populations should assess performance differences related to gender and foot shape to optimize design elements such as fiber layout, fiber count, and foam compressibility.
Footnotes
Acknowledgements
This study was supported by the IGNITE Gap Funding of Center for Technology Licensing and Doctoral Research Support Program of the College of Human Ecology, both at Cornell University. The authors appreciate the technical support of Herb Susmann for the custom PCB design.
Funding
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the Cornell University (Doctoral Research Support Program, and IGNITE Gap Funding of Center for Technology Licensing).
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Author Biographies
Appendix A. Raw signal of each fiber optic sensor during three walking trials of a participant.