Design and Performance Analysis of a Multiplexed Piezoelectric Sensor Array for Real-Time Impact Detection

Research on sensor-based impact detection systems has advanced rapidly, particularly with the use of piezoelectric sensors, which are known for their rapid response to changes in mechanical stress. These sensors are capable of converting mechanical energy into electrical signals in a very short time, making them widely used in applications such as structural health monitoring and impact detection systems [16-18]. Several studies have shown that piezoelectric sensors exhibit high sensitivity to mechanical stress variations, enabling accurate measurement of vibration and shock events [19]. These characteristics make piezoelectric sensors well-suited for real-time detection systems.

On the other hand, the advancement of wearable systems based on the Internet of Things (IoT) has further driven the integration of sensors into devices that can be directly used by end users. Wearable systems enable data acquisition to be performed directly at the point of occurrence, resulting in more contextual data compared to conventional methods. However, the implementation of wearable systems with a large number of sensors still faces limitations, particularly in terms of data acquisition and hardware resources [17]. To address the constraints of limited ADC channels and microcontroller resource management, analog multiplexing techniques are commonly employed to route multiple input channels sequentially to a single ADC module [20]. Although efficient in terms of pin usage, most existing literature has not thoroughly addressed how to maintain data integrity when dealing with piezoelectric transient signals characterized by very fast rise times and impulsive amplitudes.

In addition, signal conditioning is a critical aspect in sensor-based systems. Piezoelectric signals, which are impulsive and high in amplitude, may introduce noise, distortion, and potential damage to electronic components if not properly managed [9]. Therefore, interface circuits such as charge amplifiers are required to convert electrical charge into voltage signals that can be further processed by data acquisition systems [21]. Several studies have proposed the use of RC filters and voltage protection as basic solutions. However, the integration of voltage protection, current limiting, and signal retention mechanisms (peak hold) within multi-channel systems has not been comprehensively explored. Furthermore, microstructure-based sensors generally offer high sensitivity but limited detection range, whereas three-dimensional porous structures can extend the detection range at the cost of relatively lower sensitivity. Hence, a more optimal sensing system design approach is still required [22].

Table 1. Comparison of the proposed system with representative previous studies

Note: ADC = analog-to-digital converter, LDRs = light-dependent resistors, TDoA = time difference of arrival, ROI = region of interest, AFE = analog front end, SNR = signal-to-noise ratio, CMOS = Complementary Metal–Oxide–Semiconductor, RC = resistor–capacitor.

Table 1 summarizes the main characteristics of representative piezoelectric-based sensing approaches discussed in the literature. The reviewed studies demonstrate different approaches to impact sensing, including impact-force characterization, structural impact localization, flexible piezoelectric sensing, signal conditioning, and multi-channel readout. These studies show that piezoelectric sensing can support both impact detection and spatially distributed sensing, but the system architectures and evaluation objectives vary according to their respective applications. In contrast, the proposed system combines multi-channel sensing, signal conditioning, sequential analog multiplexing, and sensor-zone localization within a wearable ballistic impact detection platform.

Based on the reviewed studies, including previous SmartVest systems for airsoft applications, the remaining gap is focused on the integration of dense multi-channel piezoelectric sensing with signal conditioning and sequential multiplexed acquisition while maintaining reliable detection of short-duration impact signals. Previous SmartVest implementations demonstrated the feasibility of automated hit detection and remote monitoring, but used a limited number of predefined sensing points and did not address the temporal challenges associated with sequential acquisition of a high-density piezoelectric array.

Therefore, this study proposes a multiplexed piezoelectric sensor array system that integrates signal conditioning techniques and multiplexing-based acquisition methods to support reliable real-time impact detection with high detection accuracy and system stability.

4.1 Impact detection performance

The impact detection performance of the developed SmartVest system was evaluated through controlled ballistic impact experiments conducted at four shooting distances, namely 18 m, 20 m, 22 m, and 25 m. At each distance, ten shots were performed, resulting in a total of 40 impact events. The purpose of this evaluation was to determine whether the proposed multiplexed piezoelectric sensor array system could successfully detect impact events under the tested conditions.

Prior to the ballistic impact experiments, a no-impact baseline observation was conducted to evaluate the sensor output under idle conditions. During this condition, the sensor output fluctuated within approximately 200–800 ADC units. Therefore, the detection threshold was set to 1500 ADC units to provide a sufficient margin above the maximum baseline fluctuation while maintaining sensitivity to low-amplitude valid impact responses. For the 12-bit ADC with a 3.3 V reference voltage, this threshold corresponds to approximately 1.21 V. In comparison, the maximum baseline fluctuation of 800 ADC units corresponds to approximately 0.64 V. This separation indicates that the selected threshold can reduce the possibility of false detection caused by low-amplitude noise or non-impact fluctuations.

An impact event was considered successfully detected when at least one piezoelectric sensor generated a peak ADC value exceeding the predefined threshold of 1500 ADC units. For each impact event, the sensor with the highest ADC value was identified as the primary response, while other activated sensors within the same event were considered secondary responses caused by impact energy propagation.

The impact detection success rate was calculated using Eq. (11). Based on the experimental results, all 40 ballistic impact events generated peak ADC responses above the predefined threshold. Therefore, the detection success rate was obtained as follows:

Detection Success Rate $=\frac{40}{40} \times 100 \%=100 \%$

These results confirm that the proposed system is capable of reliably detecting all impact events under the experimental conditions.

image032.png

Figure 4. Peak analog-to-digital converter (ADC) response of detected impact events across 40 ballistic trials at shooting distances of 18 m, 20 m, 22 m, and 25 m

Figure 4 presents the peak ADC responses obtained from all 40 ballistic impact trials. As shown in the figure, all impact events produced peak ADC values above the detection threshold of 1500 ADC units, confirming that no missed detection occurred during the experiments. Several trials reached the ADC saturation value of 4095, indicating that high-energy impact signals were successfully limited by the signal conditioning circuit without exceeding the ADC acquisition range.

Overall, the proposed system achieved a 100% impact detection success rate under the tested ballistic impact conditions. However, this value represents the detection success rate for actual impact events, not a complete classification accuracy. Further evaluation under longer no-impact conditions and various environmental disturbance scenarios is recommended to quantify the false positive rate more comprehensively.

4.2 Impact localization analysis

In addition to evaluating detection performance, the system was further analyzed based on its capability to determine impact locations and the spatial distribution of impact energy.

The localization analysis was performed by identifying the sensor with the highest amplitude as the primary sensor, representing the impact center. Meanwhile, other sensors activated around the primary point were categorized as secondary sensors, indicating the affected area (impact effect) caused by the propagation of impact energy.

The impact localization results at various shooting distances are presented in Figure 5, illustrating the distribution of sensor activations at distances of 18 m, 20 m, 22 m, and 25 m. Each color in the figure represents a different impact event, allowing the activation patterns among shots to be visually analyzed.

To maintain visualization clarity, Figure 5 presents only several representative shots at each distance, while the complete dataset from all 40 trials was still utilized in the overall system performance analysis.

image033.png

Based on Figure 5, the system is capable of identifying impact locations across various vest zones, including the waist, navel, ribs, and shoulder regions. The detected primary sensor indicates the center of impact, whereas the activation of surrounding secondary sensors reflects the propagation of mechanical energy from the impact point to adjacent areas. These results demonstrate that the system can effectively distinguish the primary impact center from the surrounding energy propagation effects.

In several impact events, more than one sensor was activated simultaneously, indicating a multi-sensor activation phenomenon caused by impact energy propagation. This pattern provides a more comprehensive representation of energy distribution compared to single-sensor-based detection systems.

Furthermore, the sensor activation patterns exhibit variations across different shooting distances. At shorter distances, the distribution of activated sensors tends to be wider, whereas at longer distances, the sensor activation becomes more localized due to the reduced impact energy received.

These results demonstrate that the SmartVest system is not only capable of detecting impact events, but also of providing a spatial representation of impact energy distribution. Therefore, the system can be utilized as a real-time piezoelectric sensor-based impact mapping solution.

Nevertheless, under lower impact energy conditions, the number of activated secondary sensors tends to decrease. This indicates that the system sensitivity to energy distribution is still influenced by the magnitude of the impact energy received by the sensors, which remains an aspect for further improvement in future studies.

4.3 Scanning performance

Scanning performance is one of the critical aspects of the proposed system because the signals generated by piezoelectric sensors exhibit transient characteristics with extremely short durations. Therefore, the analog multiplexer-based sequential scanning mechanism needs to provide sufficient temporal response to support the acquisition of transient impact signals.

The implemented acquisition system consists of 37 piezoelectric sensor channels connected through three analog multiplexers with a channel configuration of 16, 16, and 5 channels. During sequential acquisition, a programmed delay of 120 µs is applied after each channel reading. Therefore, the accumulated contribution of the programmed inter-channel delay for 37 channels is 4.44 ms. In addition, a 5 ms peak detection window is implemented to capture and process transient impact responses.

This relationship demonstrates that the number of sensor channels directly affects the temporal burden of sequential acquisition. For the implemented 37-channel configuration, the programmed inter-channel delay contributes 4.44 ms to the acquisition process. The experimental activation timestamps presented in Table 6 are therefore interpreted in relation to the programmed acquisition parameters rather than as a direct measurement of the complete scan-cycle duration.

Table 6. Timing parameters of the implemented acquisition system

The RC interface also affects the temporal response of the conditioned piezoelectric signal. The implemented circuit uses two 1 MΩ resistors and a 100 nF capacitor. Considering the two resistors as the resistive path seen by the capacitor, the equivalent resistance is approximately 500 kΩ, resulting in an RC time constant of approximately 50 ms. This time constant provides temporal persistence of the conditioned signal during the sequential acquisition process.

The calculated RC time constant is directly related to the temporal behavior observed during the multi-sensor acquisition test. With an RC time constant of approximately 50 ms, the conditioned signal is expected to persist for a longer duration than the original short-duration piezoelectric pulse. This temporal persistence provides an opportunity for the sequential acquisition process to detect the signal when the corresponding sensor channel is accessed. The experimentally recorded multi-sensor activation intervals in Table 7 provide an observation of this temporal behavior, with most sensor activations occurring within 0–1 ms and the maximum recorded interval reaching 50 ms.

Reliable acquisition of transient impact signals depends on whether the conditioned signal remains detectable when the corresponding sensor channel is accessed during sequential scanning. In the implemented system, the temporal response is influenced by both the programmed inter-channel delay and the RC conditioning circuit. Therefore, the recorded multi-sensor activation timestamps are used as an experimental indicator of the temporal distribution of the detected impact responses.

To evaluate the scanning-related response of the system, experimental data obtained at a shooting distance of 20 m were used as a representative scenario. This distance was selected because the recorded data showed multi-sensor activation across several areas of the vest surface. The number of activated sensors and the time interval between the first and last recorded sensor activation within each impact event are summarized in Table 7.

Table 7. Multi-sensor activation and time spread per shot at 20 m

Based on Table 7, several impact events activated more than one sensor within a single shooting event. At a shooting distance of 20 m, the maximum number of activated sensors was four, as observed in shot 9. This result indicates that the system was able to record multiple sensor responses within the same impact event. Meanwhile, across all 40 experimental trials conducted at different shooting distances, the maximum number of activated sensors was 5, as summarized in Section 4.7.

Furthermore, the recorded time interval between activated sensors, denoted as Δt, was generally close to 0 ms for most shots. Only shot 6 produced a Δt of 1 ms, while shot 9 produced the maximum observed Δt of 50 ms. These results show that multiple sensor activations were recorded within a short temporal interval for most impact events, indicating that the system could record distributed sensor responses within the same impact event spatially.

Figure 6 illustrates the time spread between the first and last recorded sensor activations for each impact event at a shooting distance of 20 m. Most events exhibited a Δt close to 0 ms, while the maximum observed value was 50 ms in shot 9. This result indicates that the system was able to record multiple sensor activations with different temporal spreads within an impact event.

It should be noted that the Δt value represents the time spread between recorded sensor activations rather than the direct hardware scanning time of the multiplexer. Therefore, Δt is used only as an experimental indicator of the temporal distribution of multi-sensor responses. The timing analysis in this study is based on the implemented acquisition parameters and experimentally recorded sensor activation timestamps. Direct hardware-level measurements of multiplexer switching time, complete scan-cycle duration, ADC acquisition interval, and end-to-end detection latency were not performed and are therefore not presented as measured quantitative results. Direct measurement using an oscilloscope or logic analyzer is recommended for future work to further quantify the hardware-level timing performance.

image037.png

Figure 6. Time interval between activated sensors per shot at 20 m shooting distance

4.4 Signal conditioning effectiveness

The effectiveness of the signal conditioning circuit was evaluated based on the peak voltage values generated by the piezoelectric sensors under various shooting distance conditions. A summary of the minimum, maximum, and average voltage values is presented in Table 8.

Table 8. Summary of peak voltage across shooting distances

Based on Table 8, all measured voltage values remain within the safe operating range of the microcontroller ADC, namely below 3.3 V. This indicates that the voltage protection mechanism operates effectively in suppressing voltage spikes generated by high-energy impact events. Theoretically, the output voltage is limited by the characteristics of the Zener diode, such that the maximum measured voltage can be approximated by the clamping voltage, as expressed in Eq. (13):

$V_{\text {clamp}} \approx V_{\text {zener}} \approx 3.3 \mathrm{~V}$   (13)

This condition is further supported by the measurement results, which show saturation at the maximum ADC value, indicating that the signal reached the acquisition limit without exceeding the safe operating range of the system.

Figure 7 illustrates the relationship between the shooting distance and the average peak voltage generated by the sensors. The graph shows that the voltage values remain relatively stable across different distance variations, without significant degradation. This indicates that the quality of the received signals is well maintained despite variations in impact energy.

In addition, during several impact events, the measured voltage values approached the maximum ADC limit. This indicates that the system is capable of consistently capturing high-energy impact signals.

image039.png

Figure 7. Graph of average peak voltage vs. distance

These results are also consistent with the findings presented in Section 4.3, where multiple sensors could be detected almost simultaneously within a single impact event. This indicates that the transient signals remain sufficiently detectable throughout the scanning process.

Overall, the implemented signal conditioning circuit is capable of maintaining the stability of the piezoelectric sensor output signals, allowing them to be properly processed by the microcontroller under all testing distance variations. Nevertheless, the occurrence of several voltage values reaching the maximum ADC limit suggests that improving the signal acquisition resolution may be considered for future system development.

4.5 Sensor array response

The sensor array response was analyzed to evaluate how the system detects impact locations across different areas of the vest surface. During the experiments, the activated sensors were not always located at the same positions, but instead depended on the impact location generated by each shot. The distribution of sensor activation counts based on body zones is summarized in Table 9.

Table 9. Distribution of primary impact zones across body regions

Based on Table 9, sensor activations are distributed across all vest surface zones, with the highest number of activations occurring in the navel area, followed by the waist and ribs regions. To provide a more quantitative analysis, this distribution can be expressed as the percentage of activations relative to the total number of events. The activation percentage can be defined as shown in Eq. (14):

Activation Percentage $=\frac{N_{\text {activation }}}{N_{\text {tatal }}} \times 100 \%$   (14)

In addition, several impact events resulted in the simultaneous activation of multiple sensors. This condition indicates that the impact energy is not only received by a single sensor, but also propagates to neighboring sensors. Such activation patterns represent the impact effect, where the primary sensor indicates the impact center, while the additional activated sensors (secondary sensors) describe the energy distribution across the surrounding area.

The distribution of sensor activations across each zone is also illustrated in Figure 8. The graph shows that the sensor responses are not concentrated in a single area, but rather distributed across multiple regions of the vest surface.

image041.png

Figure 8. Distribution of sensor activation across different body zones

These results indicate that the implemented sensor array configuration is capable of uniformly recording impact locations across various areas of the vest surface. Therefore, the system is not only able to detect the occurrence of impacts, but also capable of providing more specific information regarding the impact location.

Overall, the experimental results demonstrate that the sensor array utilized in the SmartVest operates consistently in detecting the spatial distribution of impact locations across the entire observation area. This capability supports the implementation of the system as a wearable device for real-time impact monitoring.

4.6 Signal stability across shooting distances

The effect of shooting distance variation on the system response was analyzed to evaluate the stability of signal acquisition under different testing conditions. The analysis was conducted using the average peak voltage values obtained at each shooting distance. The relationship between shooting distance and average peak voltage is illustrated in Figure 7.

Based on the obtained results, the average peak voltage values at each shooting distance remain within a relatively stable range. Although slight voltage variations are observed among the testing distances, these changes do not exhibit a significant decreasing trend as the shooting distance increases. This indicates that the system is capable of maintaining impact detection sensitivity under various testing conditions.

The highest average voltage value was obtained at a shooting distance of 22 m, while slight decreases were observed at distances of 18 m and 25 m. The observed variations may also be affected by differences in impact location and impact conditions among individual trials.

To further strengthen the signal stability analysis, the standard deviation (σ) was computed for each shooting distance using Eq. (15) to quantify the level of data variation relative to the average value. The statistical summary across all four shooting distances is presented in Table 8.

As shown in Table 8, the standard deviation values across the four shooting distances range from 0.39 V to 0.58 V, indicating a relatively consistent level of variation among repeated impact measurements. The lowest standard deviation was observed at 22 m (σ = 0.39 V), while the highest was observed at 18 m (σ = 0.58 V). These values indicate that the measured peak voltages exhibited some variation across repeated trials, but the variation remained within a relatively limited range under the tested conditions. Importantly, no measured trial at any shooting distance produced a peak voltage below the detection threshold of 1.21 V, indicating that all tested impact events generated sufficient signal amplitude for detection.

The average peak voltages ranged from 2.53 V to 3.05 V, corresponding to a maximum difference of 0.52 V across the four tested distances. The mean values did not exhibit a consistent decreasing trend as the shooting distance increased. The observed variations may also be affected by differences in impact location and impact conditions among individual trials. Considering the relatively small difference between the average values and the observed standard deviations, the results indicate that the system maintained relatively stable signal responses across the tested shooting distances. However, these results should be interpreted within the tested distance range and experimental conditions, and they do not establish the absence of a distance-dependent effect under other conditions.

$\sigma=\sqrt{\frac{1}{N} \sum_{i=1}^N\left(V_i-\bar{V}\right)^2}$   (15)

Table 10. Threshold characterization under different operating conditions

Note: ADC = analog-to-digital converter.

Overall, the experimental results demonstrate that the developed system is capable of maintaining stable signal responses across all shooting distance variations. This indicates that the implemented sensor array configuration and signal acquisition circuit are sufficiently reliable to support impact detection in the wearable-based SmartVest system.

To further evaluate the robustness of the detection threshold, additional tests were conducted under several operating conditions, including a stationary condition, slow walking while wearing the SmartVest, light impact, and test shooting. The objective of this evaluation was to determine whether low-intensity wearable motion could trigger impact detection. The experimental results are summarized in Table 10.

The results show that the stationary condition generated ADC values between 0 and 400 (0.00–0.32 V), while slow walking while wearing the SmartVest produced values ranging from 400 to 1200 ADC (0.32–0.97 V). Both conditions remained below the predefined detection threshold of 1500 ADC (1.21 V) and therefore did not trigger impact detection. In contrast, light impact and test shooting generated ADC responses satisfying ADC ≥ 1500, resulting in impact detection.

These results indicate that the selected threshold can distinguish the tested low-intensity wearable motion from impact events. In particular, the slow-walking test demonstrates that normal low-intensity movement of the SmartVest did not produce an ADC response sufficient to trigger impact detection under the tested conditions.

4.7 System performance summary

Based on all experimental results and analyses conducted, the SmartVest system demonstrates consistent and reliable performance in detecting impact events in real time. A summary of the system performance is presented in Table 11 to provide a quantitative overview of the capabilities of the proposed system.

Table 11. Summary performance

Note: ADC = analog-to-digital converter.

Based on Table 11, all impact events were successfully detected throughout the testing process, and no missed impact events were observed under the tested experimental conditions. This result indicates that the implemented acquisition architecture was able to capture the evaluated impact events using the current 37-sensor configuration. However, this observation does not establish that signal loss would not occur under longer scanning sequences, rapid consecutive impacts, or more complex operating conditions.

In addition, the system was capable of detecting up to five sensors simultaneously within a single impact event, demonstrating its ability to capture the spatial propagation of impact energy. The maximum recorded activation ($\Delta t$) of 50 ms provides an experimental observation of the temporal distribution of multi-sensor responses. However, this value does not represent the complete hardware scanning time. The observed activation intervals are therefore interpreted together with the programmed acquisition parameters and the RC time constant to characterize the temporal behavior of the implemented system.

From the signal perspective, the measured voltage range was between 1.26 V and 3.30 V, with an average value of 2.82 V. These values indicate that the generated signals are sufficiently strong for processing while still remaining within the safe operating limits of the system. The occurrence of ADC saturation during several experiments further demonstrates that the voltage protection circuit effectively limits the signal to the maximum allowable level.

Furthermore, the absence of missed impact events under the tested conditions indicates that the integrated sensing, signal conditioning, and data acquisition system successfully detected the evaluated impact events. Overall, these results demonstrate that the proposed system is capable of detecting impact events with a 100% detection success rate under the tested conditions and exhibits stable, responsive, and safe performance for real-time wearable-based impact monitoring applications.

4.8 Discussion

Although the experimental results demonstrate reliable performance under the tested conditions, it is important to acknowledge the expected system behavior under several practical operating scenarios that were not directly evaluated in this study.

First, regarding rapid consecutive impacts occurring within a short time interval, the system's ability to capture such events depends on the relationship between the RC time constant (τ), the conditioned signal duration, and the sequential acquisition process. If two impacts occur within a sufficiently short interval, their detectability depends on whether the corresponding sensor signals remain above the detection threshold when their channels are accessed. Multiple impacts may therefore be captured when the conditioned signals remain detectable during sequential acquisition. However, an impact occurring after the conditioned signal has decayed and before the corresponding channel is accessed again may not be detected. This limitation becomes more critical as the number of sensors increases and the acquisition burden increases.

The relationship between the RC time constant and sequential acquisition can also be interpreted from the implemented circuit parameters. With an equivalent resistance of approximately 500 kΩ and a 100 nF capacitor, the calculated RC time constant is approximately 50 ms. This value indicates that the conditioned signal can persist for a measurable period after the initial impact, increasing the opportunity for the sequential acquisition process to detect the signal when the corresponding sensor channel is accessed. The experimentally observed multi-sensor activation intervals in Table 7 provide an experimental observation of the temporal distribution of detected responses. These observations can be discussed in relation to the calculated RC time constant, although the activation intervals should not be interpreted as direct measurements of the RC decay or the complete scanning cycle.

Second, regarding external vibrations and environmental disturbances, the current system relies on a fixed detection threshold of 1500 ADC units (approximately 1.21 V) to distinguish valid impact events from background noise. Under controlled testing conditions, the baseline noise fluctuated within 200–800 ADC units (0.16–0.64 V), providing a sufficient margin above the noise floor. However, in environments with persistent mechanical vibrations, electromagnetic interference, or body movement artifacts, the baseline noise level may increase, potentially causing false positive detections or requiring adaptive threshold adjustment. The current design does not implement dynamic threshold adaptation, which represents a limitation for deployment in highly dynamic operational environments.

Third, regarding system scalability, the proposed architecture currently employs three CD74HC4067 multiplexers to manage 37 sensing channels through three ADC input pins. Expanding the sensor array beyond the current configuration would require additional multiplexer units and potentially longer scanning cycles, which may violate the condition $T_{signal}>T_{\{scan \}}$ for very short-duration transients. Therefore, future system designs should incorporate scanning speed optimization, parallel acquisition channels, or event-driven triggering mechanisms. In addition, future developments should include direct measurement of channel reading time $T_{\{read\}}$, total scanning cycle $T_{\{scan\}}$, and end-to-end detection latency using dedicated oscilloscope instrumentation. Overall, the identified limitations—including the absence of rapid consecutive impact testing, the use of a fixed detection threshold, and the lack of direct timing measurements—define the scope of the current evaluation and indicate clear directions for future work.

The system also demonstrates multi-sensor activation capability within a single impact event. This phenomenon reflects the propagation of mechanical energy throughout the sensor array, enabling the system not only to detect the primary impact location, but also to identify the surrounding affected areas. This capability confirms that the system can perform spatial sensing, providing more comprehensive information compared to conventional single-point detection systems. Despite its various advantages, several limitations were also identified. The occurrence of ADC saturation during high-energy impacts indicates that the current resolution still limits the system’s capability to distinguish higher impact energy levels. In addition, the use of a sequential scanning method introduces a dependency between the number of sensors and the total acquisition time. As the number of sensors increases, optimization of the scanning speed or the implementation of parallel acquisition methods will be required to maintain real-time performance. Therefore, future developments may focus on improving the acquisition architecture to overcome these limitations.

Overall, the results of this study demonstrate that the proposed system is capable of achieving a balance between hardware simplicity and reliable real-time performance, although opportunities for further improvement still remain.

The authors would like to express their gratitude to Telkom University for its support in providing research facilities and resources. The authors also sincerely acknowledge the STAS RG Center of Excellence at Telkom University for its technical support, assistance in data collection, and provision of research equipment. Appreciation is also extended to all parties involved in the testing process of this study.