A Multi-Terminal Intrinsically Safe Collaborative Acquisition Method for Personal Voice Exposure

By selecting synchronization beacons based on environmental signal characteristics in high-risk operating environments, performing passive synchronization and sampling rate compensation, dividing terminal groups and performing consistency-weighted fusion, the power consumption and synchronization conflicts of multi-terminal acoustic exposure acquisition systems are resolved, achieving high-precision data alignment and reliable fusion.

CN123111456APending Publication Date: 2026-11-17HUADU (XIAMEN) TECH CO LTD +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611561620.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-10-10
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

In high-risk operating environments, existing technologies for multi-terminal personal voice exposure acquisition systems face a conflict between power consumption optimization and synchronization quality, leading to sampling rate fluctuations and affecting the spatiotemporal alignment accuracy of data and the reliability of fusion results.

Method used

By acquiring the frequency domain energy concentration characteristics of the environmental background signal, a time synchronization beacon is selected for passive time synchronization and sampling rate offset compensation. The same sound field and cross sound field terminal groups are divided and weighted fusion is performed. By adjusting the acquisition parameters and synchronization quality coupling constraints in a feedforward manner, a dynamic balance between acquisition accuracy and node power consumption is achieved.

Benefits of technology

Without external time synchronization, it achieves high-precision time alignment and data fusion, improving the reliability and energy efficiency of the acquisition system, and is suitable for intrinsically safe operating environments such as mines and tunnels.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123111456A_ABST
    Figure CN123111456A_ABST
Patent Text Reader

Abstract

This invention relates to the field of acoustic exposure monitoring technology, and particularly to a multi-terminal intrinsically safe collaborative acquisition method for personal acoustic exposure. The method acquires environmental background signals and selects time synchronization beacons based on frequency domain energy concentration, performing passive time synchronization and sampling rate offset compensation to obtain an aligned sound pressure sequence. It calculates the spatial correlation of the sound field to divide the terminal group and performs anomaly detection to determine multidimensional measurement uncertainty. Based on this uncertainty, it performs consistency-weighted fusion of the same sound field data. It continuously estimates the rate of change of sound field activity to predict future activity, and adjusts the acquisition parameters feedforward before reaching a risk threshold. The adjustment process is constrained by synchronization quality coupling to maintain beacon acquisition accuracy. This invention resolves the conflict between power consumption optimization and synchronization quality in the absence of external time synchronization, achieving high-precision spatiotemporal alignment and dynamic power consumption balance.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of acoustic exposure monitoring technology, and in particular to a multi-terminal intrinsically safe collaborative acquisition method for personal acoustic exposure. Background Technology

[0002] In high-risk work environments such as mines, tunnels, and chemical plants, accurate assessment of personal acoustic exposure dose is crucial for preventing occupational noise-induced hearing loss. Due to the flammability and explosiveness risks in these environments, data acquisition terminals must meet intrinsically safe design requirements, with strict limitations on their transmission power, circuit energy storage, and power consumption. Furthermore, underground or enclosed spaces typically cannot receive external timing signals such as GPS, and laying wired synchronization networks is costly and difficult to maintain. Therefore, distributed intrinsically safe data acquisition terminals generally employ passive time synchronization mechanisms based on environmental background signals to achieve spatiotemporal alignment of data from multiple nodes.

[0003] Existing technologies face an irreconcilable technical contradiction in practical applications: On the one hand, in order to extend the battery life of intrinsically safe terminals, the system usually adopts a feedback power scheduling strategy, that is, when the ambient sound level is low or the battery power is insufficient, the sampling rate is reduced to save energy. On the other hand, the passive time synchronization mechanism is highly dependent on the ability to capture the high-frequency features of environmental beacon signals. When the sampling rate decreases, the accuracy of cross-correlation delay estimation will deteriorate significantly, leading to a decrease in synchronization quality or even loss of synchronization.

[0004] This degradation in synchronization quality caused by power consumption optimization, in turn, forces the system to increase the sampling rate to restore synchronization, thus forming a vicious cycle of "downsampling-loss of synchronization-upsampling-power consumption-downsampling again". This not only fails to achieve the expected energy-saving goal, but also causes drastic fluctuations in the spatiotemporal alignment accuracy of the data, which seriously affects the reliability of subsequent sound field fusion and individual dose calculation.

[0005] In addition, existing solutions often simply use arithmetic average or fixed weighting when fusing multi-terminal data, without fully considering the differential measurement uncertainties introduced by each terminal due to factors such as synchronization residuals, channel anomalies or positional deviations. They also fail to effectively distinguish and process the boundary of terminal data in different sound fields, resulting in the fusion results deviating from the true sound exposure level.

[0006] Therefore, there is an urgent need for a multi-terminal intrinsically safe collaborative acquisition method for personal voice exposure that can overcome the conflict between power consumption optimization and synchronization quality and achieve high-precision spatiotemporal alignment and dynamic power consumption balance without external time synchronization. Summary of the Invention

[0007] The technical problem to be solved by the present invention is to provide a multi-terminal intrinsically safe collaborative acquisition method for personal voice exposure, which can achieve a dynamic balance between acquisition accuracy and node power consumption and high-reliability fusion of multi-source data.

[0008] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows: A multi-terminal intrinsically safe collaborative acquisition method for personal voice exposure includes the following steps: A background environmental signal continuously collected by multiple acquisition terminals within a target area is acquired. Based on the frequency domain energy concentration characteristics of the background environmental signal, a time synchronization beacon is selected from multiple candidate physical excitations. Passive time synchronization and sampling rate offset compensation are performed between the multiple acquisition terminals based on the time synchronization beacon to obtain a time-aligned sound pressure time series. Based on the time-aligned sound pressure time series, the sound field spatial correlation degree between terminal pairs is calculated. According to the sound field spatial correlation degree, the multiple acquisition terminals are divided into same-sound-field terminal groups and cross-sound-field terminal groups. Based on historical baseline data, sound level difference anomaly detection is performed on the same-sound-field terminal groups to determine the multidimensional measurement uncertainty of the multiple acquisition terminals. Based on the multidimensional measurement uncertainty, the acoustic exposure data of the same sound field terminal group are weighted and fused to obtain the fused sound level; The sound field activity change rate of the target area is continuously calculated, and the sound field activity of the future time window is predicted based on the sound field activity change rate. Before the predicted sound field activity reaches the preset risk threshold, the acquisition parameters of the multiple acquisition terminals are adjusted in a feedforward manner. The feedforward adjustment process is constrained by synchronization quality coupling: when the evaluated synchronization quality index is lower than a preset critical value, the degradation of the acquisition parameters is forcibly limited in order to maintain the acquisition accuracy of the time synchronization beacon.

[0009] To solve the above-mentioned technical problems, another technical solution adopted by the present invention is as follows: A multi-terminal intrinsically safe personal sound exposure collaborative acquisition device includes a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps in the multi-terminal intrinsically safe personal sound exposure collaborative acquisition method described above.

[0010] The beneficial effects of this invention are as follows: This invention provides a multi-terminal intrinsically safe collaborative acquisition method for personal acoustic exposure. Without relying on external time reference sources such as GPS or IEEE 1588, it extracts the frequency domain energy concentration characteristics of the environmental background signal, selects a time synchronization beacon from multiple candidate physical excitations such as mechanical vibration beacons and acoustic environmental beacons, and performs passive time synchronization and sampling rate offset compensation among multiple terminals based on the selected beacon, thereby obtaining a high-precision time-aligned sound pressure time series. Furthermore, based on the spatial correlation of the sound field, the terminals are divided into in-sound-field terminal groups and cross-sound-field terminal groups, and sound level anomaly detection is performed in conjunction with historical baseline data. The multidimensional measurement uncertainty of each terminal is determined, and then a consistency-weighted fusion is performed based on this uncertainty, which significantly improves the reliability and anti-anomaly capability of the multi-terminal acoustic exposure fusion results. At the same time, by continuously estimating the rate of change of acoustic field activity and predicting the acoustic field activity in future time windows, the acquisition parameters are adjusted in a feedforward manner before the actual risk increases. Synchronization quality coupling constraints are introduced, and when the synchronization quality index is lower than the preset critical value, the degradation of acquisition parameters is forcibly limited, thereby breaking the vicious cycle of "downsampling - loss of synchronization - upsampling - power consumption - downsampling again" and achieving a dynamic balance between acquisition accuracy and node power consumption. This is particularly suitable for intrinsically safe and restricted operating environments such as mines, tunnels, and chemical plants. Attached Figure Description

[0011] Figure 1 This is a flowchart of a multi-terminal intrinsically safe personal voice exposure collaborative acquisition method according to an embodiment of the present invention. Detailed Implementation

[0012] To explain in detail the technical content, objectives, and effects of the present invention, the following description is provided in conjunction with the embodiments and accompanying drawings.

[0013] Before detailing the embodiments of this application, some related concepts will first be explained: Intrinsically safe: This refers to equipment that, under normal operation or specified fault conditions, cannot ignite a specified explosive gas mixture due to electrical sparks and thermal effects. It typically limits circuit energy, transmission power, and energy storage element parameters.

[0014] Sound exposure: The cumulative amount of sound energy that workers are exposed to in a noisy environment over a certain period of time, commonly expressed as equivalent sound level. L eq Or 8-hour sound exposure dose L EX,8h express.

[0015] Passive time synchronization: It does not rely on external time sources, but uses existing periodic vibrations or acoustic signals in the environment as synchronization beacons, and achieves time alignment between terminals by estimating the time difference of the signals arriving at different terminals.

[0016] GCC-PHAT: Generalized Cross-Correlation with PhaseTransform, is a cross-correlation method that enhances time delay estimation of peak sharpness by using frequency domain phase weighting.

[0017] SRO: Sampling Rate Offset, refers to the inconsistency in sampling clock frequency caused by the deviation of the local crystal oscillator frequency of different terminals.

[0018] Fisher z-transform: Performs an inverse hyperbolic tangent transform on the normalized correlation coefficient to make the distribution of the correlation index closer to a normal distribution, which makes it easier to set a stable threshold.

[0019] Mahalanobis distance: a statistical distance that considers the covariance structure between variables and is used to measure the degree of deviation of the current observation vector from the historical baseline distribution.

[0020] Equivalent sound level: The continuous steady-state sound level that has the same sound energy as time-varying noise over a given time period, measured in dB.

[0021] Uncertainty: A non-negative parameter characterizing the dispersion of measurement results. In this invention, it includes calibration uncertainty, microphone channel state uncertainty, spatial position uncertainty, time synchronization residual uncertainty, and environmental correction uncertainty.

[0022] Please refer to Figure 1 Embodiment 1 of the present invention is as follows: A multi-terminal intrinsically safe collaborative acquisition method for personal voice exposure is proposed, which constructs a complete collaborative infrastructure from cluster management to signal processing and spatiotemporal alignment. In this embodiment, multiple acquisition terminals work collaboratively within a target area. Due to the lack of external time synchronization sources such as GPS, the system first needs to establish an internal unified time reference, and then utilize physical stimuli in the environment to achieve high-precision passive synchronization.

[0023] In this embodiment, a virtual master terminal election and hysteresis dynamic rotation step is included before step S1 to ensure the stability of multiple acquisition terminals working together.

[0024] Specifically, during the system startup phase, each data acquisition terminal broadcasts an identity message via an intrinsically safe low-speed wireless channel. This message contains information about the remaining battery capacity. E i Storage capacity S i Intrinsically safe circuit health status indicator (0 indicates abnormal, 1 indicates normal) and self-test signal quality indicators Q i(Reflects the background noise level of the terminal's accelerometer). Each terminal independently calculates the comprehensive election cost based on a multidimensional cost function, and the terminal with the lowest cost is selected as the virtual master terminal.

[0025] For example, a multidimensional cost function can be defined as: ; in, e i , h i , q i , s i These are the normalized battery capacity, health status, signal quality, and remaining storage capacity, respectively. ; ; ; ; in, Q ref The preset accelerometer signal quality reference value, E full The full charge capacity of the battery is a preset fixed constant. S full This represents the total available storage capacity of the terminal.

[0026] The weighting coefficients satisfy , .

[0027] This weighting design reflects the special constraints of intrinsically safe devices: prioritizing the health of intrinsically safe circuits to ensure that only devices in good safety condition can serve as reference sources; secondly, considering battery capacity to maintain cluster lifespan; thirdly, signal quality to ensure synchronization performance; and finally, storage capacity.

[0028] When the difference between the minimum cost and the second minimum cost is less than the preset tie threshold, the health, battery capacity and signal quality are compared in turn for arbitration. If they are still the same, the one with the smaller identifier is selected, thereby avoiding election conflicts.

[0029] During operation, the virtual master terminal continuously monitors its own status. The master terminal rotation is triggered only when the remaining battery capacity is lower than the rotation trigger threshold or the health level is lower than the health trigger lower limit, and this condition is met for multiple consecutive cycles. Similarly, to restore master terminal status, the battery and health level must be higher than the corresponding recovery threshold and continuously confirmed.

[0030] This hysteresis mechanism effectively prevents frequent role switching caused by instantaneous fluctuations in sensor readings. More importantly, a two-stage confirmation mechanism is employed during rotation: the incoming suboptimal terminal first initiates the broadcast of the synchronization reference signal, and the current master terminal only stops broadcasting and switches to slave mode after receiving the confirmation frame. This mechanism ensures that the synchronization reference signal remains present throughout the entire rotation process, avoiding the risk of network-wide synchronization failure due to reference interruption and providing a continuous and stable time base for subsequent collaborative data acquisition.

[0031] After completing the cluster topology initialization, the core collaborative data acquisition process begins.

[0032] S1. Acquire environmental background signals continuously collected by multiple acquisition terminals within the target area. Based on the frequency domain energy concentration characteristics of the environmental background signals, select the time synchronization beacon from multiple candidate physical excitations.

[0033] Specifically, this embodiment does not rely on manually deployed dedicated time synchronization facilities, but instead utilizes the mechanical vibrations and acoustic noise commonly found in industrial environments as natural beacons. To quantify the "frequency domain energy concentration," this embodiment calculates the self-power spectral density of both the mechanical vibration signal and the acoustic environment signal, and calculates the peak sharpness of both within a preset analysis frequency band based on the self-power spectral density. Peak sharpness reflects the degree of concentration of signal energy at a specific frequency; higher sharpness indicates stronger recognizability of the signal as a time marker and better resistance to background noise interference.

[0034] The system selects the signal type with the highest spectral peak sharpness as the current time synchronization beacon. For example, in rock drilling scenarios, the low-frequency periodic impact component of mechanical vibration signals often has higher spectral peak sharpness than radiated noise, and is therefore preferentially selected as the beacon; while in some wind turbine steady-state operation scenarios, the specific blade passage frequency of acoustic signals may be more significant.

[0035] Furthermore, when the spectral peak sharpness of the current time synchronization beacon falls below a preset sharpness threshold, it indicates a decline in the quality of the current beacon source (e.g., equipment shutdown or change in operating conditions), and the system automatically switches to an alternative analysis frequency band or alternative signal type. This adaptive selection and switching mechanism enables the system to dynamically adapt to complex and ever-changing industrial sound field environments, always locking onto the optimal physical excitation as the synchronization reference, fundamentally solving the problem of a single beacon source easily failing in non-stationary environments.

[0036] Passive time synchronization and sampling rate offset compensation are performed between multiple acquisition terminals based on a time synchronization beacon to obtain a time-aligned sound pressure time series. This step aims to eliminate clock skew and crystal oscillator frequency differences between terminals.

[0037] In practical implementation, the timestamp of the time synchronization beacon's appearance is first recorded. The cross-correlation function between the terminal pairs is calculated using Generalized Cross-Correlation-Phase Transform (GCC-PHAT). The peak position is taken as the coarse time delay, and cubic spline interpolation is used near the peak to refine the residual time offset. The GCC-PHAT algorithm suppresses the effects of reverberation and multipath effects through phase weighting, while cubic spline interpolation improves the time delay estimation accuracy to the sub-sampling interval level, overcoming the quantization error caused by discrete sampling. Addressing the inherent frequency deviation problem of the low-cost crystal oscillator in intrinsically safe terminals, this embodiment employs a dual cross-correlation processor architecture for joint estimation: estimating the sampling rate offset (SRO) within a long observation window and estimating the residual time offset within a short observation window. The long window utilizes the frequency difference accumulated over a long period for accurate velocity measurement, while the short window quickly tracks instantaneous phase jitter.

[0038] Let the first k The estimated time delay for a long observation window is The corresponding time is Construct a time-delay observation model: ; in, This is the initial time offset. The sampling rate offset is... This is the error term;

[0039] based on K An observation matrix is ​​constructed from the time delay measurements of a long observation time window. A and delay vector b The initial time offset and sampling rate offset are jointly calculated using the least squares method. The matrix construction and solution formulas involved here are as follows: ; in: ; ; Compensated clock model: ; Among them, the observation matrix A The time delay vector consists of a column vector of all 1s and an observation time vector. b It consists of time delay measurements within each long window. Using this model, the system can simultaneously calculate the fixed time difference and relative frequency difference between two terminals.

[0040] Finally, based on the calculated initial time offset and sampling rate offset, the local sampling clock is continuously phase-calibrated. This joint calculation and continuous calibration mechanism not only eliminates the alignment error at the initial moment, but more importantly, compensates for the crystal oscillator drift that accumulates over time. This ensures that the sound pressure data collected by all terminals maintain microsecond-level time alignment during long-term monitoring tasks, laying a solid spatiotemporal reference for subsequent sound field correlation analysis and data fusion.

[0041] Through the above steps, this embodiment establishes a complete collaborative acquisition infrastructure that includes virtual master terminal management, adaptive selection of environmental beacons, and high-precision time-frequency joint compensation. This architecture achieves reliable alignment of multi-terminal data without external time synchronization, and the resulting time-aligned sound pressure time series will serve as input data for sound field correlation verification, uncertainty fusion, and adaptive scheduling in subsequent embodiments.

[0042] In complex industrial acoustic environments, simply averaging all terminal data often introduces significant errors, because different terminals may be in different acoustic zones, or individual terminals may have measurement anomalies.

[0043] S2. Based on the time-aligned sound pressure time series, calculate the sound field spatial correlation between terminal pairs, and divide multiple acquisition terminals into same-sound-field terminal groups and cross-sound-field terminal groups according to the sound field spatial correlation.

[0044] Specifically, the system does not directly rely on the original cross-correlation peak value as the criterion. Instead, it constructs a mean-free cross-covariance matrix between terminal pairs within the observation time window and calculates the normalized sound field correlation degree. To overcome the characteristics of industrial noise signals, which typically exhibit a non-Gaussian distribution and whose statistical properties fluctuate over time, this embodiment performs a Fisher z-transform on the normalized sound field correlation degree to obtain a transformed correlation degree index. The Fisher z-transform can transform the skewed distribution of the correlation coefficient into an approximately normal distribution, ensuring that the correlation degree threshold maintains statistical stability under different signal-to-noise ratio environments and avoiding misjudgments in sound field classification caused by fluctuations in ambient background noise. When the transformed correlation degree index is higher than the preset correlation threshold, the corresponding terminal pair is determined to be in the same sound field and is classified into the same sound field terminal group; otherwise, they are considered to be in different acoustic regions.

[0045] After the sound field is divided, a rigorous data quality screening is required for the same sound field terminal group. This embodiment uses historical baseline data to detect sound level anomalies in the same sound field terminal group.

[0046] In practice, the system calculates the sound level difference vector of the same sound field terminal pair within multiple consecutive observation windows, and estimates the baseline mean vector and baseline covariance matrix based on historical normal data, thereby calculating the Mahalanobis distance for the current window. Mahalanobis distance, instead of the traditional Euclidean distance, is used as the anomaly criterion because there is often an inherent correlation between the sound level differences of multiple channels caused by the sound field propagation characteristics. Euclidean distance assumes that each dimension is independent and identically distributed, which can easily misjudge normal sound field fluctuations as anomalies or miss real faults masked by correlations. Mahalanobis distance, by introducing the inverse of the covariance matrix, effectively eliminates the interference of correlations between variables and can accurately identify outliers that deviate from the normal sound field evolution pattern. When the calculated Mahalanobis distance exceeds a preset chi-square threshold, the corresponding terminal pair is marked as having a measurement channel anomaly, thus providing a basis for subsequent differential weighting.

[0047] The multidimensional measurement uncertainties of multiple acquisition terminals were further determined, and the acoustic exposure data of the same sound field terminal group were weighted and fused based on the multidimensional measurement uncertainties to obtain the fused sound level. Quantifying the multidimensional measurement uncertainties is the core to ensuring the reliability of the fusion result.

[0048] Specifically, the system calculates the microphone calibration uncertainty separately. u cal Microphone channel state uncertainty u mic Spatial location uncertainty u pos Time synchronization residual uncertainty u sync and environmental correction uncertainty u env The total combined uncertainty is obtained by combining these five types of components. u i : ; in: ; Or, to simplify the process: ; ; in, The coefficient representing the mapping from Mahalanobis distance to microphone uncertainty is... D ref The reference Mahalanobis distance is indicated, and m represents the synchronization quality mapping index.

[0049] The values ​​of each component are not fixed but dynamically correlated with the system's operating state: the microphone channel state uncertainty is calculated based on the aforementioned Mahalanobis distance; a larger Mahalanobis distance indicates that the channel deviates further from the normal baseline, and its uncertainty is assigned a higher value. The time synchronization residual uncertainty is calculated based on the evaluated synchronization quality index; the worse the synchronization quality, the lower the confidence level of the delay estimate, and the greater the corresponding uncertainty. This dynamic mapping mechanism ensures that the fusion weights can reflect the actual measurement quality of each terminal in real time.

[0050] After obtaining the total synthesis uncertainty of each terminal, the system selects different fusion strategies based on the correlation of errors between terminals. When the measurement errors of each terminal in the same sound field terminal group are independent of each other, the inverse variance weighting is used to calculate the fused sound level, that is, the terminal with the smaller uncertainty is given a higher weight.

[0051] However, in actual deployments, adjacent terminals are often subject to the same background noise interference or share the same synchronization beacon, leading to correlations between measurement errors. In this case, assuming independent errors would result in overly optimistic fusion results. Therefore, when measurement errors between terminals are correlated, this embodiment constructs a measurement error covariance matrix. The diagonal elements of this matrix represent the square of the total combined uncertainty for each terminal, while the off-diagonal elements are calculated based on the error correlation coefficients between terminals and the corresponding total combined uncertainty. Based on this measurement error covariance matrix, the generalized least squares method is used to calculate the fused sound level and fusion uncertainty. Generalized least squares fusion fully considers the relevant information in the error structure and can provide the statistically optimal unbiased estimate under error correlation conditions.

[0052] For a group of terminals within the same sound field If the errors are independent, inverse variance weighting is used; if the errors are correlated, the covariance matrix is ​​estimated. It employs generalized least squares fusion to output a fused sound level. L eq,fused and fusion uncertainty u fused .

[0053] When the measurement errors of each terminal are independent, inverse variance weighting is used: ; ; ; When there is correlation between errors between terminals, generalized least squares fusion is used. Let the measurement error covariance matrix be... diagonal element is The off-diagonal elements are: ; but: ; ; in, w i For the fusion weight of terminal i, For a group of terminals within the same sound field, represented in the form of a dataset, This represents the correlation coefficient of errors between terminals.

[0054] For terminals classified as cross-sound field terminal groups, their direct participation in fusion would disrupt the spatial consistency of the sound field because they are located in a region with significantly different acoustic characteristics from the main sound field.

[0055] Therefore, this embodiment also includes cross-sound field terminal processing and individual exposure dose time-weighted correction steps. Specifically, for terminals in a cross-sound field terminal group, each terminal reports its sound exposure data independently, without forced fusion across regions.

[0056] Simultaneously, the system records the spatial positions of the sound field boundary terminal pairs. For the sound field boundary region, inverse distance-weighted interpolation is used to generate an estimate of the transition region, and the upper bound of the uncertainty is marked on the interpolation result to characterize the speculative nature of the data in this region. In the final assessment of individual sound exposure of workers, the individual sound exposure dose is time-weighted and synthesized based on the worker's residence time in each sound field. The calculation formula is as follows: ;

[0057] in, L EX,8h For an individual's 8-hour sound exposure dose, T 0 represents the standard working time. T s For the operators in the sound field s The length of stay in the country L eq,s For sound field s The equivalent sound level after fusion. Furthermore, to ensure the integrity of the dose assessment results, the dose uncertainty of an individual's 8-hour sound exposure dose needs to be calculated using the uncertainty propagation law, based on the uncertainty of each sound field fusion result. This approach respects the physical isolation characteristics of the sound fields while accurately restoring the cumulative noise exposure level of personnel during mobile operations through time weighting, achieving a precise closed loop from group monitoring data to individual health risk assessment.

[0058] S4. Continuously estimate the rate of change of sound field activity in the target area, predict the sound field activity in the future time window based on the rate of change of sound field activity, and adjust the acquisition parameters of multiple acquisition terminals in a feedforward manner before the predicted sound field activity reaches the preset risk threshold.

[0059] Specifically, traditional feedback power scheduling often passively reduces the sampling rate only after detecting that the ambient sound level has decreased or the battery power is insufficient. This delayed response can easily lead to the loss of key acoustic events due to the untimely recovery of the sampling rate when the ambient noise suddenly increases, or waste of power due to the late downsampling after the environment has stabilized.

[0060] To overcome this deficiency, this embodiment employs a feedforward control strategy, estimating the time derivative of sound field activity, i.e., the rate of change of sound field activity, online using exponentially weighted moving average (EWMA) or sliding window linear regression. For example, sliding window linear regression is used to fit the sound field activity over the most recent M moments, and its slope characterizes the evolution trend of environmental noise energy. Subsequently, based on the current rate of change of sound field activity and a preset feedforward advance time, the predicted sound field activity for the future time window is calculated. This feedforward advance time can be dynamically adjusted according to the amplitude of the rate of change; the larger the rate of change, the longer the advance time, ensuring that the system has sufficient time to complete parameter switching. When the prediction results show that the future sound field activity will be lower than the low-risk threshold and the rate of change meets the descent condition, the system proactively downgrades the sampling rate in the collected parameters to an energy-saving sampling rate. This proactive adjustment method allows the terminal to maximize energy savings during environmental quiet periods while ensuring that no high-risk noise events are missed, completely eliminating the risk of data loss and energy efficiency loss caused by response lag in feedback scheduling.

[0061] While adjusting the acquisition parameters, the system also needs to execute a tiered computing migration strategy based on the remaining battery capacity of multiple acquisition terminals. This tiered computing migration strategy includes a normal level, a migration level, and a deep energy-saving level. This strategy aims to dynamically allocate the load ratio between edge computing and terminal computing based on the energy status.

[0062] Specifically, when the remaining battery capacity is higher than the normal level threshold (such as 50% full charge capacity), the terminal is in the normal level, performs sound exposure integration calculation locally, and only reports the equivalent sound level result to the edge node to minimize wireless communication power consumption. When the remaining battery capacity is between the migration level threshold (e.g., 30%) and the normal level threshold, the terminal enters the migration level. At this time, the terminal stops the local high-energy-consuming integration calculation and instead compresses and uploads the original sound pressure sequence, which is then completed by the edge node to perform the integration calculation, thereby releasing the terminal's computing power and extending its battery life. When the remaining battery capacity is lower than the threshold of the deep energy-saving level, the terminal enters the deep energy-saving level, executes the low duty cycle sampling mode, and the edge node uses the spatial interpolation method to estimate the acoustic exposure value of the terminal, while marking the interpolation estimation uncertainty of the interpolation result.

[0063] Through this three-level progressive load migration mechanism, the system can achieve balanced energy consumption at the cluster level, avoid individual terminals running out of power too early and creating monitoring blind spots, and ensure that basic spatial coverage can still be maintained even in extremely low power conditions.

[0064] However, unrestricted power consumption optimization can easily damage the foundation of passive time synchronization. Therefore, in this embodiment, the feedforward adjustment process is constrained by synchronization quality coupling: when the evaluated synchronization quality index is lower than a preset threshold, the degradation of the acquisition parameters is forcibly limited to maintain the acquisition accuracy of the time synchronization beacon. This is the core mechanism of this application to break the vicious cycle of "downsampling-out-upsampling" in the traditional system.

[0065] Specifically, the system continuously calculates the peak sharpness index or peak-to-order ratio of the cross-correlation function of the time synchronization beacon as a synchronization quality index. Peak sharpness reflects the identifiability of the beacon signal in the time domain, while peak-to-order ratio characterizes the significance of the main peak relative to the side lobes; both are direct physical quantities for measuring the reliability of passive synchronization. When the synchronization quality index falls below a preset threshold, it indicates that the current environmental beacon is weak or heavily interfered with, and the synchronization link is in a vulnerable state. At this time, a synchronization quality coupling constraint is triggered, forcibly limiting the sampling rate to no less than the minimum sampling rate required to maintain synchronization. The underlying mechanism is that passive time synchronization relies on the accurate capture of the high-frequency characteristics of the environmental beacon. If the sampling rate is blindly reduced to meet energy-saving requirements, it will lead to spectral aliasing or feature blurring of the beacon signal, which will drastically deteriorate the accuracy of time delay estimation or even cause complete loss of synchronization. Once synchronization is lost, all subsequent sound field correlation and data fusion will fail, forcing the system to increase the sampling rate again to search for the beacon, resulting in greater energy waste and data gaps.

[0066] Therefore, this coupling constraint establishes a dynamic trade-off logic of "accuracy first, energy saving second": even if the feedforward prediction display environment is quiet and the battery power is low, as long as the synchronization quality is insufficient, the system will still reject aggressive degradation instructions and lock the sampling rate above the safety threshold. This design, which uses synchronization quality as a hard constraint for power consumption scheduling, fundamentally decouples the zero-sum game between energy saving and synchronization, ensuring that the multi-terminal collaborative acquisition system always has a stable spatiotemporal reference during long-term operation, and achieving true adaptive dynamic balance.

[0067] Embodiment 2 of the present invention is as follows: A storage medium storing a computer program thereon, characterized in that, when the computer program is executed, it implements the steps in the multi-terminal intrinsically safe personal voice exposure collaborative acquisition method described in Embodiment 1 above.

[0068] In summary, the beneficial effects of the present invention include: This invention solves the problem of missing reference for spatiotemporal alignment of multiple terminals under conditions without external time synchronization by selecting the best time synchronization beacon based on the frequency domain energy concentration characteristics of environmental background signals and combining passive time synchronization and sampling rate offset compensation mechanisms. The principle is to use the mechanical vibration or acoustic noise commonly found in industrial environments as natural beacons, and to screen out the most recognizable signal sources through frequency domain feature analysis, avoiding the cost and safety risks of manually deploying time synchronization facilities. At the same time, by jointly estimating the time delay and sampling rate offset, the cumulative error caused by the inherent frequency deviation of the crystal oscillator is eliminated, providing a high-precision time reference for subsequent data fusion.

[0069] This invention divides terminal groups by calculating the spatial correlation of the sound field and detects sound level anomalies based on historical baseline data to determine multidimensional measurement uncertainties. Then, it implements consistent weighted fusion, which improves the overall reliability of sound exposure data. The principle is to use the mean-removed covariance and Fisher z-transform stability correlation criterion, combined with Mahalanobis distance to identify channel anomalies, and dynamically map multidimensional uncertainty components such as calibration, channel, location, synchronization and environment to the fusion weights. This makes the fusion results automatically biased towards high-confidence data sources. At the same time, it independently processes and marks the boundaries of terminals across sound fields, avoiding the erroneous mixing of data from different acoustic regions and significantly improving the accuracy of individual dose calculation.

[0070] This invention solves the conflict between power consumption optimization and synchronization quality by continuously estimating the rate of change of sound field activity and adjusting the acquisition parameters in a feedforward manner, while applying synchronization quality coupling constraints during the adjustment process. Its principle lies in abandoning the traditional post-triggered feedback scheduling and instead predicting risks based on the trend of environmental state changes and adjusting the sampling rate in advance, avoiding data loss caused by response lag. At the same time, when the synchronization quality index is lower than the critical value, the sampling rate is forcibly limited to not be lower than the minimum threshold required to maintain beacon acquisition, preventing the deterioration of delay estimation accuracy caused by low sampling rate. Thus, it maximizes energy saving effect while ensuring the stability of the synchronization link, and achieves a dynamic optimal balance between acquisition accuracy and node endurance.

[0071] The above description is merely an embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent modifications made based on the content of the present invention specification and drawings, or direct or indirect applications in related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A multi-terminal intrinsically safe collaborative acquisition method for personal voice exposure, characterized in that, Including the following steps: S1. Acquire the environmental background signal continuously collected by multiple acquisition terminals in the target area. Based on the frequency domain energy concentration characteristics of the environmental background signal, select the time synchronization beacon from multiple candidate physical excitations. Perform passive time synchronization and sampling rate offset compensation between the multiple acquisition terminals based on the time synchronization beacon to obtain a time-aligned sound pressure time series. S2. Based on the time-aligned sound pressure time series, calculate the sound field spatial correlation degree between terminal pairs, divide the multiple acquisition terminals into same-sound-field terminal groups and cross-sound-field terminal groups according to the sound field spatial correlation degree, and perform sound level difference anomaly detection on the same-sound-field terminal groups based on historical baseline data to determine the multidimensional measurement uncertainty of the multiple acquisition terminals. S3. Based on the multidimensional measurement uncertainty, perform consistency weighted fusion on the acoustic exposure data of the same sound field terminal group to obtain the fused sound level; S4. Continuously calculate the rate of change of sound field activity in the target area, predict the sound field activity in the future time window based on the rate of change of sound field activity, and adjust the acquisition parameters of the multiple acquisition terminals in a feedforward manner before the predicted sound field activity reaches the preset risk threshold. The feedforward adjustment process is constrained by synchronization quality coupling: when the evaluated synchronization quality index is lower than a preset critical value, the degradation of the acquisition parameters is forcibly limited in order to maintain the acquisition accuracy of the time synchronization beacon.

2. The multi-terminal intrinsically safe personal voice exposure collaborative acquisition method according to claim 1, characterized in that, The selection of time synchronization beacons from multiple candidate physical excitations based on the frequency domain energy concentration characteristics of the environmental background signal includes: Calculate the self-power spectral density of the mechanical vibration signal and the acoustic environment signal respectively; Based on the self-power spectral density, calculate the spectral peak sharpness of the mechanical vibration signal and the acoustic environment signal within the preset analysis frequency band; The signal type with the highest spectral peak sharpness is selected as the current time synchronization beacon. When the spectral peak sharpness of the current time synchronization beacon is lower than a preset sharpness threshold, it automatically switches to the alternative analysis frequency band or alternative signal type.

3. The multi-terminal intrinsically safe personal voice exposure collaborative acquisition method according to claim 2, characterized in that, The passive time synchronization and sampling rate offset compensation between the multiple acquisition terminals based on the time synchronization beacon includes: Record the timestamp of the occurrence of the time synchronization beacon, calculate the cross-correlation function between the terminal pairs through generalized cross-correlation-phase transformation, take the peak position as the coarse time delay, and refine the residual time offset by cubic spline interpolation near the peak. A dual cross-correlation processor architecture is adopted to estimate the sampling rate offset within a long observation time window and the residual time offset within a short observation time window; Let the first k The estimated time delay for a long observation window is The corresponding time is Construct a time-delay observation model: ; in, This is the initial time offset. The sampling rate offset is... This is the error term; based on K The delay measurements of a long observation time window are used to construct an observation matrix and a delay vector. The initial time offset and the sampling rate offset are then jointly calculated using the least squares method. Based on the calculated initial time offset and sampling rate offset, the local sampling clock is continuously phase-calibrated.

4. The multi-terminal intrinsically safe personal voice exposure collaborative acquisition method according to claim 1, characterized in that, The method involves calculating the spatial correlation between terminal pairs based on the time-aligned sound pressure time series, dividing the multiple acquisition terminals into in-sound-field terminal groups and cross-sound-field terminal groups according to the spatial correlation, and detecting sound level anomalies in the in-sound-field terminal groups based on historical baseline data, including the following steps: Within the observation time window, a mean-free cross-covariance matrix is ​​constructed between terminal pairs, and the normalized acoustic field correlation degree is calculated. Perform a Fisher z-transform on the normalized sound field correlation to obtain the transformed correlation index; When the correlation index is higher than the preset correlation threshold, it is determined that the corresponding terminal pair is in the same sound field and is assigned to the same sound field terminal group. For the terminal pairs in the same sound field terminal group, the sound level difference vector is calculated within multiple consecutive observation windows, the baseline mean vector and baseline covariance matrix are estimated based on historical normal data, and the Mahalanobis distance of the current window is calculated. When the Mahalanobis distance exceeds a preset chi-square threshold, the corresponding terminal pair is marked as having a measurement channel abnormality.

5. The multi-terminal intrinsically safe personal voice exposure collaborative acquisition method according to claim 4, characterized in that, Determining the multidimensional measurement uncertainty of the multiple acquisition terminals includes the following steps: The microphone calibration uncertainty, microphone channel state uncertainty, spatial position uncertainty, time synchronization residual uncertainty, and environmental correction uncertainty of the multiple acquisition terminals are calculated and combined to obtain the total combined uncertainty. The microphone channel state uncertainty is calculated based on the Mahalanobis distance, and the time synchronization residual uncertainty is calculated based on the synchronization quality index, which is the peak sharpness index or peak order ratio of the cross-correlation function of the time synchronization beacon. Based on the aforementioned multidimensional measurement uncertainty, the acoustic exposure data of the same sound field terminal group are subjected to consistency-weighted fusion, including the following steps: When the measurement errors of each terminal in the same sound field terminal group are independent of each other, the inverse variance weighted calculation is used to calculate the fused sound level; When there is a correlation between measurement errors between terminals, a measurement error covariance matrix is ​​constructed. The diagonal elements of the measurement error covariance matrix are the squares of the total combined uncertainty of each terminal, and the off-diagonal elements are calculated based on the error correlation coefficient between terminals and the corresponding total combined uncertainty. Based on the measurement error covariance matrix, the fused sound level and fusion uncertainty are calculated using the generalized least squares method.

6. The multi-terminal intrinsically safe personal voice exposure collaborative acquisition method according to claim 4, characterized in that, Step S3 is followed by cross-field terminal processing and time-weighted correction of individual exposure dose: For each terminal in the cross-sound field terminal group, each terminal independently reports sound exposure data and records the spatial location of the sound field boundary terminal pairs; For the sound field boundary region, inverse distance weighted interpolation is used to generate an estimate of the transition region, and the upper bound of the uncertainty is marked on the interpolation result; Based on the dwell time of workers in each sound field, the individual sound exposure dose is calculated using a time-weighted composite method, and the formula is as follows: ; in, L EX,8h For an individual's 8-hour sound exposure dose, T 0 represents the standard working time. T s For the operators in the sound field s The length of stay in the country L eq,s For sound field s The equivalent sound level after fusion; Based on the uncertainty of each sound field fusion result, the dose uncertainty of the individual's 8-hour sound exposure dose is calculated through uncertainty propagation.

7. The multi-terminal intrinsically safe personal voice exposure collaborative acquisition method according to claim 1, characterized in that, Step S4 includes the following steps: The rate of change of the sound field activity is estimated online based on exponentially weighted moving average or sliding window linear regression methods. The predicted sound field activity is calculated based on the rate of change of sound field activity and the amount of feedforward advance time. When the rate of change of the sound field activity meets the preset decrease condition and the predicted sound field activity is lower than the preset low-risk threshold, the sampling rate in the acquisition parameters is downgraded to the energy-saving sampling rate. Based on the remaining battery capacity of the multiple acquisition terminals, a hierarchical calculation migration strategy is executed, which includes a normal level, a migration level, and a deep energy-saving level.

8. The multi-terminal intrinsically safe personal voice exposure collaborative acquisition method according to claim 7, characterized in that, The feedforward adjustment process is constrained by synchronous quality coupling and includes the following steps: The peak sharpness index or peak order ratio of the cross-correlation function of the time synchronization beacon is continuously calculated as the synchronization quality index. When the synchronization quality index is lower than the preset critical value, the synchronization quality coupling constraint is triggered. Under the aforementioned synchronization quality coupling constraint, the sampling rate is forcibly limited to not be lower than the minimum sampling rate required to maintain synchronization, in order to prevent the low sampling rate from causing a deterioration in the delay estimation accuracy of passive time synchronization; In the hierarchical migration strategy, when the remaining battery capacity is lower than the preset migration level threshold, the terminal compresses and uploads the original sound pressure sequence, which is then integrated by the edge node; when the remaining battery capacity is lower than the preset deep energy-saving level threshold, the terminal performs low duty cycle sampling, and the edge node uses spatial interpolation to estimate the sound exposure value and marks the interpolation estimation uncertainty.

9. The multi-terminal intrinsically safe personal voice exposure collaborative acquisition method according to claim 1, characterized in that, Before step S1, there are also steps for virtual master terminal election and hysteresis dynamic rotation: During the startup phase, each acquisition terminal broadcasts an identity message containing information such as remaining battery capacity, storage capacity, intrinsically safe circuit health status, and self-test signal quality indicators. Based on a multidimensional cost function, the comprehensive election cost is independently calculated, and the terminal with the lowest cost is selected as the virtual master terminal. During operation, the virtual master terminal continuously monitors its own battery remaining capacity and health. When the triggering conditions are met and are met for multiple consecutive cycles, the master terminal rotation is triggered. The triggering conditions include the remaining battery capacity being lower than the rotation trigger threshold or the battery health being lower than the health trigger lower limit; When the recovery conditions are met and the conditions are met for multiple consecutive cycles, the main terminal qualification is allowed to be restored. The recovery conditions include the remaining battery capacity being higher than the rotation recovery threshold and the health level being higher than the minimum health level. During the rotation, a two-stage confirmation mechanism is adopted, which involves handing over the responsibility first and then releasing it. The next suboptimal terminal to take over first starts broadcasting the synchronization reference signal, and the current master terminal stops broadcasting and switches to slave terminal mode after receiving the confirmation frame.

10. A storage medium having a computer program stored thereon, characterized in that, When the computer program is executed, it implements the steps of the multi-terminal intrinsically safe personal voice exposure collaborative acquisition method according to any one of claims 1-9.