A cloud-edge-end cooperative directional sound intelligent adaptive management and control system

Abstract

The application discloses a kind of based on cloud-edge-end cooperation's directional sound intelligent adaptive management and control system, belong to noise control technical field.System includes noise monitoring microstation, cloud intelligent management and control platform, intelligent edge control box and directional sound array.Noise monitoring microstation real-time acquisition peripheral noise data;Cloud intelligent management and control platform runs threefold intelligent adaptive control algorithm, according to the comparison of noise data and threshold, dynamically generates control instruction;Intelligent edge control box receives and executes instruction, then based on last effective parameter independent operation local closed-loop control when network interruption;Directional sound array projects sound to control area inside with directional beam.Three control algorithms include: dynamic fast down-regulation when exceeding, exploratory optimization up-regulation when there is safety margin, fine tuning to approach optimal balance point when steady state.The application optimizes internal sound field effect intelligently under the premise of rigid guaranteeing that peripheral noise meets the standard, and has the ability of autonomous network, realizes closed-loop adaptive control.

Description

Technical Field

[0001] This invention relates to the field of noise control technology, and in particular to a directional sound intelligent adaptive control system based on cloud-edge-device collaboration. Background Technology

[0002] With urban development, many school playgrounds are adjacent to residential areas. During flag-raising ceremonies, physical education classes, sports meets, and extracurricular activities, the noise from broadcasts, music, and crowds on the playground can easily cause serious disturbance to nearby residents. This is a common and long-standing conflict between schools and communities in densely populated cities.

[0003] To address the aforementioned issues, current common solutions have significant limitations: passive noise reduction methods, such as simply lowering the volume of the playground's loudspeakers, may temporarily reduce external noise, but directly result in students at the far end of the playground not being able to hear instructions, seriously affecting the quality of teaching and activities, and creating a dilemma of "either disturbing residents or affecting school activities"; while management methods that rely on manual inspections or resident complaints lack the ability to perceive the true level of external noise in real time, cannot achieve rapid and automatic intervention when noise exceeds the standard, have a delayed response and limited methods, and cannot simultaneously ensure that external noise meets standards and the internal sound field effect.

[0004] Recently, directional sound technology has begun to be used to address noise pollution issues on school playgrounds. For example, the directional broadcasting system deployed at Wenli School in Longgang District, Shenzhen, can focus sound energy into a narrow beam on the target area of ​​the playground, with a measured sound pressure difference of over 26 decibels between the main lobe and the back. Nanjing City also adopted phased array beamforming technology and combined it with quantitative simulation design for pre-evaluation during its pilot "quiet playground" renovation project. These practices have demonstrated the effectiveness of directional sound technology in achieving a "sound spotlight" effect. However, existing solutions are mostly open-loop systems, meaning they preset sound beams with fixed volume and angle, lacking real-time perception of external noise and intelligent feedback control mechanisms based on this perception. The system cannot dynamically optimize based on environmental changes such as wind speed and temperature, or actual noise feedback from residential areas, and may fail due to equipment aging and environmental changes over long-term operation. Furthermore, existing systems lack an effective cloud-edge collaborative architecture, making it difficult to guarantee the continuity and reliability of local closed-loop control in abnormal situations such as network interruptions.

[0005] Furthermore, existing technical solutions generally do not consider sound source type identification and filtering mechanisms. When peripheral monitoring points collect non-target noise such as transportation or construction noise, it is easily misjudged as noise from the playground itself, triggering unnecessary noise reduction controls and affecting the robustness of the system and the stability of the sound experience within the school. Therefore, there is an urgent need for a closed-loop management system that can achieve intelligent sensing, precise projection, dynamic control, and possess cloud-edge collaboration and offline autonomy capabilities. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings of the prior art and provide a directional sound intelligent adaptive control system based on cloud-edge-device collaboration.

[0007] The objective of this invention is achieved through the following technical solution: a directional sound intelligent adaptive control system based on cloud-edge-device collaboration, comprising a noise monitoring micro-station, wherein the noise monitoring micro-station is connected to a cloud-based intelligent control platform, the cloud-based intelligent control platform is connected to an intelligent edge control box, and the intelligent edge control box is connected to a directional sound array; The noise monitoring micro-stations are deployed in preset sensitive areas near the control area to collect noise data in real time; The directional acoustic array is deployed at the boundary of the controlled area to project sound into the controlled area in the form of directional beams. The intelligent edge control box is deployed locally as a local control hub, and has a built-in lightweight AI inference engine and simplified rule engine. When the network connection is normal, the intelligent edge control box receives and executes control commands issued by the cloud intelligent management and control platform. When the network connection is interrupted, the intelligent edge control box independently runs local closed-loop control based on the last stored valid control parameters, and automatically synchronizes logs and policies after the network is restored. The cloud-based intelligent management and control platform operates a triple intelligent adaptive control algorithm, which dynamically adjusts the output parameters of the directional acoustic array based on the comparison results of noise data and noise threshold, thereby forming a closed-loop management and control system based on real-time perception and intelligent decision-making.

[0008] Preferably, multiple noise monitoring micro-stations are set up. Each noise monitoring micro-station integrates a noise microphone, a meteorological sensor, and a wireless communication module to collect noise equivalent sound level, maximum sound level, spectrum data, and meteorological data in real time, and uploads them through a wireless network.

[0009] Preferably, multiple directional sound arrays are configured, employing phased array beamforming technology or parametric array speaker arrays to project sound energy into the controlled area in a narrow beam, thereby reducing the lateral and rearward sound pressure levels. The cloud-based intelligent control platform is equipped with a visual interface to display real-time noise maps, equipment status monitoring, alarm logs, and statistical analysis reports, and provides remote policy configuration functions.

[0010] Preferably, the triple intelligent adaptive control algorithm includes the following steps: Dynamic fast response steps: When the noise data meets the preset exceeding conditions, the system enters the dynamic fast response mode and generates a control command to lower the output parameters of the directional acoustic array; Trial optimization step: When the noise data continuously meets the preset safety margin condition within the preset safety time, the system enters the trial optimization mode and generates a control command to adjust the output parameters of the directional acoustic array in order to find the optimal output. Steady-state fine optimization steps: When the system is not in dynamic fast response mode or trial optimization mode, and the noise data fluctuates within the preset range for more than the stable time, and there are no records of exceeding the standard trigger, the output parameters of the directional acoustic array are slightly adjusted according to the preset optimization rules to reach the long-term optimal balance point and maximize the acoustic environment benefits.

[0011] Preferably, the preset exceeding conditions and response mechanism in the dynamic rapid response step are as follows: When the real-time equivalent sound level of any noise monitoring micro-station L current satisfy L current > L th +Δ, and if the duration exceeds the set time threshold, a downward control is triggered; or, When the real-time equivalent sound level of any noise monitoring micro-station L current satisfy L current > L th When +2Δ is applied, the downward control is immediately triggered; or, When the calculated noise change trend meets the preset trend conditions, the pre-volume reduction control is triggered in advance, and a dynamic reduction step size is generated. Step down : And limit the maximum downregulation to prevent mutations; in L th Δ is the noise threshold, and Δ is the dead zone threshold. Step base Adjust the step size based on the base. k This represents the dynamic step-size gain coefficient.

[0012] Preferably, the preset safety margin condition in the trial-and-error optimization step is: the real-time noise values ​​of all noise monitoring microstations meet the real-time equivalent sound level for a continuous period exceeding a preset time. L current Noise threshold L th - Dead zone threshold Δ; or, simultaneously satisfying that the sound pressure level of the reference point within the control area is lower than the preset target value; The optimization strategy for the exploratory optimization step includes the following steps: With a preset trial step size Step upIncrease the current volume V current Get trial output V test ; During the preset observation period, monitor the real-time equivalent sound level of all noise monitoring micro-stations. L current ; If all monitoring data consistently meet the safety margin requirements, then this upgrade will be accepted. V current = V test And repeat the above steps; If, after a certain upgrade, any monitoring data enters the dead zone or fails to meet the preset safety margin condition, the output parameter value before the upgrade is returned and locked as the current optimal value. Then, it is determined whether the exit condition is met. If it is met, the optimization mode is exited and the system enters a steady state.

[0013] Preferably, the preset range in the steady-state fine optimization step is: L th ±Δ / 2; the preset optimization rule is: periodically based on the average equivalent continuous sound level over a past period of time. Leq avg and the number of times the instantaneous value exceeds the limit Lmax count Perform fine adjustments to keep the average ambient noise level close to the noise threshold. L th But it does not exceed the standard level.

[0014] Preferably, the cloud-based intelligent management and control platform and the intelligent edge control box collaborate using either a policy-based distribution mode or a target-based distribution mode. In the policy-based distribution mode, the cloud-based intelligent management and control platform distributes specific control quantities. In the target-based distribution mode, the cloud-based intelligent management and control platform distributes threshold values ​​and scene tags, and the intelligent edge control box calculates the control quantities based on local real-time data. The cloud-based intelligent management and control platform automatically adjusts various control parameters in a targeted manner based on preset time period strategies or activity scenario strategies; The cloud-based intelligent management and control platform constructs a time series prediction model based on historical noise data to predict future noise peak times and adjusts the output parameters of the directional acoustic array in advance to smooth the control curve.

[0015] Preferably, the triple intelligent adaptive control algorithm also incorporates an environmental factor compensation mechanism, including: Wind speed compensation: When the wind speed exceeds the set wind speed and the wind direction is towards a preset sensitive area near the controlled area, the noise threshold is adjusted. L th Introduce a compensation offset; Temperature and humidity compensation: When the temperature change exceeds the set temperature and humidity, the beam pointing angle of the directional acoustic array is finely adjusted.

[0016] Preferably, the cloud-based intelligent management and control platform also runs a sound source identification and filtering algorithm to extract features of the collected audio segments and identify the sound source type through a classifier; if it is identified as non-target type noise and the confidence level is higher than the confidence level threshold, the noise data of the monitoring point will not participate in the control decision.

[0017] The beneficial effects of this invention are: 1) By constructing a three-tiered collaborative closed-loop management architecture of "cloud-edge-device," the system achieves second-level proactive response and dynamic balancing for noise control. Due to the deployment of real-time noise monitoring micro-stations near the controlled area, and the high-speed transmission of data and control commands through a collaborative network between the cloud-based intelligent management platform and the intelligent edge control box, the system can immediately trigger a dynamic rapid response mechanism when noise exceeds the standard at any monitoring point. This automatically generates control commands to lower the output parameters of the directional acoustic array, causing the noise to quickly drop below the safe threshold, solving the problems of reliance on manual inspection and delayed response in traditional solutions. Simultaneously, the simplified rule engine built into the intelligent edge control box can independently run local closed-loop control based on the last stored valid control parameters when the network connection is interrupted, and automatically synchronize logs and policies after the network is restored, greatly enhancing the system's robustness and operational continuity.

[0018] 2) Through the core triple intelligent adaptive control algorithm, intelligent optimal adjustment of the sound effect within the campus is achieved while rigidly ensuring that the external noise meets the standard. Among them, the trial optimization step cautiously increases the volume with a preset trial step size under the condition that the external noise value is continuously lower than the safety margin of the difference between the noise threshold and the dead zone threshold. When the data at any monitoring point enters the dead zone, it immediately reverts and locks the current optimal value, which solves the problem of the traditional solution of "one-size-fits-all" volume reduction that sacrifices the effect of activities within the campus. The steady-state fine optimization step, when the system is running stably for a long time, makes periodic adjustments based on the equivalent continuous sound level average value and the number of times the instantaneous value exceeds the limit, not exceeding the preset micro-amplitude upper limit, so that the average external noise value continuously approaches the threshold but does not exceed the standard, thereby achieving the dual-objective optimization of "quiet outside and clear inside".

[0019] 3) Through the visual interface of the cloud-based intelligent management and control platform, intangible noise data is transformed into real-time noise maps, equipment status monitoring, alarm logs and statistical analysis reports, and remote policy configuration functions are provided, enabling managers to remotely and comprehensively grasp the acoustic environment, transforming passive complaints into proactive prevention and control, and greatly improving the digitalization and refinement of noise management.

[0020] 4) The environmental factor compensation mechanism further enhances the system's adaptability to complex scenarios. When the wind speed exceeds the set value and the wind direction points to the sensitive area, a compensation offset is introduced to the noise threshold to avoid false noise reduction caused by wind interference or sound propagation attenuation. When the temperature change exceeds the set range, the beam pointing angle of the directional sound array is finely adjusted to ensure accurate coverage of the target area within the campus under different climatic conditions.

[0021] 5) By using sound source identification and filtering algorithms, features are extracted from the collected audio segments and the sound source type is identified. When the noise is identified as non-target type such as transportation or construction and the confidence level is higher than the threshold, the data is excluded from the control decision. This effectively avoids false triggering of control due to occasional noise outside the sensitive area, and significantly improves the robustness of the system operation and the stability of the sound effect experience within the school.

[0022] 6) Through preset time period strategies and activity scenario strategies, the system can automatically adjust control parameters such as noise thresholds according to specific activity patterns such as class schedules, daily routines, or "sports meets"; at the same time, through a time series prediction model built based on historical noise data, it can predict future noise peak times and smoothly adjust output parameters in advance, enabling the system to have a forward-looking adaptability to periodic and sudden changes in the sound environment, ensuring control effectiveness and sound effect experience under long-term operation. Attached Figure Description

[0023] Figure 1 This is a system architecture diagram using a school playground as an example; Figure 2 Here is a flowchart of the triple intelligent adaptive control algorithm; Figure 3 This is a schematic diagram of a directional sound intelligent control implementation case, taking a campus as an example. Detailed Implementation

[0024] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] See Figures 1-3 The present invention provides a technical solution: a directional sound intelligent adaptive control system based on cloud-edge-device collaboration, including a noise monitoring micro-station, the noise monitoring micro-station being connected to a cloud-based intelligent control platform, the cloud-based intelligent control platform being connected to an intelligent edge control box, and the intelligent edge control box being connected to a directional sound array; The noise monitoring micro-stations are deployed in preset sensitive areas near the control area to collect noise data in real time; The directional acoustic array is deployed at the boundary of the controlled area to project sound into the controlled area in the form of directional beams. The intelligent edge control box is deployed locally as a local control hub, and has a built-in lightweight AI inference engine and simplified rule engine. When the network connection is normal, the intelligent edge control box receives and executes control commands issued by the cloud intelligent management and control platform. When the network connection is interrupted, the intelligent edge control box independently runs local closed-loop control based on the last stored valid control parameters, and automatically synchronizes logs and policies after the network is restored. The cloud-based intelligent management and control platform operates a triple intelligent adaptive control algorithm, which dynamically adjusts the output parameters of the directional acoustic array based on the comparison results of noise data and noise threshold, thereby forming a closed-loop management and control system based on real-time perception and intelligent decision-making.

[0026] In this embodiment, as Figure 1 As shown, taking a school playground as an example (not limited to school playgrounds; applicable to any situation requiring control of sound leakage from a specific area while ensuring internal sound quality, such as cultural activities in city parks and squares, commercial events or concerts in sports stadiums, broadcasting areas in train stations / airports, and targeted announcements in open-plan office spaces), a closed-loop management system of "real-time perception - intelligent decision-making - precise execution" is constructed to solve the problem of noise pollution from playground activities. This system collects noise data in real time through noise monitoring micro-stations (sensing ends) deployed around the playground; the data is uploaded to a cloud-based intelligent management platform (decision center), where a triple intelligent adaptive control algorithm is used for analysis and decision-making, generating control commands; these commands are then sent to intelligent edge control boxes (edge ​​nodes), ultimately driving the directional sound array (execution end) to adjust its sound beam parameters (such as volume). Through this closed loop, the system can flexibly optimize and improve the sound field coverage within the playground while rigidly ensuring that external noise does not exceed standards, achieving digital and automated management. This forms a dynamic equilibrium system integrating perception, decision-making, and execution.

[0027] Noise monitoring micro-stations (sensing terminals): Deployed in sensitive areas surrounding the playground (such as the boundary of adjacent residential buildings). Each micro-station integrates a high-precision noise microphone, a weather sensor (optional), and a LoRaWAN or 4G communication module. Its core function is to collect noise equivalent sound level (Leq), maximum sound level (Lmax), spectrum, and other data in real time and upload them via a wireless network. The micro-stations adopt a low-power design and can be powered by mains electricity or solar panels, enabling long-term unattended operation.

[0028] Directional sound array (execution end): Deployed at the boundary of the playground, it adopts phased array beamforming technology or parametric array speaker array. The system can accurately project sound energy into the playground activity area in a narrow beam (such as within ±15°), while the sound pressure level in the sides and rear can drop by more than 20 decibels, achieving a "sound spotlight" effect.

[0029] Intelligent Edge Control Box (Edge Decision Node): Deployed in the campus computer room, serving as the local control hub. Its core is an IoT controller with edge computing capabilities and a built-in lightweight AI inference engine. It is responsible for receiving cloud commands and driving the directional acoustic array; simultaneously, in the event of a network outage, it can execute local closed-loop control based on preset rules and the last effective strategy, ensuring uninterrupted basic system functions and greatly enhancing system robustness.

[0030] Cloud-based intelligent management and control platform (cloud-based decision-making and display center): Serving as the system's "intelligent brain." The platform aggregates all monitoring data, runs the core adaptive control algorithm, and stores historical data for analysis and learning. It provides a full-featured visual interface, including real-time noise maps, equipment status monitoring, alarm logs, statistical analysis reports, and remote policy configuration functions.

[0031] In some embodiments, multiple noise monitoring micro-stations are set up. Each noise monitoring micro-station integrates a noise microphone, a meteorological sensor, and a wireless communication module to collect noise equivalent sound level, maximum sound level, spectrum data, and meteorological data in real time, and uploads them through a wireless network.

[0032] In some embodiments, multiple directional sound arrays are provided, employing phased array beamforming technology or parametric array speaker arrays to project sound energy into the controlled area in a narrow beam, thereby reducing the lateral and rearward sound pressure levels; the cloud-based intelligent control platform is equipped with a visual interface to display real-time noise maps, equipment status monitoring, alarm logs and statistical analysis reports, and provides remote policy configuration functions.

[0033] In some embodiments, the triple intelligent adaptive control algorithm includes the following steps: Dynamic fast response steps: When the noise data meets the preset exceeding conditions, the system enters the dynamic fast response mode and generates a control command to lower the output parameters of the directional acoustic array; Trial optimization step: When the noise data continuously meets the preset safety margin condition within the preset safety time, the system enters the trial optimization mode and generates a control command to adjust the output parameters of the directional acoustic array in order to find the optimal output. Steady-state fine optimization steps: When the system is not in dynamic fast response mode or trial optimization mode, and the noise data fluctuates within the preset range for more than the stable time, and there are no records of exceeding the standard trigger, the output parameters of the directional acoustic array are slightly adjusted according to the preset optimization rules to reach the long-term optimal balance point and maximize the acoustic environment benefits.

[0034] In this embodiment, a triple intelligent adaptive control algorithm is employed to achieve a closed loop of "real-time perception - intelligent decision-making - precise execution". The algorithm includes three stages: dynamic rapid response, tentative optimization, and steady-state fine optimization. The definitions of some parameters involved in the triple intelligent adaptive control algorithm are shown in Table 1. Table 1

[0035] In some embodiments, the preset exceeding conditions and response mechanism in the dynamic fast response step are specifically as follows: When the real-time equivalent sound level of any noise monitoring micro-station L current satisfy L current > L th +Δ, and if the duration exceeds the set time threshold, a downward control is triggered; or, When the real-time equivalent sound level of any noise monitoring micro-station L current satisfy L current > L th When +2Δ is applied, the downward control is immediately triggered; or, When the calculated noise change trend meets the preset trend conditions, the pre-volume reduction control is triggered in advance, and a dynamic reduction step size is generated. Step down : And limit the maximum downregulation to prevent mutations; in L th Δ is the noise threshold, and Δ is the dead zone threshold. Step base Adjust the step size based on the base. k This represents the dynamic step-size gain coefficient.

[0036] In this embodiment, the criteria for determining whether a value exceeds the limit are as follows: Triggering condition A (main condition): Any monitoring point i satisfies L current(i,t) > L th +Δ and last for ≥2 seconds (to prevent accidental triggering of transient spikes).

[0037] Triggering condition B (acceleration condition): If a monitoring point exists L current(i,t) > L th +2Δ will trigger immediately, without waiting.

[0038] Triggering condition C (strengthening trend): If L trend >2dB / 3s (preset trend condition) and the current value has entered the dead zone, triggering pre-dropout early. Dynamically adjust the step size. Step down The physical meaning of the calculation formula is: the more the standard is exceeded, the faster the reduction will be, but the maximum reduction is limited to ≤ 5dB (to prevent sudden changes).

[0039] Example: If L current =70, L th =65, Δ=2, Step base =1.5, k =0.3, then: .

[0040] Control output and execution method: Control commands can include adjustment modes (uniform adjustment or zoned adjustment) and adjustment amounts. Execution chain: Cloud / Edge → Edge control box (parses commands) → sent to directional acoustic array via Modbus / RTU or TCP protocol → array DSP adjusts output level.

[0041] In some embodiments, the preset safety margin condition in the trial-and-error optimization step is: the real-time noise values ​​of all noise monitoring microstations meet the real-time equivalent sound level for a continuous period exceeding a preset time. L current Noise threshold L th - Dead zone threshold Δ; or, simultaneously satisfying that the sound pressure level of the reference point within the control area is lower than the preset target value; The optimization strategy for the exploratory optimization step includes the following steps: With a preset trial step size Step up Increase the current volume V current Get trial output V test ; During the preset observation period, monitor the real-time equivalent sound level of all noise monitoring micro-stations. L current ; If all monitoring data consistently meet the safety margin requirements, then this upgrade will be accepted. V current = V test And repeat the above steps; If, after a certain upgrade, any monitoring data enters the dead zone or fails to meet the preset safety margin condition, the output parameter value before the upgrade is returned and locked as the current optimal value. Then, it is determined whether the exit condition is met. If it is met, the optimization mode is exited and the system enters a steady state.

[0042] In this embodiment, the exit condition is two consecutive rollbacks or reaching a preset upper volume limit. The optimization trigger condition is: all monitoring points continuously meet the condition for T ≥ 5 minutes. L current < L th −Δ; Additional on-campus feedback condition (optional, the on-campus feedback condition must be met in addition to the previous condition): On-campus remote reference microphone L school < L school_target (e.g., <60dB) indicates insufficient sound pressure level within the school; among which L school The sound pressure level at the far end of the campus is the equivalent sound pressure level actually measured by a reference microphone deployed within the controlled area (such as the far end of the playground). The default value is the real-time measurement value, which is used to reflect the actual volume felt by the audience in the venue. L school_target The target sound pressure level within the school is the minimum sound pressure level expected to ensure normal hearing within the controlled area (such as the far end of the playground), used in conjunction with... L school The comparison determines whether the volume needs to be increased; the default value is 60dB, and the reference range is 50-75dB. L school < L school_target Furthermore, when the ambient noise level meets the safety margin, the system triggers a trial boost to optimize the listening experience within the school.

[0043] Optimization Strategy (Simplified Version of Hill Climbing Method): 1. Initialization: Step up =0.5dB, V current The current volume.

[0044] 2. Probing action: V test = V current + Step up .

[0045] 3. Waiting observation period: Continuously monitor for 10 seconds and record data from all monitoring points. L current .

[0046] 4. Judgment rule: If all monitoring points consistently meet the following conditions... L current < L th -Δ → Accept this promotion, V current = V test Repeat step 2.

[0047] If any monitoring point enters a dead zone or exceeds the standard ( L current ≥ L th If the value is -Δ, then the output parameter value before the current improvement is returned to its previous value and locked as the current optimal value. Then, it is checked whether the exit condition is met. If it is, the optimization mode is exited and a steady state is entered. The optimization objective function is: This means ensuring that the noise level at all monitoring points is below the noise threshold. L th Under the premise of maximizing the average sound pressure level within the school L school_avg .in For all times, the constraint condition must be satisfied at any point in time, that is, there must be no instantaneous exceedance during the entire control process; The maximum noise level at the peripheral monitoring points is the measured equivalent sound level of all noise monitoring micro-stations deployed in sensitive areas (such as residential area boundaries). L current The maximum value in the range is used to ensure that even the most sensitive points do not exceed the limit.

[0048] Objective function interpretation: This invention maximizes the average sound pressure level within the campus by probing and increasing steps, while ensuring that the noise at all external monitoring points remains below a set threshold. This design satisfies both rigid requirements (no disturbance to residents) and flexible optimization (optimal listening experience within the campus).

[0049] In some embodiments, the preset range in the steady-state fine optimization step is: L th ±Δ / 2; the preset optimization rule is: periodically based on the average equivalent continuous sound level over a past period of time. Leq avg and the number of times the instantaneous value exceeds the limit Lmax count Perform fine adjustments to keep the average ambient noise level close to the noise threshold. L thBut it does not exceed the standard level.

[0050] In this embodiment, the definition of entering steady state is: the system is neither in fast response mode nor in trial-and-error optimization mode; the noise values ​​at all monitoring points are within... L th Fluctuations within ±Δ / 2 lasted for more than 3 minutes; no records of exceeding the limit were found.

[0051] Optimization method: Input: data from the past 5 minutes Leq avg : Equivalent continuous sound level average value Lmax count : Number of times the instantaneous value exceeds the limit, i.e., the instantaneous value exceeds L th The number of times −Δ / 2, The steady-state optimization rule is: ,in is the volume fine-tuning amount, the change in volume with each adjustment (positive for increase, negative for decrease), in dB; otherwise represents all other cases, indicating all cases other than the two conditions in the steady-state optimization rule; Rule Interpretation: Condition 1: When the average noise level is below the safety lower limit ( L th If there is no record of instantaneous value exceeding the limit (−Δ), it indicates that the current volume is too low and the environment is stable, and the system is boosted by 0.2 dB; Condition 2: When the instantaneous value exceeds the limit more than 3 times, it indicates that noise spikes are frequent, and the system drops by 0.2dB; Other cases: Volume remains unchanged; This rule, through small and gradual adjustments, makes the long-term average noise approach a certain value. L th −Δ / 2 maximizes the acoustic environment benefits without exceeding the limits. The adjustment cycle is typically once every 2 minutes, with each adjustment amplitude not exceeding 0.2 dB, which is virtually imperceptible to the human ear.

[0052] The steady-state optimization objective is: That is, to keep the ambient noise close to the threshold but not exceed the standard for a long period of time, so as to maximize the benefits of the acoustic environment.

[0053] In some embodiments, the cloud-based intelligent management and control platform and the intelligent edge control box collaborate using either a policy-based distribution mode or a target-based distribution mode. In the policy-based distribution mode, the cloud-based intelligent management and control platform distributes specific control quantities. In the target-based distribution mode, the cloud-based intelligent management and control platform distributes threshold values ​​and scene tags, and the intelligent edge control box calculates the control quantities based on local real-time data. The cloud-based intelligent management and control platform automatically adjusts various control parameters in a targeted manner based on preset time period strategies or activity scenario strategies; The cloud-based intelligent management and control platform constructs a time series prediction model based on historical noise data to predict future noise peak times and adjusts the output parameters of the directional acoustic array in advance to smooth the control curve.

[0054] In this embodiment, exception handling and collaborative logic: local closed loop of the "edge" when the network is interrupted: the edge control box has a built-in simplified rule engine that stores the last valid control parameters received from the cloud (such as...). L th , Δ, Step base When the network is interrupted, the edge control box independently runs a local threshold judgment program, performs automatic reduction or conservative optimization based on locally collected noise data, and caches the action logs. After the network is restored, the edge control box automatically uploads the cached logs and synchronizes them with the cloud policy.

[0055] Cloud-Edge Decision-Making Division (Two Modes): **Policy Deployment Mode (Recommended):** The cloud calculates precise control values ​​and deploys them, while the edge handles execution and local closed-loop processing; suitable for scenarios with limited edge computing power and reliance on cloud algorithms. **Target Deployment Mode:** The cloud only deploys thresholds and scenario labels, while the edge calculates control values ​​based on local real-time data; suitable for scenarios with strong edge computing power and requiring low-latency responses.

[0056] Time-based and scenario-based strategies: The system can preset various strategies, such as: Time-based strategy: automatically adjusts according to the course schedule. L th During class L th =70dB, during lunch break L th =60dB. Activity scenario: When the "Sports Meet" mode is selected on the platform, the system automatically switches to wide beam, high volume, and enables a more conservative threshold.

[0057] Learning and Prediction: Based on historical noise data, the cloud platform uses time series prediction models (such as LSTM) to predict the peak noise times of daily activities such as morning exercises, and starts to slowly increase the volume 10 minutes in advance to make the control curve smoother.

[0058] In some embodiments, the triple intelligent adaptive control algorithm further incorporates an environmental factor compensation mechanism, including: Wind speed compensation: When the wind speed exceeds the set wind speed and the wind direction is towards a preset sensitive area near the controlled area, the noise threshold is adjusted. L th Introduce a compensation offset; Temperature and humidity compensation: When the temperature change exceeds the set temperature and humidity, the beam pointing angle of the directional acoustic array is finely adjusted.

[0059] In this embodiment, wind speed compensation: if the wind speed is >3 m / s and the wind direction is from the playground to the residential area, the monitoring value may be affected by wind noise interference or sound propagation attenuation. The control algorithm introduces a +1dB compensation offset to avoid false sound reduction.

[0060] Temperature and humidity compensation: speed of sound If the temperature change exceeds ±5°C, the beam pointing angle can be finely adjusted (e.g., ±2°) to ensure coverage of the target area within the campus.

[0061] In some embodiments, the cloud-based intelligent management and control platform also runs a sound source identification and filtering algorithm to extract features of the collected audio segments and identify the sound source type through a classifier; if it is identified as non-target type noise and the confidence level is higher than the confidence level threshold, the noise data of the monitoring point will not participate in the control decision.

[0062] In this embodiment, sound source identification and filtering: audio feature extraction and recognition: the micro-station collects a 3-second audio clip (16kHz, 16bit), extracts features such as MFCC, and identifies the sound source type through a lightweight CNN or GMM classifier: traffic noise, construction noise, natural sound, human voice / broadcast, etc.

[0063] Filtering logic: If the noise is identified as non-target noise (traffic / construction) and the confidence level is higher than the threshold (e.g., >0.8), then the noise data of that monitoring point is filtered out. L current It does not participate in control decisions, but only records logs; otherwise, it participates in control normally.

[0064] Effect: Prevents the school's broadcast system from being mistakenly lowered due to horns blaring on roads outside residential areas, thus improving system robustness.

[0065] The following is an example of an application in a campus environment, such as... Figure 3 The image shows a real-world application case study of 25 directional sound units, 7 noise monitoring micro-stations, and 1 intelligent edge control box. These 25 directional sound units, 7 noise monitoring micro-stations, and 1 intelligent edge control box together constitute a campus directional sound intelligent control system covering two playgrounds.

[0066] Working process: 25 directional sound devices are deployed around the playground to form a directional sound array, which constitutes the acoustic environment of the playground when emitting sound. 7 micro-stations are deployed outside the directional sound array to monitor the noise data of the playground's perimeter and upload it to the cloud platform in real time. The platform generates control strategies based on the noise data and sends them to the edge intelligent control box, which dynamically controls the volume of the directional sound devices. This process is repeated cyclically to ensure that the acoustic environment of the playground does not interfere with the outside world and has no impact on the volume requirements of activities within the playground. At the same time, the platform generates a real-time visualized noise map based on the noise data, showing the changes in the acoustic environment of the school playground.

[0067] Data definition and protocol format: Terminal (microsite) → Cloud (MQTT / HTTP): As an example, uploaded data can include the following fields: device identifier, timestamp, equivalent sound level, maximum sound level, spectrum data, battery voltage, signal strength, location information, meteorological data, etc. For example: Device ID: MS_201; Timestamp: 2026-03-11T14:35:20Z; Leq_1s: 62.3 dB; Lmax_1s: 68.5 dB; Spectrum: [45.2, 48.1, 52.3, ...]; Battery voltage: 12.4V; GPS: 30.5721,104.0665; Temperature: 22.5°C; Wind speed: 2.1 m / s.

[0068] Cloud → Edge (Edge Control Box) Control Commands: As an example, control commands may include command identifiers, command types (e.g., volume adjustment), adjustment modes (global / zone), adjustment amounts, target areas, and policy parameters. For example: Command ID: CTRL_202603111435; Timestamp: 2026-03-11T14:35:30Z; Command Type: Volume Adjustment; Adjustment Mode: Zone; Adjustment Amount: -2.2 dB; Target Area: Zone A, Zone B; Policy: {Threshold: 65 dB, Dead Zone: 2 dB}; Validity Period: 2026-03-11T14:45:00Z.

[0069] Edge → Execution End (Directional Sound Array) Drive Commands: The edge control box sends drive commands to the directional sound array via standard industrial protocols (such as Modbus RTU / TCP). For example, the volume value of each sound column can be set by writing to the holding register via TCP. The specific register address and value mapping can be defined according to the device, and this invention does not limit this.

[0070] The above description is merely a preferred embodiment of the present invention. It should be understood that the present invention is not limited to the forms disclosed herein and should not be construed as excluding other embodiments. It can be used in various other combinations, modifications, and environments, and can be altered within the scope of the concept described herein through the above teachings or related technologies or knowledge. Modifications and variations made by those skilled in the art that do not depart from the spirit and scope of the present invention should be within the protection scope of the appended claims.

Claims

1. A directional sound intelligent adaptive control system based on cloud-edge-device collaboration, characterized in that: This includes a noise monitoring micro-station, which is connected to a cloud-based intelligent management and control platform. The cloud-based intelligent management and control platform is connected to an intelligent edge control box, which is connected to a directional sound array. The noise monitoring micro-stations are deployed in preset sensitive areas near the control area to collect noise data in real time; The directional acoustic array is deployed at the boundary of the controlled area to project sound into the controlled area in the form of directional beams. The intelligent edge control box is deployed locally as a local control hub, and has a built-in lightweight AI inference engine and simplified rule engine. When the network connection is normal, the intelligent edge control box receives and executes control commands issued by the cloud intelligent management and control platform. When the network connection is interrupted, the intelligent edge control box independently runs local closed-loop control based on the last stored valid control parameters, and automatically synchronizes logs and policies after the network is restored. The cloud-based intelligent management and control platform operates a triple intelligent adaptive control algorithm, which dynamically adjusts the output parameters of the directional acoustic array based on the comparison results of noise data and noise threshold, thereby forming a closed-loop management and control system based on real-time perception and intelligent decision-making.

2. The directional sound intelligent adaptive control system based on cloud-edge-device collaboration according to claim 1, characterized in that: Multiple noise monitoring micro-stations are set up. Each noise monitoring micro-station integrates a noise microphone, a meteorological sensor, and a wireless communication module to collect noise equivalent sound level, maximum sound level, spectrum data, and meteorological data in real time, and uploads them through a wireless network.

3. The directional sound intelligent adaptive control system based on cloud-edge-device collaboration according to claim 1, characterized in that: Multiple directional sound arrays are configured, employing phased array beamforming technology or parametric array speaker arrays to project sound energy into the controlled area in a narrow beam, thereby reducing the lateral and rearward sound pressure levels. The cloud-based intelligent management and control platform is equipped with a visual interface to display real-time noise maps, equipment status monitoring, alarm logs, and statistical analysis reports, and provides remote policy configuration functions.

4. The directional sound intelligent adaptive control system based on cloud-edge-device collaboration according to claim 1, characterized in that: The aforementioned triple intelligent adaptive control algorithm includes the following steps: Dynamic fast response steps: When the noise data meets the preset exceeding conditions, the system enters the dynamic fast response mode and generates a control command to lower the output parameters of the directional acoustic array; Trial optimization step: When the noise data continuously meets the preset safety margin condition within the preset safety time, the system enters the trial optimization mode and generates a control command to adjust the output parameters of the directional acoustic array in order to find the optimal output. Steady-state fine optimization steps: When the system is not in dynamic fast response mode or trial optimization mode, and the noise data fluctuates within the preset range for more than the stable time, and there are no records of exceeding the standard trigger, the output parameters of the directional acoustic array are slightly adjusted according to the preset optimization rules to reach the long-term optimal balance point and maximize the acoustic environment benefits.

5. The directional sound intelligent adaptive control system based on cloud-edge-device collaboration according to claim 4, characterized in that: The preset exceeding conditions and response mechanism in the dynamic fast response step are as follows: When the real-time equivalent sound level of any noise monitoring micro-station L current satisfy L current > L th +Δ, and when the duration exceeds the set time threshold, the control is lowered. or, When the real-time equivalent sound level of any noise monitoring micro-station L current satisfy L current > L th At +2Δ, the downward control is immediately triggered; or, When the calculated noise change trend meets the preset trend conditions, the pre-volume reduction control is triggered in advance, and a dynamic reduction step size is generated. Step down : And limit the maximum downregulation to prevent mutations; in L th Δ is the noise threshold, and Δ is the dead zone threshold. Step base Adjust the step size based on the base. k This represents the dynamic step-size gain coefficient.

6. The directional sound intelligent adaptive control system based on cloud-edge-device collaboration according to claim 4, characterized in that: The preset safety margin condition in the exploratory optimization step is: the real-time noise value of all noise monitoring micro-stations meets the real-time equivalent sound level for a continuous period exceeding a preset time. L current Noise threshold L th - Dead zone threshold Δ; or, simultaneously satisfying that the sound pressure level of the reference point within the control area is lower than the preset target value; The optimization strategy for the exploratory optimization step includes the following steps: With a preset trial step size Step up Increase the current volume V current Get trial output V test ; During the preset observation period, monitor the real-time equivalent sound level of all noise monitoring micro-stations. L current ; If all monitoring data consistently meet the safety margin requirements, then this upgrade will be accepted. V current = V test And repeat the above steps; If, after a certain upgrade, any monitoring data enters the dead zone or fails to meet the preset safety margin condition, the output parameter value before the upgrade is returned and locked as the current optimal value. Then, it is determined whether the exit condition is met. If it is met, the optimization mode is exited and the system enters a steady state.

7. The directional sound intelligent adaptive control system based on cloud-edge-device collaboration according to claim 4, characterized in that: The preset range in the steady-state fine optimization step is: L th ±Δ / 2; the preset optimization rule is: periodically based on the average equivalent continuous sound level over a past period of time. Leq avg and the number of times the instantaneous value exceeds the limit Lmax count Perform fine adjustments to keep the average ambient noise level close to the noise threshold. L th But it does not exceed the standard level.

8. The directional sound intelligent adaptive control system based on cloud-edge-device collaboration according to any one of claims 1-7, characterized in that: The cloud-based intelligent management and control platform and the intelligent edge control box collaborate using either a policy-based or target-based approach. In the policy-based approach, the cloud-based intelligent management and control platform issues specific control parameters. In the target-based approach, the cloud-based intelligent management and control platform issues threshold values ​​and scene tags, and the intelligent edge control box calculates the control parameters based on local real-time data. The cloud-based intelligent management and control platform automatically adjusts various control parameters in a targeted manner based on preset time period strategies or activity scenario strategies; The cloud-based intelligent management and control platform constructs a time series prediction model based on historical noise data to predict future noise peak times and adjusts the output parameters of the directional acoustic array in advance to smooth the control curve.

9. The directional sound intelligent adaptive control system based on cloud-edge-device collaboration according to any one of claims 1-7, characterized in that: The aforementioned triple intelligent adaptive control algorithm also incorporates an environmental factor compensation mechanism, including: Wind speed compensation: When the wind speed exceeds the set wind speed and the wind direction is towards a preset sensitive area near the controlled area, the noise threshold is adjusted. L th Introduce a compensation offset; Temperature and humidity compensation: When the temperature change exceeds the set temperature and humidity, the beam pointing angle of the directional acoustic array is finely adjusted.

10. The directional sound intelligent adaptive control system based on cloud-edge-device collaboration according to any one of claims 1-7, characterized in that: The cloud-based intelligent management and control platform also runs a sound source identification and filtering algorithm, which is used to extract the features of the collected audio segments and identify the sound source type through a classifier; if it is identified as non-target type noise and the confidence level is higher than the confidence level threshold, the noise data of that monitoring point will not participate in the control decision.

Images