Dormitory noise monitoring data analysis method and system

By constructing a noise spatiotemporal distribution characteristic map and a multi-level noise source positioning method in the dormitory environment and dynamically adjusting the noise evaluation standards, the problems of insufficient accuracy and poor adaptability in dormitory noise monitoring are solved, accurate noise identification and environmental optimization are achieved, and the adaptability and privacy protection of noise management are improved.

CN120748446AInactive Publication Date: 2025-10-03SHENZHEN AOSIEN PURIFYING TECH CO LTD
View PDF 0 Cites 3 Cited by

Patent Information

Application Number
CN202511165846.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-08-20
Publication Date
2025-10-03
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

Existing noise monitoring technology lacks customized solutions in dormitory environments and cannot adapt to the differentiated needs of different time periods and activity states. It also has problems with insufficient accuracy in noise source location, type identification, and impact assessment, making it difficult to provide accurate basis for noise control.

Method used

By extracting features of noise signals in the dormitory space, constructing a noise spatiotemporal distribution feature map, combining multi-level noise source positioning methods and integrated learning prediction models, dynamically adjusting noise evaluation standards, realizing pattern analysis and trend prediction of noise source tracing data, and establishing a closed-loop feedback system for noise control.

Benefits of technology

It achieves accurate identification and evaluation of dormitory noise, improves the accuracy and adaptability of noise monitoring, can prevent noise interference in advance, optimize the noise environment, and solves the problems of insufficient adaptability and privacy protection of traditional methods in dormitory environments.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN120748446A_ABST
    Figure CN120748446A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of data analysis, and discloses a dormitory noise monitoring data analysis method and system, and the method comprises the steps: carrying out the noise signal feature extraction of a dormitory space, and obtaining a noise type recognition result and noise parameter data; analyzing the distribution characteristics of the noise in time and space to obtain a noise space-time distribution characteristic spectrum and a noise standard exceeding judgment result; determining a noise source area and a noise type attribute based on the noise space-time distribution characteristic spectrum and a noise standard exceeding judgment result, and obtaining dormitory noise traceability data; performing mode analysis and trend prediction on the dormitory noise traceability data to obtain dormitory noise early warning information and high-incidence time period prediction data; and creating a dormitory noise environment optimization scheme according to the dormitory noise early warning information and the high-incidence time period prediction data. Therefore, context perception evaluation of noise interference is realized, the limitation that a traditional method neglects the environment context is overcome, and the accuracy of dormitory noise monitoring data analysis is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of data analysis technology, and in particular to a dormitory noise monitoring data analysis method and system. Background Art

[0002] Existing noise monitoring technologies are mainly aimed at scenarios such as industrial environments, urban roads, and community residences, and lack customized solutions for the special environment of dormitories. Traditional noise monitoring methods usually use a single fixed threshold judgment standard, which cannot adapt to the differentiated noise requirements in different time periods and different activity states in dormitory environments. At the same time, existing technologies have problems such as insufficient accuracy and poor adaptability in noise source positioning, noise type identification, and noise impact assessment, making it difficult to provide an accurate basis for noise control. In addition, traditional noise monitoring systems generally lack the ability to perceive the patterns of student activities and are unable to dynamically adjust noise evaluation standards according to the characteristics of different functional periods in the dormitory, resulting in poor noise control effects. Summary of the Invention

[0003] The present application provides a dormitory noise monitoring data analysis method and system, thereby realizing context-aware evaluation of noise interference, overcoming the limitation of traditional methods that ignore environmental context, and improving the accuracy of dormitory noise monitoring data analysis.

[0004] A first aspect of the present application provides a dormitory noise monitoring data analysis method, the dormitory noise monitoring data analysis method comprising: Noise feature extraction is performed on the original noise signal collected in the dormitory space to obtain noise type identification results and noise parameter data; Analyzing the distribution characteristics of noise in time and space based on the noise type identification result and the noise parameter data to obtain a noise spatiotemporal distribution characteristic map and a noise exceeding standard determination result; Determine the noise source area and noise type attributes based on the noise spatiotemporal distribution characteristic map and the noise exceeding standard determination result, and obtain dormitory noise source tracing data; Performing pattern analysis and trend prediction on the dormitory noise source tracing data to obtain dormitory noise warning information and high-incidence period prediction data; A dormitory noise environment optimization plan is created based on the dormitory noise warning information and high-incidence period prediction data.

[0005] A second aspect of the present application provides a dormitory noise monitoring data analysis system, the dormitory noise monitoring data analysis system comprising: The feature extraction module is used to extract noise features from the original noise signals collected in the dormitory space to obtain noise type identification results and noise parameter data; An analysis module is used to analyze the distribution characteristics of noise in time and space according to the noise type identification result and the noise parameter data, and obtain a noise spatiotemporal distribution characteristic map and a noise exceeding standard determination result; A processing module is used to determine the noise source area and noise type attributes based on the noise spatiotemporal distribution characteristic map and the noise exceeding standard determination result, and obtain dormitory noise source tracing data; A prediction module is used to perform pattern analysis and trend prediction on the dormitory noise source tracing data to obtain dormitory noise warning information and high-incidence period prediction data; A creation module is used to create a dormitory noise environment optimization plan based on the dormitory noise warning information and high-incidence period prediction data.

[0006] Compared with the prior art, the present application has the following beneficial effects: Different from the traditional fixed threshold noise monitoring method, the present invention dynamically adjusts the noise evaluation standard according to the characteristics of different functional time periods in the dormitory (study, rest, sleep), realizes the accurate identification and evaluation of dormitory noise, and effectively solves the technical problem of insufficient adaptability of traditional methods in specific scenarios. Through the two-stage computational framework of progressive constraint relaxation, the present invention decomposes the noise analysis task into two stages: coarse-grained noise event detection and fine-grained feature analysis, and adopts a hybrid representation method to process different types of noise parameters, which significantly reduces the computational complexity of the system and solves the technical problem of real-time processing of large-scale noise data. The present invention constructs a time period activity perception matrix and an activity switching protection interval, associates the noise pattern with the dormitory activity cycle, and introduces the dormitory activity state transition probability parameter, thereby realizing context-aware evaluation of noise interference and overcoming the limitation of traditional methods that ignore environmental context. Based on a multi-level noise source localization method, combined with sound intensity gradient analysis and triangulation techniques, this invention accurately identifies noise source areas, overcoming the limited localization accuracy of traditional methods in complex indoor environments. By analyzing the temporal and spatial dimensions of noise source tracing data and integrating it with an ensemble learning prediction model, the invention can identify noise patterns and predict peak periods, shifting from post-processing to pre-emptive prevention, addressing the technical bottleneck of traditional noise management, which relies on passive response. This invention establishes a closed-loop feedback system for noise control, utilizing an intelligent recommendation engine for control strategies and an A / B testing evaluation mechanism. This system continuously optimizes control strategies and achieves sustained improvement in dormitory acoustic environments, overcoming the technical limitations of traditional noise management, which lacks adaptive optimization capabilities. During noise monitoring and analysis, the invention only extracts acoustic feature parameters to determine noise type and location; the original sound content is not recorded or stored. This effectively resolves the technical contradiction between environmental monitoring and privacy protection by implementing regional noise control rather than individual identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0008] The structures, proportions, sizes, etc. depicted in the drawings of this specification are only used to match the contents disclosed in the specification so as to facilitate understanding and reading by persons familiar with this technology. They are not intended to limit the conditions under which the present invention can be implemented and therefore have no substantive technical significance. Any structural modifications, changes in proportional relationships, or adjustments in size should still fall within the scope of the technical contents disclosed in the present invention without affecting the effects and objectives that can be achieved by the present invention.

[0009] Figure 1 1 is a flow chart of a dormitory noise monitoring data analysis method provided by an embodiment of the present invention; Figure 2 It is a schematic block diagram of the structure of the dormitory noise monitoring data analysis system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0010] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0011] The flowcharts shown in the accompanying drawings are for illustrative purposes only and do not necessarily include all contents and operations / steps, nor must they be executed in the order described. For example, some operations / steps may be decomposed, combined, or partially merged, so the actual execution order may vary depending on the actual situation.

[0012] It should also be understood that the terms used in this specification are for the purpose of describing specific embodiments only and are not intended to limit the present application. As used in this specification and the appended claims, the singular forms "a," "an," and "the" are intended to include the plural forms unless the context clearly indicates otherwise.

[0013] It should be further understood that the term "and / or" used in this specification and the appended claims refers to any and all possible combinations of one or more of the associated listed items, and includes these combinations. Figure 1In one embodiment of the dormitory noise monitoring data analysis method of the present application, the following steps are included: Step 100: extract noise features from the original noise signal collected in the dormitory space to obtain noise type identification results and noise parameter data; It is understandable that the execution subject of this application can be a dormitory noise monitoring data analysis system, or a terminal or a server, which is not limited here. The embodiment of this application is described by taking the server as the execution subject as an example.

[0014] Specifically, based on an analysis of the dormitory's spatial acoustic structure, an ECM high-sensitivity microphone was placed on the ceiling at the geometric center of each dormitory. Additional auxiliary collection nodes were deployed in corridors and public areas, creating a high-fidelity sampling network that covers the entire space and provides 360-degree omnidirectional sound pickup. During the data collection phase, each node performs real-time sampling at a high sampling rate of 48kHz and automatically segments the signal within the sound pressure range of 20–130dB. Instantaneous sound pressure data is collected every 10 seconds, and a complete sampling record with dormitory number, floor identification, and timestamp is generated every minute to ensure spatial data annotation and temporal continuity. The sampled data is then filtered through an A-weighted filter to remove frequencies insensitive to the human ear and highlight mid- and high-frequency information. Based on this data, the equivalent continuous A-level (LAeq) is calculated, and acoustic parameters such as Lmax, Lmin, L10, L50, and L90 are statistically analyzed to characterize the noise distribution and fluctuation characteristics, forming preliminary noise acoustic signature data. A built-in interference detection mechanism analyzes short-term amplitude gradients at sound pressure abrupt changes and, combined with a spectral feature comparison model, identifies and removes non-continuous interference signals, such as door and window openings, dropped objects, and brief movements of people, in real time, preserving stable, valid sound source signals. For signals identified as valid, a wavelet transform is applied across the entire frequency band using multi-scale frequency band decomposition technology to effectively distinguish the target spectrum from background noise. Suppression filtering is then applied within the background frequency band to achieve dynamic noise reduction and restore a cleansed version of the signal, generating noise-cleaned data. Contextual relevance testing is performed on the cleansed data to assess structural continuity and logical consistency within the time window. Collaborative cross-validation is performed on data from multiple sensor nodes over the same time period to eliminate inconsistent and anomalous sampling segments. After local pre-processing, data is uploaded to a cloud data center via a GPRS wireless module or wired Ethernet. In the event of a network outage, a resumable transmission mechanism automatically activates, caching data locally and retransmitting it in batches upon restoration, ensuring data integrity and temporal continuity. The data uploaded to the server constitutes the noise preprocessing data input, which is then taken over by the subsequent feature extraction and classification analysis module. The system uses this preprocessed data to extract time-domain features (such as root mean square, crest factor, and waveform factor), frequency-domain features (such as power spectral density and frequency band energy distribution), and joint time-frequency features (an energy spectrum matrix based on wavelet packet decomposition and short-time Fourier transform). These are then input into a trained noise classification model. K-means clustering and support vector machine models are then applied in stages for coarse and fine classification. The final output is noise parameter data, which includes key information such as noise type identification results (such as human voice, equipment sound, music sound, etc.), occurrence time, signal confidence, and a list of characteristic parameters.

[0015] Based on long-term historical dormitory noise data, combined with dormitory schedules and student behavior patterns, the day is divided into study, rest, and sleep periods. Taking into account the differences in noise tolerance within each period, a differentiated noise tolerance function is constructed, which in turn forms a noise threshold model for dormitory functional periods. This model uses dB as the base unit and introduces a weighted disturbance index to set specific penalty factors for specific noise types. For example, a lower tolerance limit is set for high-frequency conversations and electronic audio during sleep than during the day. Based on this threshold model, the noise preprocessing data is input into a two-stage analysis framework with progressive constraint relaxation. In the first stage, this framework preliminarily marks suspected noise events through a fixed threshold detection mechanism. In the second stage, constraints are relaxed and redundancy is eliminated from the pre-labeled data through local noise fluctuation analysis and time series consistency checks. This generates structured preliminary noise event marker data and labels each event record with the functional period to which the corresponding time point belongs. Noise events exceeding the threshold for the current time period are screened and subjected to fine-grained feature analysis, including pitch contour changes, burst duration, and spectral concentration. This extracts key noise event feature data, which reveals the core disturbance causes and behavioral pattern characteristics of the noise. To fully represent the unique complexity of dormitory noise, discrete features such as conversational tone fluctuations, electronic device audio output patterns, and door opening and closing rhythms are used as discrete feature variables. Meanwhile, continuous variables such as the mean fluctuation, low-frequency energy content, and duration of ambient background noise are used to construct a hybrid representation of noise feature data. This hybrid feature data is then fed into a context-sensitive dual-constraint allocation model. This model leverages prior-trained noise type feature labels and incorporates activity state transition probability parameters within the dormitory, such as the transition probability from quiet study to light socializing or from rest to deep sleep. Within this transition framework, the current noise disturbance level and state conflict are evaluated, resulting in a context-sensitive noise impact assessment. Based on the modeling and analysis of daily dormitory activity patterns, the system identifies behavioral transition time windows that occur multiple times a day, such as the transition from evening study to nighttime rest and from nighttime rest to deep sleep, and establishes an activity switching protection interval accordingly. Within this interval, the system automatically improves the detection sensitivity to weak noise disturbances, lowers the original noise threshold, and introduces a dynamic noise weight adjustment mechanism to generate noise sensitivity adjustment parameters for the activity switching period. This parameter is then applied to the correction process of the aforementioned context-related noise impact assessment results to generate a more sensitive and accurate switching perception noise interference assessment result for the behavioral transition period. The switching perception noise interference assessment result is fused with the noise type feature data, and through feature reweighting and confidence reassignment algorithms, a noise type recognition result and noise parameter data are generated.

[0016] Step 200: Analyze the temporal and spatial distribution characteristics of noise based on the noise type identification result and noise parameter data to obtain a noise temporal and spatial distribution characteristic map and a noise exceeding standard determination result; Specifically, using noise type results and corresponding acoustic parameter data as basic components, combined with sensor data deployed at multiple points within the dormitory, a noise propagation association network was constructed to reflect the noise propagation relationship between sensor nodes. Each node in this network represents a monitoring point or dormitory unit, and each edge corresponds to the statistical correlation of noise propagation between two nodes within a certain period of time. This correlation is quantified by analyzing the synchronization and time delay of sound pressure intensity changes, forming a directed graph structure including node weights, edge correlations, and propagation directions. A multi-scale time window sliding analysis of the noise propagation association network was performed on the time dimension. The autocorrelation coefficient, fluctuation trend index, and periodic intensity of the noise type occurrence frequency were calculated at time scales of 10 minutes, 1 hour, and 24 hours, respectively. This generated a noise time evolution map. This map reveals the concentrated occurrence patterns of specific noise types on daily or weekly cycles and captures local fluctuation characteristics such as sudden noise at night and high-frequency social noise on weekends. Furthermore, using the synchronous noise variation data between different nodes in the aforementioned propagation correlation network, the cross-correlation coefficient, propagation delay, and energy attenuation parameters between each pair of monitoring points are calculated. Taking into account the sound insulation coefficient and path length of building structures (such as walls, doors, windows, and corridors), a graph-theoretic path construction algorithm is used to generate a noise spatial propagation path map. This path map depicts the complete path of noise diffusion from the source point to the surrounding area in physical space, including the degree of attenuation, propagation direction, and time delay distribution. The time evolution map and the spatial propagation path map are fused, and feature alignment and node embedding aggregation are performed using a tensor fusion framework or graph convolutional neural network to form a noise spatiotemporal distribution feature map. This map reflects the changing trend of noise in the temporal dimension and demonstrates its diffusion pattern within the building's spatial structure, achieving a dimensional upgrade from single-point data to a multidimensional data structure. Based on this map, the system performs noise exceedance determination, invoking daytime and nighttime noise limit standards (e.g., 45dB during the day and 37dB at night, as specified in GB 50118-2010). Combined with the system's internal dormitory functional segment model, the system dynamically sets thresholds based on the interrelationships between the current time period, noise type, and spatial location, and performs point-by-point determinations for each node in the map. Furthermore, to avoid misjudgments due to short-term anomalies or occasional disturbances, a duration filter for exceeding limits and a resonance confirmation mechanism for adjacent nodes are introduced. Only when noise intensity consistently exceeds the set threshold and forms a closed propagation chain within the spatial propagation path is a valid noise exceedance event confirmed. The output of this analysis process includes the type, intensity, duration, and spatial location of the exceeding noise, as well as its propagation path and impact range within the spatiotemporal map.

[0017] Step 300: Determine the noise source area and noise type attributes based on the noise spatiotemporal distribution characteristic map and the noise exceeding standard determination result, and obtain dormitory noise source tracing data; It should be noted that a multi-dimensional analysis is performed on the noise spatiotemporal distribution characteristic map, and the sound pressure level, sound energy intensity and its gradient distribution data of each node in a specific time window and spatial partition are extracted from the map. Based on the spatial variation trend of the sound energy density, a noise energy gradient function is constructed to perform energy differential analysis on each floor, dormitory unit, corridor and other areas of the entire building, thereby identifying the rising direction of sound energy and determining the center area of ​​sound source energy. This energy concentration area is defined as the first noise source area data, and its floor, spatial unit number and high-energy center position are marked. The first noise source area data is fused with the noise exceeding standard judgment results obtained in the early stage to screen out the concentrated area of ​​exceeding standard events and its corresponding time period, and at least three adjacent sensor nodes with spatial geometric conditions are selected from its periphery to extract the sound pressure level data and sound wave propagation time difference recorded by them during the time period. Combining building acoustic structural parameters, including wall insulation coefficients, door and window opening and closing status, and floor vibration characteristics, the system constructs a set of sound source localization equations based on propagation time differences and sound pressure attenuation. Triangulation is then used to calculate the spatial coordinates of the sound source point, further converging the noise localization range and generating secondary noise source area data, which is accurate to a specific dormitory unit or corridor section. A local acoustic array is constructed using primary dormitory sensors and auxiliary corridor sensors within this secondary noise source area. Combining sound source separation techniques with beamforming algorithms, the system performs phase alignment and path inversion on multiple source signals within the same time period, re-estimating the sound wave propagation path and directional distribution, thereby applying propagation path correction to the initial localization results. After sound field reconstruction and backpropagation optimization processing, the system outputs third noise source area data, corrected for sound field interference. This data is refined to specific locations within the dormitory, such as windows, bedside, or desk surfaces, providing more accurate sound source area identification. After spatially locating and extracting the high-confidence cleansed signal segment at the third noise source area, the multidimensional acoustic features in the time, frequency, and time-frequency domains are matched against a pre-built acoustic feature library containing reference sound source templates for various typical dormitory scenarios, such as the pitch contours of conversational sounds, the frequency band structure of musical sounds, and the steady-state spectra of electrical appliance operating sounds. Through feature similarity calculation and confidence interval assessment, the corresponding noise type and its key acoustic properties, including type label, frequency center of gravity, duration, and energy density, are determined to form noise type attribute data. This noise type attribute data is compared and integrated with the noise type identification results obtained earlier. The third noise source area data, the determined level of exceedance, and the spatial impact range are also introduced. Combined with metadata such as the event timestamp, floor number, and dormitory number, a complete set of dormitory noise source tracing data is generated.

[0018] Step 400: Perform pattern analysis and trend prediction on dormitory noise source tracing data to obtain dormitory noise warning information and high-incidence period prediction data; Specifically, a multi-level sequence decomposition process is performed on the long-term accumulated dormitory noise source tracing data in the temporal dimension. Using an improved seasonal decomposition algorithm (such as STL decomposition), the trend, period, and residual terms in the raw noise data are deconstructed to extract daily, weekly, and semester-specific patterns reflecting changes in dormitory noise intensity and type. Furthermore, spatial clustering and hotspot analysis are performed on the floor numbers, room numbers, and geographic locations annotated in the dormitory noise source tracing data to generate spatial distribution characteristics of dormitory noise. By constructing two-dimensional heat maps and three-dimensional sound pressure distribution maps, the spatial distribution characteristics of noise intensity are modeled, allowing the identification of typical high-incidence areas, such as dormitory units near elevators, entrance corridors, or concentrated entertainment areas on each floor. Together, temporal patterns and spatial characteristics form the spatiotemporal background foundation for noise evolution. Based on this, behavioral sequence mining was performed on the joint distribution of different noise types. Using association rule analysis and sequence pattern matching, we extracted dormitory noise behavioral pattern sequences. These sequences represent the stable occurrence order of certain noise types within specific time periods and spatial conditions, such as the typical gathering behavior sequence of "conversation → music → laughter → door slamming" or the chain of daily activities of "footsteps → collisions → equipment operation." These sequences not only describe the statistical co-occurrence relationships of noise but also reveal the causal logic of behavioral events along the timeline. The extracted noise behavioral pattern sequences were fed into an ensemble learning prediction model for trend forecasting. This model combines a long short-term memory (LSTM) network to capture temporal dependencies, XGBoost to capture nonlinear correlations, and a Prophet model to capture cyclical and seasonal trends. A weighted ensemble of multiple models was used to improve prediction accuracy. A sliding window strategy and error feedback mechanism were introduced during training to optimize prediction stability. The model outputs noise warning information on timescales such as one hour, four hours, and 24 hours into the future, encompassing key predictive factors such as the probability of exceeding standards, the type of noise that may occur, the spatial region, and the impact range. Based on the quantitative risk values ​​in this noise warning information, a multi-level hierarchical warning strategy is constructed. The prediction results are divided into three levels: low risk, medium risk, and high risk, based on the probability of noise exceeding standards and the degree of potential environmental impact. Different warning trigger thresholds, response times, and intervention intensities are set for each level, forming structured forecast data for high-incidence periods.

[0019] Step 500: Create a dormitory noise environment optimization plan based on dormitory noise warning information and high-incidence period prediction data.

[0020] Specifically, the noise warning information output by the prediction model and forecast data for high-incidence periods are fed into an intelligent control strategy recommendation engine. This engine, integrating decision tree analysis with reinforcement learning algorithms, comprehensively identifies multiple factors in the input data, including noise level distribution, noise type classification, spatial propagation characteristics, and trend intensity. Based on a multi-objective optimization mechanism, it generates multi-level noise control strategy recommendations. These recommendations include constraints for noise-sensitive periods, spatial layout optimization, individual behavioral interventions, and intelligent system-linked response plans. Specific treatment paths are also defined for different noise types, such as educational guidance strategies for human voice noise and intelligent noise reduction strategies for electronic sound sources. The preliminary control strategy recommendations are then processed hierarchically through a strategy classification and response level calculation module. Based on the noise disturbance level (mild, moderate, or severe), spatial weighting factors (such as location in high-density corridors or at the edge of quiet zones), and the urgency of the forecast period, all recommended strategies are mapped into tiered noise control implementation plans. These plans are structured as a hierarchical response model, corresponding to daily reminders, system linkage, and manual intervention levels. After the plan is confirmed, remote deployment of the control implementation plan is achieved through a PC or mobile platform. This plan is then linked to the intelligent dormitory noise control system to precisely implement the policy in the target area. During deployment, the corresponding execution functions are dynamically triggered based on the current level of noise violations detected. These include environmental prompts (such as intelligent sound bar alerts and light indicators), intelligent linkages (such as muting common area speakers and adjusting air conditioning fan speeds), and alert notifications (such as sending disturbance alerts to dormitory administrators or students via an app). All execution records are simultaneously generated into dormitory noise intervention records and stored in a database. Using these intervention records as a core sample, an evaluation index system is established encompassing four key dimensions: noise compliance rate, reduction rate of noise violations, student environmental satisfaction, and dormitory management efficiency. A comparison and analysis (AB) testing algorithms are then used to compare the effectiveness of different policy combinations under actual deployment conditions. During the AB testing process, the target dormitories are automatically divided into experimental and control groups. The improvement in compliance and intervention response frequency of different policies over the same period of time are compared to assess their actual effectiveness and calculate confidence intervals. The noise control strategy evaluation results are then output, serving as important feedback for subsequent policy optimization. The assessment results are matched and analyzed with the current dormitory environment. During the analysis process, the system calls the spatial semantic information stored in the environmental perception module, such as structural layout, dormitory population density, and functional area division. Combined with the noise disturbance characteristics and intervention feedback results, it performs in-depth correlation modeling between dormitory noise and overall environmental quality, explores the potential interaction mechanism between noise problems and air quality, lighting environment, and work and rest patterns, and then generates a dormitory noise environment optimization plan with three dimensions: structural suggestions, behavioral guidance, and intelligent strategy optimization.

[0021] In the embodiment of the present application, the present invention is different from the traditional fixed threshold noise monitoring method. It dynamically adjusts the noise evaluation standard according to the characteristics of different functional time periods (study, rest, sleep) in the dormitory, realizes the accurate identification and evaluation of dormitory noise, and effectively solves the technical problem of insufficient adaptability of traditional methods in specific scenarios. Through the two-stage calculation framework of progressive constraint relaxation, the present invention decomposes the noise analysis task into two stages: coarse-grained noise event detection and fine-grained feature analysis. It uses a hybrid representation method to process different types of noise parameters, significantly reducing the system's computational complexity and solving the technical problem of real-time processing of large-scale noise data. The present invention constructs a time period activity perception matrix and an activity switching protection interval, associates the noise pattern with the dormitory activity cycle, and introduces the dormitory activity state transition probability parameter, thereby realizing context-aware evaluation of noise interference and overcoming the limitation of traditional methods that ignore environmental context. Based on a multi-level noise source localization method, combined with sound intensity gradient analysis and triangulation techniques, this invention accurately identifies noise source areas, overcoming the limited localization accuracy of traditional methods in complex indoor environments. By analyzing the temporal and spatial dimensions of noise source tracing data and integrating it with an ensemble learning prediction model, the invention can identify noise patterns and predict peak periods, shifting from post-processing to pre-emptive prevention, addressing the technical bottleneck of traditional noise management, which relies on passive response. This invention establishes a closed-loop feedback system for noise control, utilizing an intelligent recommendation engine for control strategies and an A / B testing evaluation mechanism. This system continuously optimizes control strategies and achieves sustained improvement in dormitory acoustic environments, overcoming the technical limitations of traditional noise management, which lacks adaptive optimization capabilities. During noise monitoring and analysis, the invention only extracts acoustic feature parameters to determine noise type and location; the original sound content is not recorded or stored. This effectively resolves the technical contradiction between environmental monitoring and privacy protection by implementing regional noise control rather than individual identification.

[0022] In a specific embodiment, the process of executing step 100 may specifically include the following steps: The original noise signal is collected through the ECM high-sensitivity microphones placed in the geometric center of the dormitory space and the public area of ​​the corridor. The original noise signal is sampled in sections to obtain the original noise sampling data containing the dormitory number, floor and timestamp; Perform A-weighted filtering on the original noise sampling data, calculate the equivalent continuous A sound level and statistical parameters, and obtain the noise acoustic characteristic data; According to the short-term sound pressure mutation and spectrum characteristics of noise acoustic characteristic data, the non-continuous interference signal is identified and filtered through the interference recognition algorithm to obtain the effective noise signal; Decompose the signal frequency band based on the noise effective signal and remove the background noise to obtain noise-cleaned data; The noise purification data is evaluated for contextual relevance and data consistency, and transmitted through a GPRS wireless transmission module or a wired network, while a network disconnection protection mechanism is activated to obtain noise preprocessing data. Feature extraction and classification analysis are performed on the noise preprocessing data to obtain noise type recognition results and noise parameter data.

[0023] Specifically, a high-density, spatially representative raw noise collection system was constructed. An ECM high-sensitivity microphone was deployed at the geometric center of each dormitory, equidistant from the wall and moderately from the ceiling. Additional microphone nodes were evenly distributed in corresponding corridors or public passageways, resulting in a three-dimensional sampling structure with full spatial coverage and redundant fault tolerance. The ECM high-sensitivity microphones feature 360° omnidirectional pickup, low noise floor, wide dynamic response range, and high frequency resolution, ensuring high-fidelity sound pressure data acquisition within a signal range of 20 to 130 decibels. The sound signal was sampled continuously in real time at a sampling frequency of 48kHz. To ensure manageability and temporal traceability of the raw noise data, the continuous sound wave signal was segmented and sampled in fixed time windows, with a 10-second interval as the base unit. At the end of each time period, the sampled data was packaged and automatically bound to the corresponding spatial and temporal identifiers, including dormitory number, floor number, sensor number, and a timestamp accurate to the second, to generate structured raw noise sample data. After being cached locally at the collection node, the raw data enters the first stage of preprocessing. Once the raw data enters the processing pipeline, it undergoes A-weighted filtering. This involves applying an A-weighting function consistent with the human auditory sensitivity curve to the spectral signal. This frequency-weights the spectral signal, enhancing the contribution of the mid-frequency component (approximately 500Hz–6kHz) to the overall sound pressure value and weakening the influence of the low- and high-frequency components. This ensures that the processed result more closely reflects the actual noise disturbance perceived by the human body. Based on this, the equivalent continuous A-level (LAeq) is calculated for each sampling segment, and typical acoustic parameters of the sound pressure signal, including L10 (the level at which the sound pressure exceeds 10% of the time), L50, L90, Lmax (peak sound pressure), and Lmin (minimum sound pressure), are statistically analyzed to form the initial noise acoustic feature dataset. Dormitory environments contain a large number of occasional, non-continuous sound sources, such as door openings, dropped objects, and pedestrians. These signals, while high in sound pressure, do not constitute continuous interference. Therefore, an interference recognition algorithm is introduced to identify and remove these anomalous components based on the acoustic feature data. The specific method includes short-term energy mutation detection and spectrum offset analysis: the slope fitting analysis is performed on the sound pressure change rate of each sampling signal, and its spectrum stability index in the wavelet domain is calculated. If a high-amplitude but low-duration transient signal is detected, and the spectrum structure is highly matched with the atypical noise event template in the target classification model, the segment is automatically marked as "non-persistent interference", and the valid signal set is eliminated. The retained data constitutes the noise valid signal.Multi-scale signal frequency band decomposition is performed on the noise-signaled signal. Wavelet transforms (such as the Daubechies basis) are used to perform multi-level frequency domain segmentation, constructing subband signals from low-frequency layers to high-frequency layers. Energy thresholding and noise terrain modeling are then used to identify and eliminate background noise in low-energy disturbance areas, removing steady-state interference such as building noise and ventilation system noise, thereby restoring noise-cleaned data with a high signal-to-noise ratio. Contextual relevance analysis and data consistency assessment are performed on the noise-cleaned data. Contextual relevance compares the cosine similarity or Euclidean distance trends between feature vectors in adjacent time windows to determine whether the current data segment has abrupt discontinuities with its predecessors and successors. If there are no obvious discontinuities and the feature trends are smooth, the context is considered consistent. Data consistency is assessed by performing multi-source comparisons with signals from adjacent sensor nodes during the same time period to eliminate local anomalies caused by device errors or synchronization errors. After the data passes the consistency check, it is uploaded to the server using a GPRS wireless communication module or a wired network. If a network interruption occurs during transmission, the system activates a network-resuming protection mechanism, writing the locally cached temporary data to non-volatile storage and automatically retransmitting it after the connection is restored, ensuring uninterrupted data continuity and generating noise preprocessed data. The preprocessed data is input into the feature extraction and classification analysis module, which includes a time-domain feature extractor, a frequency-domain feature analyzer, and a time-frequency joint encoder. These extract features such as waveform factor, crest factor, energy distribution moment, spectral density, and frequency band centroid, and combine them into a noise feature vector. The noise feature vector is input into the trained noise recognition model, and a two-level structural model is adopted: the first layer performs coarse classification of the signal (such as human voice, equipment, environment, and activity) through improved K-means clustering. The second layer uses the support vector machine classifier to complete specific fine-category identification. For example, the human voice category is further divided into conversation, quarrel, laughter, etc., and the equipment category is identified as speakers, fans, computer sounds, etc., and each recognition result is assigned a corresponding confidence, duration, and waveform feature summary. Finally, the structured noise type recognition result and noise parameter data are output.

[0024] In a specific embodiment, the process of performing feature extraction and classification analysis on the noise preprocessing data to obtain the noise type recognition result and noise parameter data may specifically include the following steps: Based on the historical noise data of the dormitory, a day is divided into study period, rest period and sleep period, and a differentiated noise tolerance weight is assigned to each period to obtain the noise threshold model of the dormitory functional period; Perform a two-stage progressive constraint relaxation calculation on the noise preprocessed data to obtain preliminary noise event marker data; Based on the preliminary noise event marking data, fine-grained feature analysis is performed on noise events that exceed the current period threshold in the dormitory functional period noise threshold model to obtain key noise event feature data; Based on the key noise event feature data, dormitory-specific noise feature parameters such as conversation tone changes and electronic equipment audio patterns are set as discrete variables, and environmental background noise parameters are set as continuous variables to obtain mixed representation noise feature data; The mixed representation noise feature data is input into the dual-constraint allocation model, combined with the dormitory activity state transition probability parameters, and the noise type characteristics and the actual interference degree to the current functional period are calculated to obtain the context-related noise impact assessment results. Based on the analysis of dormitory activity patterns, the time interval for transitioning from study state to rest state or from rest state to sleep state is determined, and an activity switching protection interval is constructed. The noise sensitivity is dynamically increased within the activity switching protection interval to obtain the noise sensitivity adjustment parameter during the activity switching period. The context-dependent noise impact assessment results are corrected by applying the activity switching period noise sensitivity adjustment parameter to obtain the switching-aware noise interference assessment results. The noise interference evaluation results of switching perception are fused with the noise type feature data to form the noise type recognition results and noise parameter data.

[0025] Specifically, a time-zone model tailored to student life patterns was constructed. High-frequency monitoring data from a representative period was selected from a long-term, collected and stored dormitory noise database. Cluster analysis and trend fitting were used to extract the natural fluctuation structure of noise levels over a 24-hour period. Daily noise intensity data was plotted as a temporal distribution using an hourly sliding window. A local extreme value detection algorithm was applied to identify daily peaks and troughs. Based on on-campus schedules and actual behavioral observations, the day was divided into study, rest, and sleep periods, with tolerance weights set based on the noise sensitivity requirements of each period. The system assigned disturbance weights to different noise types (such as human voices, equipment noise, and ambient noise) and normalized the weights for each period to generate a three-dimensional parameter matrix. This model was then used to establish a noise threshold model for dormitory functional periods. This model not only accounts for temporal differences in tolerance but also considers the interplay between noise properties and intensity. Real-time or batch input noise preprocessed data is fed into a two-stage recognition model based on a progressive constraint relaxation strategy. In the first stage, a fixed threshold detection mechanism is used to rapidly screen for signal segments that clearly exceed the threshold. Samples exceeding the threshold are initially flagged by frame-by-frame comparison with the threshold model for the current time period. In the second stage, local statistical features are dynamically compared with historical noise curves, allowing for error-tolerant relaxation of some critical samples. A comprehensive evaluation of sample energy density fluctuation range, spectral stability, and consistency with the contextual time period is performed to eliminate false positives and address missed detections. The model ultimately outputs preliminary noise event marker data with timestamps, dormitory numbers, intensity levels, and suspected noise types. Based on this labeled data, a fine-grained feature analysis is performed on all noise events identified as exceeding the threshold. The spectral structure is decomposed using multi-layer Fourier transforms and wavelet packet decomposition techniques. Features such as peak factor, waveform symmetry, and pitch break points are extracted in the time domain. Furthermore, the samples are structurally compared with the prior templates used in the classification model to generate a set of key noise event feature data, including acoustic characteristics, possible source categories, disturbance forms, and duration. To unify the data representation of different features, attributes with strong discreteness, such as changes in conversational tone, music playback mode, speech rate and frequency band, and the rhythm of device on / off, are discretized and encoded using a multi-value classification method. At the same time, parameters with natural continuity, such as the mean background sound pressure, spectral energy distribution, and noise duration, are retained as numerical variables. This generates a hybrid representation of noise feature data. This data vector possesses structural diversity and can be read by both rule-based logic and statistical models. This hybrid representation of feature data is input into a dual-constraint allocation model for multi-dimensional mapping. This model combines the acoustic characteristics of the noise type with the temporal properties of the current dormitory functional period. By constructing an interference assessment function with a contextual weight term, it calculates the state interference value for each noise sample.The model also incorporates a dormitory activity state transition probability matrix constructed using a Markov chain. It infers the activity state of noise behavior sequences within the time window before and after the sample, calculating whether they are transitioning between study, rest, or sleep. This state transition information is then fed into the interference assessment as an auxiliary constraint, generating a context-sensitive noise impact assessment. Based on historical behavioral data and a dormitory work-life rhythm analysis model, the model clearly identifies the study-to-rest transition period and the rest-to-sleep transition period. A protection interval for activity switching, covering twice daily activity switching, is constructed. Within this interval, noise sensitivity is dynamically adjusted. For the same noise level, the tolerance threshold is lowered by 10–15%, and the perturbation weights for human voices and rhythmic sound sources are increased. This results in a set of dynamic adjustment parameters that serve as noise sensitivity adjustment parameters for activity switching periods. Based on this, the generated context-sensitive noise impact assessment results are amplitude-corrected using these dynamic sensitivity adjustment parameters, and the perturbation level boundaries are reassessed to obtain a switch-aware noise interference assessment result. This result exhibits temporal and behavioral transition adaptability, providing a more sensitive basis for identifying and responding to behavioral fluctuations during these boundary periods. The above-mentioned switching perception interference assessment results are fused with the original noise feature data, and the feature fusion matrix is ​​used to jointly model the dimensions such as time information, spatial attribution information, noise category, and interference level. The final noise type recognition results and noise parameter data are output through the integrated classification model.

[0026] In a specific embodiment, the process of executing step 200 may specifically include the following steps: Construct a noise propagation association network based on noise type identification results and noise parameter data; Based on the noise propagation correlation network, the noise autocorrelation coefficient, trend characteristics and periodicity index at different time scales are calculated to obtain the noise time evolution map; According to the mutual correlation coefficient, propagation delay and attenuation characteristics between different monitoring points in the noise propagation association network, the noise spatial propagation path diagram is calculated through graph theory modeling; The noise time evolution map and the noise spatial propagation path map are fused to obtain the noise spatiotemporal distribution characteristic map, and the noise spatiotemporal distribution characteristic map is subjected to an exceeding standard judgment process to obtain the noise exceeding standard judgment result.

[0027] Specifically, the noise type identification results and noise parameter data are structured and reorganized to extract core variables such as the location, time of occurrence, sound source type, spectral characteristics, energy intensity, and propagation direction of noise events within each monitoring period. These events are then assigned to nodes and bound to events based on the number of each sensor node and its corresponding physical geographic location information. Based on noise event data with timestamps and spatial labels, a temporally ordered and spatially traceable noise propagation association network is constructed. This network uses sensor nodes as vertices, and edges between nodes are established based on the existence of noise propagation paths. Edge weight coefficients are established through logical judgment of sound pressure synchronization, event type consistency, and propagation delay in the propagation path, forming a weighted directed graph with direction, intensity, and type, which preliminarily reflects the structural relationship of noise propagation within the building space. Based on the constructed noise propagation correlation network, its evolutionary behavior at different time scales is modeled. Local autocorrelation analysis is performed on the noise intensity time series at each sensor node, using minute, hour, and day time windows as the basic time windows. The autocorrelation coefficient is obtained by calculating the similarity of the sound pressure at each node in the current window with its own historical window. Trend fitting models (such as Holt-Winters exponential smoothing) are then used to analyze the directionality, amplitude, and reversal points of the sound pressure changes. Furthermore, frequency domain analysis techniques such as Fourier transform and wavelet decomposition are used to identify periodic indicators, extracting the dominant frequency and modulation rhythm of each noise type over different time periods. This results in a noise time evolution map that describes the trend direction, autocorrelation strength, and periodic behavior. Temporal clustering is performed by noise type, identifying typical daily activity rhythms and unique behavioral disturbance patterns. Furthermore, addressing the spatial propagation mechanism, the cross-correlation coefficient between each pair of monitoring nodes in the noise propagation correlation network is extracted to measure the consistency of their sound pressure signal changes within the same time window. The maximum value of the correlation function and its corresponding time lag are calculated using a sliding window to estimate the propagation delay. On this basis, the free-field propagation attenuation coefficient, reflection interference correction term, and material sound insulation parameters from the acoustic propagation model are introduced. Combined with the building's structural drawings and the distribution of possible sound propagation paths, the sound pressure attenuation characteristics of noise propagating from one node to another are calculated. By encoding these time delays and attenuation weights into the edge weights of the associated network and modeling the graph structure using graph theory, an optimal propagation path tree is constructed using a shortest propagation path algorithm (such as Dijkstra) combined with the minimum energy loss criterion. Ultimately, a noise spatial propagation path map is generated. This path map reflects the actual path of noise diffusion from the source along the spatial structure, and annotates the time delay, intensity attenuation, and event type transition of the received signal at each node along the path.To unify temporal and spatial information, the aforementioned noise time evolution map is spatiotemporally fused with the spatial propagation path map. A tensor fusion method is used to construct a three-dimensional tensor structure, with the first dimension representing node location, the second dimension representing the time window, and the third dimension representing the noise intensity and type vector. Convolutional aggregation is performed on each node's historical time window tensor slices to extract local spatiotemporal propagation features. A graph convolutional neural network is used to embed this tensor map, projecting the node's spatiotemporal features into a low-dimensional embedding space. Key propagation regions within the spatial structure, characterized by dense propagation paths, energy focus, or periodic resonances, are identified to generate a noise spatiotemporal distribution feature map. The sound pressure intensity values ​​for each node in the noise spatiotemporal distribution feature map within each time window are compared with the noise threshold model established for the functional period. A disturbance index is calculated based on the noise type, duration, and the number of upstream and downstream nodes influencing the propagation path. If the sound pressure intensity at a node consistently exceeds the tolerance threshold corresponding to the current functional period and the node is located at the starting point or turning point of energy transmission in the propagation path, the system marks the node as a "spatiotemporal exceedance event." To prevent misjudgments, a disturbance threshold determination mechanism is introduced. This sets a minimum duration threshold for isolated incidents and requires that their propagation path impact at least one adjacent node before they are considered valid exceedance events. This approach ultimately generates a noise exceedance determination result.

[0028] In a specific embodiment, the process of executing step 300 may specifically include the following steps: Calculate the noise energy distribution differences of each floor and each area of ​​the building based on the noise spatiotemporal distribution characteristic map and determine the first noise source area data through the noise energy gradient change; Based on the first noise source area data and the noise exceeding standard determination result, the spatial coordinates of the noise source are calculated by triangulation using the sound pressure level data, propagation delay difference, and building acoustic characteristic parameters collected by at least three adjacent sensors to obtain the second noise source area data; Based on the second noise source area data, an acoustic array consisting of a main sensor in a single dormitory and auxiliary sensors in the outdoor corridor is used, combined with sound source separation technology to correct the noise propagation path and obtain the third noise source area data; Performing pattern matching on the acoustic features collected at the third noise source area data, matching and analyzing the noise signal with a preset acoustic feature library, determining the noise type and its main acoustic characteristics, and obtaining noise type attribute data; The noise type attribute data is associated with the noise type identification results, the third noise source area data, the degree of exceeding the standard and the impact range data, and the time information and regional affiliation information are added to obtain the dormitory noise source tracing data.

[0029] Specifically, based on the spatial and temporal noise distribution characteristic map, the noise energy distribution state within each floor and area of ​​the building is systematically calculated and compared. The sum and average sound pressure levels of each sensor node in different time windows are extracted from the map tensor. A noise energy spatial distribution matrix is ​​constructed, using floor numbers and area numbers (e.g., dormitory, corridor, stairwell, etc.) as basic units. By comparing the differences in sound energy values ​​across different units within the same time period, the rate of change of energy gradients along spatial axes (e.g., longitudinal floor, height) and local plane axes (e.g., east-west corridors, adjacent dormitories) is calculated. The direction of maximum gradient is used as the main propagation vector. Spatial inversion is performed based on the starting node location of the noise exceedance event to determine the origin point with the highest concentration of sound energy along the propagation path, which is the first noise source area data. This data, based on floor number, area identifier, and node coordinates, clearly identifies the high-energy disturbance zone most likely to be the noise source within the current time window and serves as the spatial constraint for subsequent fine-tuning operations. Based on the data of the first noise source area, data from at least three physically independent and time-synchronized sensor nodes surrounding the area are extracted. Raw sound pressure level curve data and signal response times within the corresponding time window are then obtained. To perform spatial calculations using triangulation, the spatial coordinate relationships between the three nodes are combined with the arrival time differences of the sound pressure peaks. A system of time-delay difference equations is constructed using the propagation delay differences. To ensure the accuracy of the propagation model, building acoustic parameters are incorporated, including the sound insulation coefficients of walls, doors, and windows, the spatial absorption coefficient, and the volume reverberation time. These physical properties are then corrected for propagation velocity and path length parameters to construct a modified sound wave propagation model for actual building conditions. Based on this propagation model, the spatial coordinates of the noise event source are determined by solving a nonlinear system of equations for the arrival time of the sound wave front within the spatial triangular region formed by the three nodes using interpolation positioning and error-constrained minimization iterations. This outputs data for the second noise source area, which is accurate to the specific dormitory unit or even the local space within the dormitory, along with a positioning error radius and spatial confidence interval. However, due to complex architectural acoustic effects such as corridor reflections, room reverberation, and wall diffraction, sound source estimation based solely on triangulation still has certain limitations. Therefore, an acoustic array correction mechanism was introduced. Within the located second noise source area, the main sensor in the dormitory and the auxiliary sensor in the corridor outside the door were selected to construct a minimum closed array group. This micro-array formed a near-field sound source reception model. The noise signals received by the array in the same time period were input into the sound source separation module. The dominant sound source signal path was extracted through independent component analysis and beamforming algorithms. The phase difference and amplitude distribution of the signal were then inverted to identify possible reflection components and interference directions.On this basis, a spatial filter is used to further suppress interference waves from directions other than the source, thereby correcting the propagation path of the previous localization result and reproducing the corrected third noise source area data. This area not only provides high-confidence localization capabilities but also clearly indicates the precise location of the noise event within the room, such as near a door, desk, window, and other specific locations. After obtaining the precise localization area, a purified high-signal-to-noise ratio audio signal segment is extracted from this spatial node. This audio signal is then feature extracted in the time, frequency, and time-frequency domains, calculating key eigenvalues ​​such as the main frequency band, transient energy change, amplitude modulation rate, waveform period, and energy spectral density. An acoustic feature vector is also constructed. This vector is then input into a pre-set acoustic feature template library for pattern matching analysis. The feature library contains pre-set standard sample vectors for common dormitory sound source types, such as conversation, quarreling, laughter, equipment operation, music playing, and impact. Each template contains a multi-dimensional acoustic descriptor and a variation range. Using methods such as cosine similarity, Mahalanobis distance, and dynamic time warping, the system calculates the similarity between the current sound source signal and each category template in the feature library. Based on the matching results, the noise event's type, source feature category (e.g., high-frequency vibration, periodic rhythm, broadband noise), confidence level, and representative feature point values ​​are determined, generating noise type attribute data. This noise type attribute data is then compared with the noise type identification results previously output by the recognition model. A comprehensive label is then constructed using data from the third noise source region, the level of exceedance marked in the event's spatiotemporal distribution map, the number of nodes within the affected area, the sound energy level, and the duration of the interference. This label is then combined to form a complete dormitory noise source traceability data record. The system also appends metadata such as time information (year / month / day / hour / minute / second), event ID, floor number, dormitory number, and monitoring device number to this record, making it searchable, traceable, and verifiable.

[0030] In a specific embodiment, the process of executing step 400 may specifically include the following steps: Conduct time dimension analysis on dormitory noise source tracing data to obtain daily, weekly, and semester noise variation patterns. Perform spatial dimension analysis on dormitory noise source tracing data to obtain dormitory noise spatial distribution characteristic data; Based on the daily, weekly, and semester variation patterns of dormitory noise and the spatial distribution characteristics of dormitory noise, the correlation between different types of noise is analyzed to obtain a sequence of dormitory noise behavior patterns. The dormitory noise behavior pattern sequence is input into the ensemble learning prediction model to obtain dormitory noise warning information; A graded warning strategy is constructed based on the probability of noise exceeding the standard and the potential impact degree in the dormitory noise warning information. Different levels of warning trigger thresholds and intervention measures are set for low-risk, medium-risk and high-risk noise events to obtain forecast data for high-incidence periods.

[0031] Specifically, the accumulated historical noise source tracing data is analyzed temporally, extracting key elements such as the time of occurrence, event intensity, type, and duration from all traceability records with complete timestamps. A framework for analyzing the diurnal variation of dormitory noise is constructed using the day as the basic unit. By dividing the 24-hour day into several fixed time periods (e.g., one per hour), the frequency, average sound pressure, and duration of different types of noise events within each time period are statistically analyzed to form a multidimensional vector time series. Typical daytime noise rhythms are then identified through time series curve fitting and local extreme value detection. The diurnal noise variation data are aggregated and analyzed by natural week, and the distribution of events in the same time period each day is compared within the week. Time series similarity functions (e.g., DTW) and periodic regression analysis are used to extract periodically repetitive structural variation characteristics, forming a weekly variation pattern for dormitory noise. By extending the time span to the entire semester, a sliding average is applied to weekly data, incorporating labels for teaching schedules such as holidays and exam weeks. A semester-scale trend analysis model is constructed. This model outputs semester-long patterns of dormitory noise variation from various perspectives, including overall trends, periodic mutation points, and periodic peaks and troughs, forming a comprehensive temporal behavioral portrait encompassing both short-term and long-term patterns. Furthermore, spatial analysis is performed on dormitory noise source tracing data. Each monitoring unit, such as dormitory, corridor, and public area, is treated as a spatial node in a spatial structure diagram, and a two- or three-dimensional spatial topology map is constructed based on the sensor layout coordinates. The system normalizes the total number of noise events, mean intensity, frequency of exceeding standards, and disturbance type structure for each node within a fixed period to form a spatial intensity matrix. Hotspot analysis and spatial clustering techniques (such as DBSCAN) are then used to identify high-frequency event areas and anomalous propagation paths, such as dormitories with frequent sharp sound sources, clusters of nodes where noise spreads to upper and lower floors, or corridor echo zones where signal reflections are concentrated. This ultimately generates spatial distribution characteristics of dormitory noise, indicating the average noise level of each area and encoding its criticality and impact radius within the propagation network. By integrating temporal and spatial behavioral basic graphs, a causal chain model of noise behavior was established. Using noise type as an index, event sequences with continuous temporal correlation and spatial adjacency were extracted as behavioral segments. Sequential pattern mining algorithms (such as PrefixSpan) and association rule analysis (such as Apriori) were then used to mine typical noise behavior pattern sequences from the data. Each pattern was assigned labels such as frequency of occurrence, average duration, coverage area, and typical time period. This library of dormitory noise behavior pattern sequences served as a high-level feature representation describing the evolutionary mechanisms of noise events. The noise behavior pattern sequences were then fed into an ensemble learning prediction model for trend forecasting. This model combines a long short-term memory neural network (LSTM), XGBoost, and Prophet models, each of which captures nonlinear long-term dependencies, short-term mutation points, and cyclical trends in time series.The LSTM model uses the temporal feature vectors of each behavior sequence as input to learn the temporal evolution probability of different behavior chains. XGBoost establishes a nonlinear predictive mapping between noise propagation relationships between spatial nodes and external factors (such as holidays and weather). The Prophet module models the periodic structure of behavioral patterns based on fixed temporal rhythms. The integrated model, by weightedly fusing the outputs of the three sub-models, ultimately predicts the probability, likely time period, spatial location, and disturbance level of various noise events within several future time windows, generating structured dormitory noise warning information. The output includes key indicators such as warning type, prediction confidence, target area, start time, and duration. After obtaining the prediction results, the system enters the hierarchical response strategy formulation phase. The event risk value is calculated based on the noise exceedance probability and potential impact level in the warning information. A noise disturbance risk scoring model is constructed by combining the historical disturbance impact range, noise type nuisance level, and dormitory density coefficient. Based on this model, future noise events are classified into three levels: low risk (e.g., gentle talking, background music), medium risk (e.g., frequent movement, dragging heavy objects), and high risk (e.g., large gatherings, quarrels, etc.). Differentiated warning trigger thresholds and response strategies are configured for different levels: for low-risk events, only reminders are sent via the dormitory head app or a silent reminder is added to the interface. For medium-risk events, the system activates audio warning devices and records relevant logs for subsequent tracing. For high-risk events, remote announcements, automatic warning push notifications from dormitory managers, or temporary adjustments to the sampling frequency and sensitivity of sensors in the area for focused monitoring. Relevant data is archived and backed up for subsequent responsibility identification and intervention assessment. All policy responses are bound to the predicted timeline and spatial nodes, forming the output content of the high-incidence period forecast data. Its data structure includes key fields such as the predicted time period, warning level, recommended strategy, impact range, and node weight.

[0032] In a specific embodiment, the process of executing step 500 may specifically include the following steps: Input dormitory noise warning information and high-incidence period forecast data into the intelligent recommendation engine for control strategies, comprehensively analyze noise level distribution, noise type, regional characteristics and forecast trends, and obtain multi-level noise control strategy recommendation data; The multi-level noise control strategy suggestion data is processed in layers to obtain a graded noise control implementation plan; Deploy the graded noise control implementation plan to the dormitory noise control system through PC or mobile phone. Activate environmental reminders, intelligent linkage, or early warning push functions based on the degree of noise exceeding the standard, and obtain dormitory noise intervention records. Based on dormitory noise intervention records, an evaluation index system was established for noise compliance rate, reduction rate of noise exceeding standard events, environmental satisfaction, and management efficiency. An AB testing algorithm was used to compare and analyze the effects of different control strategies to obtain noise control strategy evaluation results. Combining the noise control strategy evaluation results with the current dormitory environment, a correlation analysis between dormitory noise and dormitory environmental quality was performed to obtain a dormitory noise environment optimization plan.

[0033] Specifically, dormitory noise warning information and high-incidence period forecast data are fed into an intelligent noise control strategy recommendation engine. Based on a multi-factor fusion analysis mechanism, this engine jointly models multiple variables, including noise level distribution data, noise type identification labels, spatial area attribute characteristics, historical noise behavior trajectories, and disturbance trends within the predicted time window. A graph-based enhanced recommendation strategy is then used for strategy matching. A multidimensional feature space vector is constructed to quantify the noise risk intensity of different dormitory areas within a specific timeframe. This vector is then aligned with successful cases in a historical strategy response database. Vector similarity and risk scoring functions are then used to filter the vectors. The engine then outputs a set of multi-level noise control strategy recommendations that are most adaptable to the current forecast scenario. These recommendations cover spatial levels (e.g., building, floor, and dormitory), temporal levels (e.g., early intervention during high-incidence periods), and behavioral levels (e.g., strategies to suppress target behavior types), ensuring that the recommended strategies are fully context-aware and adaptable. The recommendation data is logically structured and filtered through multi-dimensional conditions. Using a control task decomposition model, each recommendation is broken down into executable hierarchical control units based on management priority, strategy complexity, and execution conditions. Basic intervention strategies (such as environmental cues and silent reminders) are defined as a Level 1 response, moderate linkage strategies (such as audible and visual warnings and administrator reminders) as a Level 2 response, and high-intensity strategies (such as remote announcements, data storage, and patrol dispatch) as a Level 3 response. Multi-level switching rules are set based on factors such as the degree of disturbance, spatial location, and time urgency, forming a structured, hierarchical noise control implementation plan. Once generated, this implementation plan is automatically stored as a sequence of tasks to be deployed and can be remotely activated and deployed through the PC management backend or mobile app management platform. During deployment, the hierarchical implementation plan is pushed as a task package to the front-end control unit through command interaction with the established dormitory noise control platform. The front-end includes sensor modules, environmental cues modules (such as voice announcers and display screens), intelligent linkage modules (such as access control, lighting, and audio), and user communication modules (such as the student app and administrator WeChat server). Based on the degree of exceedance detected in actual scenarios, the predicted risk level during the forecasted period, and behavioral pattern characteristics, the system activates corresponding intervention measures in real time: if the risk is low, the system only implements environmental prompts or pop-up reminders; if the risk is medium, it simultaneously activates sound prompts, audio and visual alarms, and pushes event briefings to management personnel; and in high-risk situations, the system initiates lockdown strategies, such as locking specific sound source devices, starting audio recording for evidence, or triggering remote voice commands. All of these execution processes are recorded in real time and form a complete dormitory noise intervention record, which includes the type of intervention measure, execution time, target area, triggering reason, execution status, and subsequent impact tracking.To quantitatively evaluate the system's effectiveness, a noise control performance evaluation module was built into the backend. Based on intervention records, it automatically aggregates and constructs a four-dimensional evaluation index system, including "noise compliance rate," "reduction rate of exceedance incidents," "environmental satisfaction," and "management efficiency." The compliance rate measures the percentage of time that all dormitories meet the "Civil Building Sound Insulation Design Code" and the school's internal control standards. The exceedance reduction rate compares the trend in the number of exceedance incidents during the current cycle with the historical average. Environmental satisfaction is quantified using student survey data and the ratio of positive feedback from management logs. Management efficiency is modeled by combining response time, the number of incident handling processes, and the speed at which policies take effect. To verify the effectiveness of different policy configurations, an A / B testing algorithm was introduced. In actual deployment, a subset of policies were applied to a randomly selected sample of dormitories as a test group, while the control group maintained the existing policy. The system continuously analyzes the differences in noise change curves, student feedback, intervention frequency, and other multi-dimensional data after the two policy groups were implemented. Hypothesis testing was used to confirm the effectiveness, ultimately producing quantitative and statistically significant noise control strategy evaluation results. By combining the strategy evaluation results with the current dormitory environment, regression modeling is performed on environmental factors including spatial structure, occupancy density, dormitory functional zoning, student grade composition, and behavioral risk distribution, exploring potential causal relationships between noise disturbance and environmental structure. The system utilizes multivariate regression analysis and structural equation modeling to assess the effectiveness of strategies under different environmental combinations and further analyze which strategies are more sensitive to high-density areas and which are more effective for younger students. Based on this, the system automatically generates a dormitory noise environment optimization plan. This plan includes recommendations for management system optimization (such as adjusting the frequency of on-duty patrols and establishing a noise responsibility points system), physical space structure modifications (such as adding sound absorption devices, optimizing bed distribution, and improving vent sound barriers), and intelligent system linkage configuration adjustments (such as adding specific floors to key monitoring areas and installing redundant sensing devices at key nodes). All optimization recommendations are packaged into standardized output solutions and fed back into the intelligent recommendation engine through a policy feedback mechanism, forming a closed-loop strategy iteration loop, enabling the system to continuously learn and adapt to new behavioral patterns and environmental changes over the long term.

[0034] The above describes the dormitory noise monitoring data analysis method in the embodiment of the present application. The following describes the dormitory noise monitoring data analysis system in the embodiment of the present application. Figure 2 In one embodiment of the present application, a dormitory noise monitoring data analysis system includes: The feature extraction module 11 is used to extract noise features from the original noise signal collected in the dormitory space to obtain noise type recognition results and noise parameter data; An analysis module 12 is used to analyze the temporal and spatial distribution characteristics of noise based on the noise type identification result and noise parameter data, and obtain a temporal and spatial distribution characteristic map of noise and a noise exceeding standard determination result; Processing module 13, for determining the noise source area and noise type attributes based on the noise spatiotemporal distribution characteristic map and the noise exceeding standard determination result, and obtaining dormitory noise source tracing data; The prediction module 14 is used to perform pattern analysis and trend prediction on dormitory noise source tracing data to obtain dormitory noise warning information and high-incidence period prediction data; The creation module 15 is used to create a dormitory noise environment optimization plan based on the dormitory noise warning information and the high-incidence period prediction data.

[0035] Through the collaborative cooperation of the above-mentioned components, the present invention is different from the traditional fixed threshold noise monitoring method. It dynamically adjusts the noise evaluation standard according to the characteristics of different functional periods in the dormitory (study, rest, sleep), realizes the accurate identification and evaluation of dormitory noise, and effectively solves the technical problem of insufficient adaptability of traditional methods in specific scenarios. Through the two-stage computational framework of progressive constraint relaxation, the present invention decomposes the noise analysis task into two stages: coarse-grained noise event detection and fine-grained feature analysis. It uses a hybrid representation to process different types of noise parameters, significantly reducing the computational complexity of the system and solving the technical problem of real-time processing of large-scale noise data. By constructing a time period activity perception matrix and an activity switching protection interval, the present invention associates the noise pattern with the dormitory activity cycle, introduces the dormitory activity state transition probability parameter, and realizes the context-aware evaluation of noise interference, overcoming the limitation of traditional methods that ignore environmental context. Based on a multi-level noise source localization method, combined with sound intensity gradient analysis and triangulation techniques, this invention accurately identifies noise source areas, overcoming the limited localization accuracy of traditional methods in complex indoor environments. By analyzing the temporal and spatial dimensions of noise source tracing data and integrating it with an ensemble learning prediction model, the invention can identify noise patterns and predict peak periods, shifting from post-processing to pre-emptive prevention, addressing the technical bottleneck of traditional noise management, which relies on passive response. This invention establishes a closed-loop feedback system for noise control, utilizing an intelligent recommendation engine for control strategies and an A / B testing evaluation mechanism. This system continuously optimizes control strategies and achieves sustained improvement in dormitory acoustic environments, overcoming the technical limitations of traditional noise management, which lacks adaptive optimization capabilities. During noise monitoring and analysis, the invention only extracts acoustic feature parameters to determine noise type and location; the original sound content is not recorded or stored. This effectively resolves the technical contradiction between environmental monitoring and privacy protection by implementing regional noise control rather than individual identification.

[0036] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0037] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling an electronic device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, and other media that can store program code.

[0038] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A dormitory noise monitoring data analysis method, characterized in that: include: Noise feature extraction is performed on the original noise signal collected in the dormitory space to obtain noise type identification results and noise parameter data; Analyzing the distribution characteristics of noise in time and space based on the noise type identification result and the noise parameter data to obtain a noise spatiotemporal distribution characteristic map and a noise exceeding standard determination result; Determine the noise source area and noise type attributes based on the noise spatiotemporal distribution characteristic map and the noise exceeding standard determination result, and obtain dormitory noise source tracing data; Performing pattern analysis and trend prediction on the dormitory noise source tracing data to obtain dormitory noise warning information and high-incidence period prediction data; A dormitory noise environment optimization plan is created based on the dormitory noise warning information and high-incidence period prediction data.

2. The dormitory noise monitoring data analysis method according to claim 1, characterized in that: The noise feature extraction is performed on the original noise signal collected in the dormitory space to obtain the noise type identification result and noise parameter data, including: The original noise signal is collected by ECM high-sensitivity microphones placed in the geometric center of the dormitory space and the public area of ​​the corridor, and the original noise signal is sampled in sections to obtain the original noise sampling data containing the dormitory number, floor and timestamp; Performing A-weighted filtering on the original noise sampling data, and calculating the equivalent continuous A sound level and statistical parameters to obtain noise acoustic characteristic data; According to the short-term sound pressure mutation and spectrum characteristics of the noise acoustic characteristic data, the non-continuous interference signal is identified and filtered by the interference identification algorithm to obtain the noise effective signal; Decomposing the signal frequency band based on the noise effective signal and removing the background noise to obtain noise-cleaned data; The noise purification data is evaluated for context relevance and data consistency, and transmitted via a GPRS wireless transmission module or a wired network, while a network disconnection and resuming protection mechanism is activated to obtain noise preprocessed data; Feature extraction and classification analysis are performed on the noise preprocessing data to obtain noise type recognition results and noise parameter data.

3. The dormitory noise monitoring data analysis method according to claim 2, characterized in that: The feature extraction and classification analysis of the noise preprocessing data to obtain noise type recognition results and noise parameter data includes: Based on the historical noise data of the dormitory, a day is divided into study period, rest period and sleep period, and a differentiated noise tolerance weight is assigned to each period to obtain the noise threshold model of the dormitory functional period; Performing a two-stage progressive constraint relaxation calculation on the noise preprocessed data to obtain preliminary noise event marker data; Based on the preliminary noise event marking data, a fine-grained feature analysis is performed on the noise events that exceed the current period threshold in the dormitory functional period noise threshold model to obtain key noise event feature data; Based on the key noise event feature data, dormitory-specific noise feature parameters such as conversation tone changes and electronic device audio patterns are set as discrete representations, and ambient background noise parameters are set as continuous variables to obtain mixed representation noise feature data; The mixed representation noise feature data is input into a dual-constraint allocation model, combined with the dormitory activity state transition probability parameter, and the noise type characteristics and the actual interference degree to the current functional period are calculated to obtain a context-related noise impact assessment result; Based on the analysis of dormitory activity patterns, the time interval for transitioning from study state to rest state or from rest state to sleep state is determined, and an activity switching protection interval is constructed. The noise sensitivity is dynamically increased within the activity switching protection interval to obtain the noise sensitivity adjustment parameter during the activity switching period. Applying the activity switching period noise sensitivity adjustment parameter to the context-related noise impact assessment result to correct it, so as to obtain a switching-perceived noise interference assessment result; The noise interference evaluation result of the handover perception is fused with the noise type feature data to form a noise type recognition result and noise parameter data.

4. The dormitory noise monitoring data analysis method according to claim 1, characterized in that: The analyzing the distribution characteristics of noise in time and space according to the noise type identification result and the noise parameter data to obtain a noise spatiotemporal distribution characteristic map and a noise exceeding standard determination result includes: Constructing a noise propagation association network based on the noise type identification result and the noise parameter data; Based on the noise propagation correlation network, the noise autocorrelation coefficient, trend characteristics and periodicity index at different time scales are calculated to obtain a noise time evolution map; Calculating a noise spatial propagation path diagram through graph theory modeling based on the mutual correlation coefficients, propagation delays, and attenuation characteristics between different monitoring points in the noise propagation association network; The noise time evolution map and the noise spatial propagation path map are fused to obtain a noise spatiotemporal distribution characteristic map, and the noise spatiotemporal distribution characteristic map is subjected to an exceeding standard determination process to obtain a noise exceeding standard determination result.

5. The dormitory noise monitoring data analysis method according to claim 1, characterized in that: The step of determining the noise source area and noise type attributes based on the noise spatiotemporal distribution characteristic map and the noise exceeding standard determination result to obtain dormitory noise source tracing data includes: Calculate the noise energy distribution difference of each floor and each area of ​​the building according to the noise spatiotemporal distribution characteristic map and determine the first noise source area data by the noise energy gradient change; Based on the first noise source area data and the noise exceeding standard determination result, the spatial coordinates of the noise source are calculated by triangulation using sound pressure level data, propagation delay difference, and building acoustic characteristic parameters collected by at least three adjacent sensors to obtain second noise source area data; Based on the second noise source area data, an acoustic array consisting of a main sensor in a single dormitory and auxiliary sensors in an outdoor corridor is used in combination with sound source separation technology to correct the noise propagation path to obtain third noise source area data; Performing pattern matching on the acoustic features collected at the third noise source area data, matching and analyzing the noise signal with a preset acoustic feature library, determining the noise type and its main acoustic characteristics, and obtaining noise type attribute data; The noise type attribute data is associated with the noise type identification result, the third noise source area data, the exceeding degree and the impact range data, and the time information and regional affiliation information are added to obtain the dormitory noise source tracing data.

6. The dormitory noise monitoring data analysis method according to claim 1, characterized in that: The pattern analysis and trend prediction of the dormitory noise source tracing data are performed to obtain dormitory noise warning information and high-incidence period prediction data, including: Performing a time dimension analysis on the dormitory noise source tracing data to obtain daily, weekly, and semester noise variation pattern data; Performing spatial dimension analysis on the dormitory noise source tracing data to obtain dormitory noise spatial distribution characteristic data; Based on the daily, weekly, and semester variation patterns of dormitory noise and the spatial distribution characteristic data of dormitory noise, the correlation between different types of noise is analyzed to obtain a dormitory noise behavior pattern sequence; Inputting the dormitory noise behavior pattern sequence into an integrated learning prediction model for prediction to obtain dormitory noise warning information; A graded warning strategy is constructed based on the probability of noise exceeding the standard and the potential impact degree in the dormitory noise warning information. Different levels of warning trigger thresholds and intervention measures are set for low-risk, medium-risk and high-risk noise events to obtain prediction data for high-incidence periods.

7. The dormitory noise monitoring data analysis method according to claim 1, characterized in that: The step of creating a dormitory noise environment optimization plan based on the dormitory noise warning information and the high-incidence period prediction data includes: Input the dormitory noise warning information and the high-incidence period forecast data into the intelligent recommendation engine for noise control strategies, comprehensively analyze the noise level distribution, noise type, regional characteristics and forecast trends, and obtain multi-level noise control strategy recommendation data; Performing hierarchical processing on the multi-level noise control strategy suggestion data to obtain a hierarchical noise control implementation plan; Deploy the graded noise control implementation plan to the dormitory noise control system through a PC or mobile phone, activate environmental prompts, intelligent linkage or early warning push functions according to the degree of noise exceeding the standard, and obtain dormitory noise intervention records; Based on the dormitory noise intervention records, an evaluation index system for noise compliance rate, reduction rate of exceeding standard events, environmental satisfaction and management efficiency was established, and an AB test algorithm was used to compare and analyze the effects of different control strategies to obtain the noise control strategy evaluation results; Combined with the noise control strategy evaluation results and the current status of the dormitory environment, a correlation analysis between dormitory noise and dormitory environment quality is performed to obtain a dormitory noise environment optimization plan.

8. A dormitory noise monitoring data analysis system, characterized in that: Used to execute the dormitory noise monitoring data analysis method according to any one of claims 1 to 7, the dormitory noise monitoring data analysis system comprises: The feature extraction module is used to extract noise features from the original noise signals collected in the dormitory space to obtain noise type identification results and noise parameter data; An analysis module is used to analyze the distribution characteristics of noise in time and space according to the noise type identification result and the noise parameter data, and obtain a noise spatiotemporal distribution characteristic map and a noise exceeding standard determination result; A processing module is used to determine the noise source area and noise type attributes based on the noise spatiotemporal distribution characteristic map and the noise exceeding standard determination result, and obtain dormitory noise source tracing data; A prediction module is used to perform pattern analysis and trend prediction on the dormitory noise source tracing data to obtain dormitory noise warning information and high-incidence period prediction data; A creation module is used to create a dormitory noise environment optimization plan based on the dormitory noise warning information and high-incidence period prediction data.

Citation Information

Cited By

  • Deep data set construction method, safety monitoring method and related equipment

    CN122154242A

  • Sound barrier risk identification and prediction method, system and computer

    CN122192748A

  • A sound barrier risk identification prediction method, system and computer

    CN122192748B