Intelligent sound box synchronous coordination monitoring system and method based on artificial intelligence

Abstract

The invention discloses an intelligent sound box synchronous coordination monitoring system and method based on artificial intelligence, and belongs to the technical field of intelligent sound boxes, audio signals, environmental parameters, user behavior data and network states are synchronously acquired through a plurality of intelligent sound boxes distributed at different positions, and the network states are subjected to timestamp calibration and standardization preprocessing to obtain a plurality of intelligent sound boxes; a multi-dimensional spatio-temporal data set is constructed, compressed and stored; inputting the preprocessed data into a graph neural network to construct a dynamic association graph, analyzing association among the data, recognizing a current use scene in combination with environment characteristics and a user behavior mode, evaluating a multi-loudspeaker-box cooperation state and a user behavior risk level, and generating a comprehensive evaluation result; based on the evaluation result, a collaborative optimization strategy of the multiple intelligent sound boxes is generated, and an audio transmission path and resource allocation are dynamically adjusted; and constructing a space-time prediction model, performing probability prediction on a future sound box cooperation state and a user behavior trend, and when the prediction probability exceeds a preset threshold value, triggering a grading early warning mechanism and executing a corresponding early warning strategy.

Description

Technical Field

[0001] The present invention relates to the technical field of smart speakers, and in particular to an artificial intelligence-based smart speaker synchronization and coordination monitoring system and method. Background Art

[0002] With the widespread adoption of smart home devices, smart speakers have become an essential component in multiple scenarios, including home entertainment, voice assistants, and information management. Smart speakers not only provide convenient voice interaction services, but also enhance the user experience through voice recognition and intelligent features. With the increasing popularity of multiple smart speakers, the need to play audio simultaneously across multiple devices in the same environment has become a common requirement. However, due to factors such as audio signal transmission delays and signal interference, precise synchronization between different speakers is often not possible when multiple speakers are playing together. Users often notice differences in sound quality and rhythm when enjoying music or movies. This is especially true in scenarios such as home theaters, where audio asynchrony can significantly impact the overall listening experience.

[0003] In addition, current smart speakers mostly rely on fixed algorithm models to analyze user behavior or voice commands, but these models often lack sufficient flexibility and find it difficult to accurately judge complex and changeable user behavior. Especially when smart speakers are used for security monitoring or processing abnormal user behavior, these algorithm models often fail to respond in a timely manner or make misjudgments, resulting in poor monitoring effects. This not only affects the user's daily usage experience, but in some application scenarios, especially in security monitoring, it may also bring greater potential risks. Summary of the Invention

[0004] The purpose of the present invention is to provide an intelligent speaker synchronization coordination monitoring system and method based on artificial intelligence to solve the problems raised in the above background technology.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions: a method for synchronous coordination monitoring of smart speakers based on artificial intelligence, comprising: Step S100: Synchronously collect audio signals, environmental parameters, user behavior data, and network status through multiple smart speakers distributed in different locations; perform timestamp calibration and standardization preprocessing on the collected multi-source data, construct a multidimensional spatiotemporal dataset containing spatial location information, and perform preliminary feature extraction and data compression storage; Step S200: Input the preprocessed data into the graph neural network to construct a dynamic association graph, and analyze the associations between the data through graph convolution; combine environmental characteristics and user behavior patterns to automatically identify the current usage scenario, simultaneously evaluate the multi-speaker collaboration status and user behavior risk level, and generate a comprehensive evaluation result that includes scenario type, collaboration status, and risk level; Step S300: Based on the scenario evaluation results, a collaborative optimization strategy for multiple smart speakers is generated, the audio transmission path is dynamically adjusted according to the current network status, and the computing resources and playback tasks of each speaker are intelligently allocated; Step S400: Based on historical collaboration data and real-time user behavior patterns, a spatiotemporal prediction model is constructed to make probabilistic predictions on the speaker collaboration status and user behavior trends in future time periods. When the predicted probability exceeds a preset threshold, a graded warning mechanism is automatically triggered and the corresponding warning strategy is executed.

[0006] Furthermore, step S100 includes: Step S101: Assign a unique three-dimensional position identifier P to each smart speaker i =(x i ,y i ,z i ), where (x i ,y i ,z i ) represents the coordinate values ​​of smart speaker i on the x-axis, y-axis and z-axis, determined based on the indoor Cartesian coordinate system, with the origin of the coordinates set to the position of the master device, and is assigned a logical identifier ID i Forming a device identification pair (ID i ,P i ), where ID i A unique digital identifier representing smart speaker i. Each smart speaker uses a microphone array to collect ambient background sound, user voice, and playback audio signals, and employs a beamforming algorithm to suppress ambient noise. Temperature, humidity, and air pressure are collected synchronously using temperature and humidity sensors and air pressure sensors. Accelerometers and infrared sensors are used to capture user motion amplitude and movement trajectory at a preset frequency. A network interface acquires network status parameters in real time at set intervals, and all data is generated with an initial timestamp. Step S102: Use the PTP protocol to globally calibrate the timestamps of each speaker and synchronize the initial time to a unified timestamp; construct a data association matrix M data , audio, environment, behavior and network data in the same time window are sorted by device ID (ID i ,P i) for association and integration; a two-layer data storage architecture is constructed: the first storage unit uses column storage to save audio signals, environmental parameters and network status; the second storage unit uses a graph database to store user behavior sequence data, with nodes recording behavior status and edges representing state transition relationships; at the same time, device identification pairs and time calibration parameter-related metadata are stored in a key-value database; the stored raw data is standardized, and the audio signal is framed and windowed before Mel-frequency cepstral coefficients are extracted; environmental parameters are normalized by Z-score; user behavior data uses principal component analysis PCA to reduce the dimensionality of the original features, and finally the processed data is compressed and stored.

[0007] Furthermore, step S200 includes: Step S201: Input the pre-processed multi-source data into the graph neural network to construct a dynamic association graph G t =(V t ,E t ); where the node set V t ={v1,v2,...,v n} represents a smart speaker device, where v n represents the nth node, representing the nth smart speaker device; each node feature vector integrates three-dimensional position, audio features, environmental parameters and user behavior patterns to form a multi-dimensional feature representation; the edge set E t ={e ij} represents the association relationship between devices, and the edge weight is calculated by the adaptive graph attention mechanism: w t ij =f(d ij ,c t ij ,l t ij ), where w t ijt represents the weight of the associated edge between smart speakers i and j at time t, d ij represents the physical distance between smart speaker i and smart speaker j, c t ij represents the cosine similarity of the audio signals of smart speakers i and j at time t, l t ij represents the network transmission delay between smart speaker i and smart speaker j at time t; Step S202: Based on the dynamic association graph G t =(V t ,E t), the node feature vector is input into the multi-head graph attention layer, the information transfer weight between nodes is calculated through the attention mechanism, and the node embedding representation is generated through the graph convolution operation. The global pooling layer aggregates all node information to obtain the global feature vector h extracted by the graph neural network G ; Based on the extracted graph features h G , scene classification is performed through a multi-layer perceptron, and the probability distribution of the scene type is output: softmax(W2ReLU(W1h G +b1)+b2), where W1 and W2 represent the weight matrices of the multi-layer perceptron, and b1 and b2 represent the bias vectors; finally, the scene type set Y is obtained; a multi-dimensional collaborative evaluation system is simultaneously constructed, and the comprehensive collaborative score is: S=a1ΔΦ+a2C v +a3S n Quantify the speaker coordination quality, where ΔΦ represents the audio phase synchronization error, C v Indicates volume consistency, S n Indicates network stability, the weight coefficient a1+a2+a3=1, which is determined by offline optimization. The optimization goal is to minimize the evaluation error on the validation set; Step S203: By performing feature extraction, cluster analysis and data annotation on historical multi-source data, a historical abnormal pattern library B = {B1, B2, ..., Bm} is constructed, where Bm represents the mth abnormal pattern; by calculating the cosine similarity sim (b t ,B j ), combined with the risk propagation algorithm to dynamically update the risk score: R t =λR t-1 +(1-λ)max m j=1 sim(b t ,B j ), where R t represents the risk score at time t, R t ∈[0, 1]; λ is the forgetting factor, which is used to control the decay rate of historical risks; the risk level is divided into low, medium and high levels according to the preset threshold {T1, T2}, forming a graded early warning mechanism; Step S204: The scenario type set Y obtained in step S202, the calculated comprehensive collaboration score S, and the risk level determined in step S203 are subjected to spatiotemporal weighted fusion to generate the final comprehensive evaluation result; in the time dimension, the results of the latest k historical evaluation cycles are subjected to exponential weighting to obtain the time dimension fusion evaluation vector R time In the spatial dimension, the physical distance between the user's real-time location and each smart speaker is weighted to obtain the spatial dimension fusion evaluation vector R space; The evaluation results of the time dimension and the space dimension are integrated to obtain the spatiotemporal comprehensive evaluation vector: R f =γR time +(1-γ)R space , where R f Indicates that γ is the balance coefficient, which is determined by cross-validation; finally, the fused multi-dimensional indicators are integrated into the final evaluation result vector R final .

[0008] Furthermore, step S300 includes: Step S301: Based on the comprehensive evaluation results, the current scenario type is analyzed to determine the specific requirements for the collaborative operation of multiple smart speakers. The quality level of the current multi-speaker collaborative operation is determined. If the score is lower than the preset threshold, the collaborative optimization process is initiated. In addition, potential risk points are matched from the historical abnormal pattern library based on the risk level. Step S302: Real-time network status information between speakers is acquired to construct a dynamic network topology diagram, and the network connection status between nodes is updated in real time. The optimal transmission path is calculated by using an improved Dijkstra algorithm, taking into account network latency, bandwidth utilization, and path stability. This allows audio data to be transmitted to each speaker in the most stable manner with the lowest latency and highest bandwidth utilization. At the same time, based on the performance parameters (CPU computing power, memory capacity, audio decoding capability) and current load of each speaker, a priority-based task scheduling algorithm is used to assign complex audio processing tasks to speakers with performance scores above a threshold and loads below 50%, and to assign simple audio playback tasks to the remaining devices. Step S303: During the execution of the strategy, try to monitor the audio synchronization error, volume consistency, and network stability of each speaker, and through data analysis and comparison, determine whether these indicators meet the preset collaborative quality requirements; if the audio synchronization error exceeds the preset threshold, the audio playback between the speakers is synchronized through the time synchronization protocol; if the volume consistency exceeds the preset threshold, the output gain is automatically adjusted according to the volume deviation of each speaker through the volume adjustment algorithm; if network instability is detected, immediately re-evaluate the network status, trigger the path planning algorithm in step S302, and dynamically switch to the backup transmission path.

[0009] Furthermore, step S400 includes: Step S401: Integrate historical multi-dimensional spatiotemporal data, scenario evaluation results, and collaborative optimization strategy execution records to construct a spatiotemporal prediction dataset; construct a spatiotemporal joint neural network architecture model, with the input layer receiving the integrated multi-dimensional feature vector; in the time dimension, a bidirectional gated recurrent unit network is used as the time feature extraction layer. Through forward and backward propagation, the periodic changes in user behavior patterns and the evolution of speaker status in the time series are analyzed to obtain historical and future context information; in the spatial dimension, based on the constructed dynamic association graph, a graph attention network is used as the spatial feature extraction layer. The speaker position coordinates, audio signal similarity, and network connection status data are input to analyze the spatial correlation relationship between each smart speaker device and output a spatial embedding vector representing the strength of the collaborative relationship between devices; Step S402: The temporal and spatial features are integrated through the spatiotemporal fusion layer. Using concatenation and weighted attention mechanisms, the fusion ratio of temporal and spatial features is automatically adjusted to obtain a feature representation that includes information in both spatiotemporal and temporal dimensions. Finally, the model ends with a multi-task prediction layer. Based on the fused features, it outputs prediction results from two sub-networks in parallel: one sub-network focuses on predicting the speaker coordination status over a period of time, including audio synchronization error, volume consistency, and network stability indicators; the other sub-network predicts user behavior trends, including the transition probability of user activity areas and the classification of interaction patterns. Step S403: After the model completes the prediction output, the predicted probability results of the speaker collaboration status and user behavior trends in the future time period are continuously compared with the preset threshold; once it is found that the predicted probability of a certain indicator exceeds the corresponding threshold, the hierarchical warning mechanism is automatically triggered immediately, and the warning level is determined according to the threshold range, including high, medium, and low warnings, and the corresponding warning strategy is executed.

[0010] An artificial intelligence-based smart speaker synchronization and coordination monitoring system, including a data acquisition module, a scene analysis module, a collaborative optimization strategy module, and a monitoring management module; The data acquisition module uses multiple smart speakers distributed in different locations to synchronously collect audio signals, environmental parameters, user behavior data, and network status; performs timestamp calibration and standardization preprocessing on the collected multi-source data, constructs a multidimensional spatiotemporal dataset containing spatial location information, and performs preliminary feature extraction and data compression storage; The scenario analysis module inputs preprocessed data into a graph neural network to construct a dynamic association graph and analyzes the associations between data through graph convolution. It automatically identifies the current usage scenario by combining environmental characteristics with user behavior patterns, and simultaneously evaluates the multi-speaker collaboration status and user behavior risk level to generate a comprehensive evaluation result that includes scenario type, collaboration status, and risk level. The collaborative optimization strategy module generates a collaborative optimization strategy for multiple smart speakers based on the scenario evaluation results, dynamically adjusts the audio transmission path according to the current network status, and intelligently allocates computing resources and playback tasks to each speaker; The early warning management module builds a spatiotemporal prediction model based on historical collaboration data and real-time user behavior patterns, and makes probabilistic predictions on the speaker collaboration status and user behavior trends in future time periods. When the predicted probability exceeds a preset threshold, the hierarchical early warning mechanism is automatically triggered and the corresponding early warning strategy is executed.

[0011] The data acquisition module includes a data acquisition unit and a data processing unit; the data acquisition unit configures a three-dimensional location identifier and a logical identifier for each smart speaker, and synchronously collects audio signals, environmental parameters, user behavior data and network status through a microphone array, various sensors and a network interface; the data processing unit uses the PTP protocol to calibrate timestamps, constructs a data association matrix to integrate multi-source data, and constructs a two-layer data storage architecture; the raw data is standardized, including audio feature extraction, environmental parameter normalization and user behavior data dimensionality reduction, and finally completes data compression and storage.

[0012] The scenario analysis module includes a graph construction unit and a scenario evaluation unit; the graph construction unit inputs the preprocessed multi-source data into the graph neural network to construct a dynamic association graph; the scenario evaluation unit completes scene classification based on the graph features through the graph neural network and the multi-layer perceptron, and at the same time constructs a multi-dimensional collaborative evaluation system to quantify the collaborative quality of the speakers, and evaluates user behavior risks in combination with the historical abnormal pattern library, and generates comprehensive evaluation results through time-space weighted fusion.

[0013] The collaborative optimization strategy module includes a strategy generation unit, a resource scheduling unit, and a status monitoring unit. The strategy generation unit analyzes the comprehensive evaluation results to clarify the scenario collaboration requirements and judge the current quality level of multi-speaker collaborative work. If it is lower than the threshold, the optimization process is started, and potential risk points are identified to generate a basic collaborative strategy. The resource scheduling unit constructs a dynamic network topology map based on the network status, calculates the optimal audio transmission path, and intelligently allocates computing resources and playback tasks based on the speaker performance and load conditions. The status monitoring unit monitors the speaker collaboration status indicators in real time and takes corresponding adjustment measures in a timely manner when problems such as audio synchronization errors, inconsistent volume, and network instability occur. The early warning management module includes a model construction unit and a prediction and early warning unit; the model construction unit constructs a spatiotemporal prediction data set by integrating historical data, and constructs a spatiotemporal joint neural network architecture model to predict the speaker collaboration status and user behavior trends; the prediction and early warning unit uses a trained model to make probabilistic predictions, and compares the results with a preset threshold. Once the threshold is exceeded, a graded early warning mechanism is triggered, and corresponding strategies are executed according to different warning levels.

[0014] Compared with the prior art, the present invention has the following beneficial effects: By constructing a multidimensional spatiotemporal dataset containing spatial location information and using graph neural networks to construct and analyze dynamic association graphs, this invention achieves accurate assessment and scenario recognition of the collaborative status of multiple smart speakers and user behavior. This not only improves the efficiency and quality of multi-speaker collaborative work, but also dynamically adjusts audio transmission paths and computing resource allocation according to real-time network status, ensuring the synchronization and stability of audio playback. The present invention constructs a spatiotemporal prediction model to make probabilistic predictions on the speaker collaboration status and user behavior trends in future time periods, and automatically triggers a graded early warning mechanism when potential risks are predicted, enabling smart speakers to perceive and respond to possible abnormal situations in advance, further enhancing the reliability and user experience of the multi-smart speaker system. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings: Figure 1 The present invention is a method flow chart of a method for synchronous coordination monitoring of smart speakers based on artificial intelligence. DETAILED DESCRIPTION

[0016] 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 the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0017] See also Figure 1 The present invention provides a technical solution: a method for synchronous coordination monitoring of smart speakers based on artificial intelligence, comprising: Step S100: Synchronously collect audio signals, environmental parameters, user behavior data, and network status through multiple smart speakers distributed in different locations; perform timestamp calibration and standardization preprocessing on the collected multi-source data, construct a multidimensional spatiotemporal dataset containing spatial location information, and perform preliminary feature extraction and data compression storage; Step S200: Input the preprocessed data into the graph neural network to construct a dynamic association graph, and analyze the associations between the data through graph convolution; combine environmental characteristics and user behavior patterns to automatically identify the current usage scenario, simultaneously evaluate the multi-speaker collaboration status and user behavior risk level, and generate a comprehensive evaluation result that includes scenario type, collaboration status, and risk level; Step S300: Based on the scenario evaluation results, a collaborative optimization strategy for multiple smart speakers is generated, the audio transmission path is dynamically adjusted according to the current network status, and the computing resources and playback tasks of each speaker are intelligently allocated; Step S400: Based on historical collaboration data and real-time user behavior patterns, a spatiotemporal prediction model is constructed to make probabilistic predictions on the speaker collaboration status and user behavior trends in future time periods. When the predicted probability exceeds a preset threshold, a graded warning mechanism is automatically triggered and the corresponding warning strategy is executed.

[0018] Furthermore, step S100 includes: Step S101: Assign a unique three-dimensional position identifier P to each smart speaker i =(x i ,y i ,z i ), where (x i ,y i ,z i ) represents the coordinate values ​​of smart speaker i on the x-axis, y-axis and z-axis, determined based on the indoor Cartesian coordinate system, with the origin of the coordinates set to the position of the master device, and is assigned a logical identifier ID i Forming a device identification pair (ID i ,P i ), where ID i A unique digital identifier representing smart speaker i. Each smart speaker uses a microphone array to collect ambient background sound, user voice, and playback audio signals, and employs a beamforming algorithm to suppress ambient noise. Temperature, humidity, and air pressure are collected synchronously using temperature and humidity sensors and air pressure sensors. Accelerometers and infrared sensors are used to capture user motion amplitude and movement trajectory at a preset frequency. A network interface acquires network status parameters in real time at set intervals, and all data is generated with an initial timestamp. Step S102: Use the PTP protocol to globally calibrate the timestamps of each speaker and synchronize the initial time to a unified timestamp; construct a data association matrix M data, audio, environment, behavior and network data in the same time window are sorted by device ID (ID i ,P i ) for association and integration; a two-layer data storage architecture is constructed: the first storage unit uses column storage to save audio signals, environmental parameters and network status; the second storage unit uses a graph database to store user behavior sequence data, with nodes recording behavior status and edges representing state transition relationships; at the same time, metadata related to device identification pairs and time calibration parameters are stored in a key-value database; the stored raw data is standardized, and the audio signal is framed and windowed before the Mel-frequency cepstral coefficients are extracted; the environmental parameters are normalized by Z-score; the user behavior data uses principal component analysis (PCA) to reduce the dimensionality of the original features, and finally the processed data is compressed and stored.

[0019] Step S200 includes: Step S201: Input the pre-processed multi-source data into the graph neural network to construct a dynamic association graph G t =(V t ,E t ); where the node set V t ={v1,v2,...,v n} represents a smart speaker device, where v n represents the nth node, representing the nth smart speaker device; each node feature vector integrates three-dimensional position, audio features, environmental parameters and user behavior patterns to form a multi-dimensional feature representation; the edge set E t ={e ij} represents the association relationship between devices, and the edge weight is calculated by the adaptive graph attention mechanism: w t ij =f(d ij ,c t ij ,l t ij ), where w t ijt represents the weight of the associated edge between smart speakers i and j at time t, d ij represents the physical distance between smart speaker i and smart speaker j, c t ij represents the cosine similarity of the audio signals of smart speakers i and j at time t, l t ij represents the network transmission delay between smart speaker i and smart speaker j at time t; Step S202: Based on the dynamic association graph G t =(V t ,E t), the node feature vector is input into the multi-head graph attention layer, the information transfer weight between nodes is calculated through the attention mechanism, and the node embedding representation is generated through the graph convolution operation. The global pooling layer aggregates all node information to obtain the global feature vector h extracted by the graph neural network G ; Based on the extracted graph features h G , scene classification is performed through a multi-layer perceptron, and the probability distribution of the scene type is output: softmax(W2ReLU(W1h G +b1)+b2), where W1 and W2 represent the weight matrices of the multi-layer perceptron, and b1 and b2 represent the bias vectors; finally, the scene type set Y is obtained; a multi-dimensional collaborative evaluation system is simultaneously constructed, and the comprehensive collaborative score is: S=a1ΔΦ+a2C v +a3S n Quantify the speaker coordination quality, where ΔΦ represents the audio phase synchronization error, C v Indicates volume consistency, S n Indicates network stability, the weight coefficient a1+a2+a3=1, which is determined by offline optimization. The optimization goal is to minimize the evaluation error on the validation set; Step S203: By performing feature extraction, cluster analysis and data annotation on historical multi-source data, a historical abnormal pattern library B = {B1, B2, ..., Bm} is constructed, where Bm represents the mth abnormal pattern; by calculating the cosine similarity sim (b t ,B j ), combined with the risk propagation algorithm to dynamically update the risk score: R t =λR t-1 +(1-λ)max m j=1 sim(b t ,B j ), where R t represents the risk score at time t, R t ∈[0, 1]; λ is the forgetting factor, which is used to control the decay rate of historical risks; the risk level is divided into low, medium and high levels according to the preset threshold {T1, T2}, forming a graded early warning mechanism; Step S204: The scenario type set Y obtained in step S202, the calculated comprehensive collaboration score S, and the risk level determined in step S203 are subjected to spatiotemporal weighted fusion to generate the final comprehensive evaluation result; in the time dimension, the results of the latest k historical evaluation cycles are subjected to exponential weighting to obtain the time dimension fusion evaluation vector R time In the spatial dimension, the physical distance between the user's real-time location and each smart speaker is weighted to obtain the spatial dimension fusion evaluation vector R space; The evaluation results of the time dimension and the space dimension are integrated to obtain the spatiotemporal comprehensive evaluation vector: R f =γR time +(1-γ)R space , where R f Indicates that γ is the balance coefficient, which is determined by cross-validation; finally, the fused multi-dimensional indicators are integrated into the final evaluation result vector R final .

[0020] Step S300 includes: Step S301: Based on the comprehensive evaluation results, the current scenario type is analyzed to determine the specific requirements for the collaborative operation of multiple smart speakers. The quality level of the current multi-speaker collaborative operation is determined. If the score is lower than the preset threshold, the collaborative optimization process is initiated. In addition, potential risk points are matched from the historical abnormal pattern library based on the risk level. Step S302: Real-time network status information between speakers is acquired to construct a dynamic network topology diagram, and the network connection status between nodes is updated in real time. The optimal transmission path is calculated by using an improved Dijkstra algorithm, taking into account network latency, bandwidth utilization, and path stability. This allows audio data to be transmitted to each speaker in the most stable manner with the lowest latency and highest bandwidth utilization. At the same time, based on the performance parameters (CPU computing power, memory capacity, audio decoding capability) and current load of each speaker, a priority-based task scheduling algorithm is used to assign complex audio processing tasks to speakers with performance scores above a threshold and loads below 50%, and to assign simple audio playback tasks to the remaining devices. Step S303: During the execution of the strategy, try to monitor the audio synchronization error, volume consistency, and network stability of each speaker, and through data analysis and comparison, determine whether these indicators meet the preset collaborative quality requirements; if the audio synchronization error exceeds the preset threshold, the audio playback between the speakers is synchronized through the time synchronization protocol; if the volume consistency exceeds the preset threshold, the output gain is automatically adjusted according to the volume deviation of each speaker through the volume adjustment algorithm; if network instability is detected, immediately re-evaluate the network status, trigger the path planning algorithm in step S302, and dynamically switch to the backup transmission path.

[0021] Step S400 includes: Step S401: Integrate historical multi-dimensional spatiotemporal data, scenario evaluation results, and collaborative optimization strategy execution records to construct a spatiotemporal prediction dataset; construct a spatiotemporal joint neural network architecture model, with the input layer receiving the integrated multi-dimensional feature vector; in the time dimension, a bidirectional gated recurrent unit network is used as the time feature extraction layer. Through forward and backward propagation, the periodic changes in user behavior patterns and the evolution of speaker status in the time series are analyzed to obtain historical and future context information; in the spatial dimension, based on the constructed dynamic association graph, a graph attention network is used as the spatial feature extraction layer. The speaker position coordinates, audio signal similarity, and network connection status data are input to analyze the spatial correlation relationship between each smart speaker device and output a spatial embedding vector representing the strength of the collaborative relationship between devices; Step S402: The temporal and spatial features are integrated through the spatiotemporal fusion layer. Using concatenation and weighted attention mechanisms, the fusion ratio of temporal and spatial features is automatically adjusted to obtain a feature representation that includes information in both spatiotemporal and temporal dimensions. Finally, the model ends with a multi-task prediction layer. Based on the fused features, it outputs prediction results from two sub-networks in parallel: one sub-network focuses on predicting the speaker coordination status over a period of time, including audio synchronization error, volume consistency, and network stability indicators; the other sub-network predicts user behavior trends, including the transition probability of user activity areas and the classification of interaction patterns. Step S403: After the model completes the prediction output, the predicted probability results of the speaker collaboration status and user behavior trends in the future time period are continuously compared with the preset threshold; once it is found that the predicted probability of a certain indicator exceeds the corresponding threshold, the hierarchical warning mechanism is automatically triggered immediately, and the warning level is determined according to the threshold range, including high, medium, and low warnings, and the corresponding warning strategy is executed.

[0022] In an embodiment of the present invention, three smart speakers (ID1-ID3) are deployed in the living room, bedroom, study, and other locations to build a home smart audio network. Each speaker uses a microphone array to collect ambient sound, including TV sound, door and window opening and closing sounds, and user voice commands. At the same time, a temperature and humidity sensor monitors the indoor temperature and humidity. An acceleration sensor is used to capture the user's movement trajectory in different rooms, such as the path from the living room to the bedroom, and real-time acquisition of Wi-Fi network latency (average 20ms), bandwidth (50Mbps), etc. All data is generated with an initial timestamp and calibrated to a unified time through the PTP protocol. A data association matrix is ​​constructed and the device ID (ID i ,P i) Integrate data within the same time window: The audio signal collected by the ID1 speaker at 14:00:05, the temperature and humidity data at 24°C, the user's movement trajectory in the living room, and the network delay parameters are associated and stored; the audio signal is processed through frame segmentation and windowing, the environmental parameters are normalized using Z-score, and the user behavior data is reduced in dimension using PCA. Finally, the data is compressed and stored in a column-based database (storing audio and environmental data) and a graph database (storing behavior sequences); Input the preprocessed data into the graph neural network to construct a dynamic association graph G t Node features include the three-dimensional position of the speaker (ID1 coordinates (1,2,0)), audio features (spectral features of the voice command "play classical music"), environmental parameters (temperature and humidity), and user behavior patterns (moving speed 0.5m / s); edge weights are calculated using an adaptive graph attention mechanism: the physical distance between ID1 and ID2 is 3 meters, the cosine similarity of the audio signal is 0.85, and the network delay is 25ms; a global feature vector h is generated through a multi-head graph attention layer and graph convolution operations. G , after multi-layer perceptron classification, it is classified as "family leisure" scene with a probability of 0.92; synchronous calculation collaboration score: audio phase synchronization error ΔΦ = 0.03, weight a1 = 0.4, volume consistency C v =0.9, a2=0.3, network stability S n =0.85, a3=0.3, the comprehensive score S=0.82, which is higher than the set threshold of 0.7, indicating good collaboration quality; at the same time, the current user behavior characteristics are extracted: sitting for more than 30 minutes, and the cosine similarity with the "health risk" pattern B2 in the historical abnormal pattern library is 0.68. Combined with the risk propagation algorithm, λ=0.7 is set to update the risk score R t =0.52, it is judged as medium risk, and the health monitoring warning is immediately triggered; Based on the needs of the "family leisure" scenario, the analysis shows that stereo music playback is required. The current collaborative comprehensive score of 0.82 meets the requirements and there is no need to start the optimization process. However, due to the medium risk level, the historical abnormal pattern library is matched to the "sedentary reminder" strategy; the network status is monitored in real time, and a topology map is constructed to show that the delay between ID1 and ID3 is 30ms, which is higher than the set threshold of 25ms. The improved Dijkstra algorithm is used to re-plan the path to ID1→ID2→ID3, reducing the delay to 22ms; task allocation is planned, and the high computing power demand of the audio decoding task is allocated to the high-performance speaker ID3 in the study (CPU utilization rate 30%), and the playback task is allocated to the living room speaker ID1 and the bedroom speaker ID2; during the execution process, if the volume consistency deviation of ID2 exceeds 5%, its gain parameter is automatically adjusted to keep the volume consistent with ID1; at the same time, the network stability is continuously monitored to ensure smooth music playback; The system integrates historical data from the past 24 hours, including user routines and speaker collaboration strategy execution records, to construct a spatiotemporal prediction dataset. The model predicts the next hour: the probability of the user continuing to sit for a long time is 0.88, exceeding the medium-risk threshold of 0.85; the probability of speaker network fluctuation is 0.72, which is lower than the low-risk threshold of 0.75; triggering a medium-risk warning: the user is reminded to get up and move around through voice through the ID1 speaker, and a sedentary health reminder is pushed to the user's mobile phone. The music data for the next 30 minutes is cached in advance to the ID2 speaker to prevent possible network fluctuations.

[0023] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above and that the invention can be embodied in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments should be considered in all respects as illustrative and non-restrictive, and the scope of the invention is defined by the appended claims, not the foregoing description, and all variations within the meaning and range of equivalents of the claims are intended to be included therein. Any reference sign in a claim should not be construed as limiting the claim to which it relates.

Claims

1. A method for synchronous coordination monitoring of smart speakers based on artificial intelligence, characterized by: The method comprises: Step S100: Synchronously collect audio signals, environmental parameters, user behavior data, and network status through multiple smart speakers distributed in different locations; perform timestamp calibration and standardization preprocessing on the collected multi-source data, construct a multidimensional spatiotemporal dataset containing spatial location information, and perform preliminary feature extraction and data compression storage; Step S200: Input the preprocessed data into the graph neural network to construct a dynamic association graph, and analyze the associations between the data through graph convolution; combine environmental characteristics and user behavior patterns to automatically identify the current usage scenario, simultaneously evaluate the multi-speaker collaboration status and user behavior risk level, and generate a comprehensive evaluation result that includes scenario type, collaboration status, and risk level; Step S300: Based on the scenario evaluation results, a collaborative optimization strategy for multiple smart speakers is generated, the audio transmission path is dynamically adjusted according to the current network status, and the computing resources and playback tasks of each speaker are intelligently allocated; Step S400: Based on historical collaboration data and real-time user behavior patterns, a spatiotemporal prediction model is constructed to make probabilistic predictions on the speaker collaboration status and user behavior trends in future time periods. When the predicted probability exceeds a preset threshold, a graded warning mechanism is automatically triggered and the corresponding warning strategy is executed.

2. The method for synchronous coordination monitoring of smart speakers based on artificial intelligence according to claim 1, characterized in that: The step S100 includes: Step S101: Assign a unique three-dimensional position identifier P to each smart speaker i =(x i ,y i ,z i ), where (x i ,y i ,z i ) represents the coordinate values ​​of smart speaker i on the x-axis, y-axis and z-axis, determined based on the indoor Cartesian coordinate system, with the origin of the coordinates set to the position of the master device, and is assigned a logical identifier ID i Forming a device identification pair (ID i ,P i ), where ID i A unique digital identifier representing smart speaker i. Each smart speaker uses a microphone array to collect ambient background sound, user voice, and playback audio signals, and employs a beamforming algorithm to suppress ambient noise. Temperature, humidity, and air pressure are collected synchronously using temperature and humidity sensors and air pressure sensors. Accelerometers and infrared sensors are used to capture user motion amplitude and movement trajectory at a preset frequency. A network interface acquires network status parameters in real time at set intervals, and all data is generated with an initial timestamp. Step S102: Use the PTP protocol to globally calibrate the timestamps of each speaker and synchronize the initial time to a unified timestamp; construct a data association matrix M data , audio, environment, behavior and network data in the same time window are sorted by device ID (ID i ,P i ) for association and integration; a two-layer data storage architecture is constructed: the first storage unit uses column storage to save audio signals, environmental parameters and network status; the second storage unit uses a graph database to store user behavior sequence data, with nodes recording behavior status and edges representing state transition relationships; at the same time, device identification pairs and time calibration parameter-related metadata are stored in a key-value database; the stored raw data is standardized, and the audio signal is framed and windowed to extract the Mel-frequency cepstral coefficients; the environmental parameters are normalized by Z-score; the user behavior data uses principal component analysis to reduce the dimensionality of the original features, and finally the processed data is compressed and stored.

3. The method for synchronous coordination monitoring of smart speakers based on artificial intelligence according to claim 1, characterized in that: The step S200 includes: Step S201: Input the pre-processed multi-source data into the graph neural network to construct a dynamic association graph G t =(V t ,E t ); where the node set V t ={v1,v2,...,v n } represents a smart speaker device, where v n represents the nth node, representing the nth smart speaker device; each node feature vector integrates three-dimensional position, audio features, environmental parameters and user behavior patterns to form a multi-dimensional feature representation; the edge set E t ={e ij } represents the association relationship between devices, and the edge weight is calculated by the adaptive graph attention mechanism: w t ij =f(d ij ,c t ij ,l t ij ), where w t ij represents the weight of the associated edge between smart speakers i and j at time t, d ij represents the physical distance between smart speaker i and smart speaker j, c t ij represents the cosine similarity of the audio signals of smart speakers i and j at time t, l t ij represents the network transmission delay between smart speaker i and smart speaker j at time t; Step S202: Based on the dynamic association graph G t =(V t ,E t ), the node feature vector is input into the multi-head graph attention layer, the information transfer weight between nodes is calculated through the attention mechanism, and the node embedding representation is generated through the graph convolution operation. The global pooling layer aggregates all node information to obtain the global feature vector h extracted by the graph neural network G ; Based on the extracted graph features h G , scene classification is performed through a multi-layer perceptron, and the probability distribution of the scene type is output: softmax(W2ReLU(W1h G +b1)+b2), where W1 and W2 represent the weight matrices of the multi-layer perceptron, and b1 and b2 represent the bias vectors; finally, the scene type set Y is obtained; a multi-dimensional collaborative evaluation system is simultaneously constructed, and the comprehensive collaborative score is: S=a1ΔΦ+a2C v +a3S n Quantify the speaker coordination quality, where ΔΦ represents the audio phase synchronization error, C v Indicates volume consistency, S n Indicates network stability, the weight coefficient a1+a2+a3=1, which is determined by offline optimization. The optimization goal is to minimize the evaluation error on the validation set; Step S203: By performing feature extraction, cluster analysis and data annotation on historical multi-source data, a historical abnormal pattern library B = {B1, B2, ..., Bm} is constructed, where Bm represents the mth abnormal pattern; by calculating the cosine similarity sim (b t ,B j ), combined with the risk propagation algorithm to dynamically update the risk score: R t =λR t-1 +(1-λ)max m j=1 sim(b t ,B j ), where R t represents the risk score at time t, R t ∈[0, 1]; λ is the forgetting factor, which is used to control the decay rate of historical risks; the risk level is divided into low, medium and high levels according to the preset threshold {T1, T2}, forming a graded early warning mechanism; Step S204: The scenario type set Y obtained in step S202, the calculated comprehensive collaboration score S, and the risk level determined in step S203 are subjected to spatiotemporal weighted fusion to generate the final comprehensive evaluation result; in the time dimension, the results of the latest k historical evaluation cycles are subjected to exponential weighting to obtain the time dimension fusion evaluation vector R time In the spatial dimension, the physical distance between the user's real-time location and each smart speaker is weighted to obtain the spatial dimension fusion evaluation vector R space ; The evaluation results of the time dimension and the space dimension are integrated to obtain the spatiotemporal comprehensive evaluation vector: R f =γR time +(1-γ)R space , where R f Indicates that γ is the balance coefficient, which is determined by cross-validation; finally, the fused multi-dimensional indicators are integrated into the final evaluation result vector R final .

4. The method for synchronous coordination monitoring of smart speakers based on artificial intelligence according to claim 1, characterized in that: The step S300 includes: Step S301: Based on the comprehensive evaluation results, the current scenario type is analyzed to determine the specific requirements for the collaborative operation of multiple smart speakers. The quality level of the current multi-speaker collaborative operation is determined. If the score is lower than the preset threshold, the collaborative optimization process is initiated. In addition, potential risk points are matched from the historical abnormal pattern library based on the risk level. Step S302: The network status information between each speaker is acquired in real time, a dynamic network topology is constructed, the network connection status between each node is updated in real time, and the optimal transmission path is calculated by using an improved Dijkstra algorithm, taking into account network latency, bandwidth utilization, and path stability. This allows audio data to be transmitted to each speaker in the most stable manner with the lowest latency and highest bandwidth utilization. At the same time, based on the performance parameters and current load of each speaker, a priority-based task scheduling algorithm is used to assign complex audio processing tasks to speakers with performance scores above a set threshold and loads below a preset load threshold, and to assign simple audio playback tasks to the remaining devices. Step S303: During the execution of the strategy, the audio synchronization error, volume consistency, and network stability of each speaker are monitored in real time, and through data analysis and comparison, it is determined whether these indicators meet the preset collaborative quality requirements; if the audio synchronization error exceeds the preset threshold, the audio playback between the speakers is synchronized through the time synchronization protocol; if the volume consistency exceeds the preset threshold, the output gain is automatically adjusted according to the volume deviation of each speaker through the volume adjustment algorithm; if network instability is detected, the network status is immediately re-evaluated, the path planning algorithm in step S302 is triggered, and dynamic switching to the backup transmission path is performed.

5. The method for synchronous coordination monitoring of smart speakers based on artificial intelligence according to claim 1, characterized in that: The step S400 includes: Step S401: Integrate historical multi-dimensional spatiotemporal data, scenario evaluation results, and collaborative optimization strategy execution records to construct a spatiotemporal prediction dataset; construct a spatiotemporal joint neural network architecture model, with the input layer receiving the integrated multi-dimensional feature vector; in the time dimension, a bidirectional gated recurrent unit network is used as the time feature extraction layer. Through forward and backward propagation, the periodic changes in user behavior patterns and the evolution of speaker status in the time series are analyzed to obtain historical and future context information; in the spatial dimension, based on the constructed dynamic association graph, a graph attention network is used as the spatial feature extraction layer. The speaker position coordinates, audio signal similarity, and network connection status data are input to analyze the spatial correlation relationship between each smart speaker device and output a spatial embedding vector representing the strength of the collaborative relationship between devices; Step S402: The temporal and spatial features are integrated through the spatiotemporal fusion layer. Using concatenation and weighted attention mechanisms, the fusion ratio of temporal and spatial features is automatically adjusted to obtain a feature representation that includes information in both spatiotemporal and temporal dimensions. Finally, the model ends with a multi-task prediction layer. Based on the fused features, it outputs the prediction results of two sub-networks in parallel: one sub-network focuses on predicting the speaker coordination status over a period of time, including audio synchronization error, volume consistency, and network stability indicators; the other sub-network predicts user behavior trends, including the transition probability of user activity areas and the classification of interaction patterns. Step S403: After the model completes the prediction output, the predicted probability results of the speaker collaboration status and user behavior trends in the future time period are continuously compared with the preset threshold; once it is found that the predicted probability of a certain indicator exceeds the corresponding threshold, the hierarchical warning mechanism is automatically triggered immediately, and the warning level is determined according to the threshold range, including high, medium, and low warnings, and the corresponding warning strategy is executed.

6. An artificial intelligence-based smart speaker synchronization and coordination monitoring system, characterized by: The system includes a data acquisition module, a scenario analysis module, a collaborative optimization strategy module, and a monitoring and management module; The data acquisition module uses multiple smart speakers distributed in different locations to synchronously collect audio signals, environmental parameters, user behavior data, and network status; performs timestamp calibration and standardization preprocessing on the collected multi-source data, constructs a multidimensional spatiotemporal dataset containing spatial location information, and performs preliminary feature extraction and data compression storage; The scenario analysis module inputs preprocessed data into a graph neural network to construct a dynamic association graph and analyzes the associations between data through graph convolution. It automatically identifies the current usage scenario by combining environmental characteristics with user behavior patterns, and simultaneously evaluates the multi-speaker collaboration status and user behavior risk level to generate a comprehensive evaluation result that includes scenario type, collaboration status, and risk level. The collaborative optimization strategy module generates a collaborative optimization strategy for multiple smart speakers based on the scenario evaluation results, dynamically adjusts the audio transmission path according to the current network status, and intelligently allocates computing resources and playback tasks to each speaker; The early warning management module builds a spatiotemporal prediction model based on historical collaboration data and real-time user behavior patterns, and makes probabilistic predictions on the speaker collaboration status and user behavior trends in future time periods. When the predicted probability exceeds a preset threshold, the hierarchical early warning mechanism is automatically triggered and the corresponding early warning strategy is executed.

7. The method for synchronous coordination monitoring of smart speakers based on artificial intelligence according to claim 6, characterized in that: The data acquisition module includes a data acquisition unit and a data processing unit. The data acquisition unit configures a three-dimensional location identifier and a logical identifier for each smart speaker, and synchronously collects audio signals, environmental parameters, user behavior data, and network status through a microphone array, various sensors, and a network interface. The data processing unit uses the PTP protocol to calibrate timestamps, constructs a data association matrix to integrate multi-source data, and builds a two-layer data storage architecture. The raw data is standardized, including audio feature extraction, environmental parameter normalization, and user behavior data dimensionality reduction, and finally data compression and storage are completed.

8. The method for synchronous coordination monitoring of smart speakers based on artificial intelligence according to claim 6, characterized in that: The scenario analysis module includes a graph construction unit and a scenario evaluation unit; the graph construction unit inputs the preprocessed multi-source data into the graph neural network to construct a dynamic association graph; the scenario evaluation unit classifies scenes based on the graph features through the graph neural network and the multi-layer perceptron; at the same time, a multi-dimensional collaborative evaluation system is constructed to quantify the collaborative quality of the speakers, and the user behavior risk is evaluated in combination with the historical abnormal pattern library, and a comprehensive evaluation result is generated through time-space weighted fusion.

9. The method for synchronous coordination monitoring of smart speakers based on artificial intelligence according to claim 6, characterized in that: The collaborative optimization strategy module includes a strategy generation unit, a resource scheduling unit, and a status monitoring unit. The strategy generation unit analyzes the comprehensive evaluation results to clarify the scenario collaboration requirements and judge the current quality level of multi-speaker collaborative work. If it is below the threshold, it starts the optimization process, identifies potential risk points, and generates a basic collaborative strategy. The resource scheduling unit constructs a dynamic network topology diagram based on the network status, calculates the optimal audio transmission path, and intelligently allocates computing resources and playback tasks based on speaker performance and load conditions; the status monitoring unit monitors the speaker collaboration status indicators in real time, and takes corresponding adjustment measures in a timely manner when problems such as audio synchronization errors, inconsistent volume, and network instability occur.

10. The method for synchronous coordination monitoring of smart speakers based on artificial intelligence according to claim 6, characterized in that: The early warning management module includes a model construction unit and a prediction and early warning unit; the model construction unit constructs a spatiotemporal prediction data set by integrating historical data, and constructs a spatiotemporal joint neural network architecture model to predict the speaker collaboration status and user behavior trends; the prediction and early warning unit uses a trained model to make probabilistic predictions, and compares the results with a preset threshold. Once the threshold is exceeded, a graded early warning mechanism is triggered, and corresponding strategies are executed according to different warning levels.

Images