A control method of an intelligent sound system

Abstract

The present application relates to the technical field of sound control, in particular to a control method of an intelligent sound system, comprising: inputting environment parameters collected in a current scene, control instructions and sound real-time running parameters into a three-layer structure of an input layer, a state layer and an output layer, extracting control variables, mapping the control variables with corresponding scenes, and establishing scene mapping data; calling an execution process of each control instruction in the scene mapping data, constructing a state transition graph according to sound state changes; confirming personnel positions under any state node in the state transition graph, performing spatial clustering according to the positions of the personnel, and calculating transition probabilities associated with the position clusters; associating sound placement positions with the transition probabilities, and determining sound control results; and converting the control results into state node control sequences, and determining position control strategies in combination with control time lengths. The efficiency and accuracy of sound control are realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of sound control, in particular to a control method of an intelligent sound system. BACKGROUND

[0002] With the integration of audio technology and Internet of Things technology, intelligent sound has been widely used in conference rooms, living rooms, shopping malls, education and training, etc. For example, conference rooms need to ensure the voice clarity when multiple people gather to discuss, home scenarios need to adapt to the volume balance of different rooms, and shopping malls need to balance the background music atmosphere and the guide voice prompt. In the current control of intelligent sound, manual control accounts for a high proportion, which is low in efficiency and relies on manual intervention. Its control dimension is single, and it is difficult to realize multi-scene adjustment.

[0003] For example, Chinese patent publication No. CN119136109A discloses an intelligent sound control method and system and storage medium. The present application focuses on the control effect of the interaction control of intelligent sound on home devices, can filter out the interaction control record with abnormal control effect based on the historical interaction control record, and then perform abnormal tracing and correlation analysis of the complexity of the voice control instruction. Thus, the adjustment direction of the sound interaction control is obtained.

[0004] For example, Chinese patent publication No. CN119789001A discloses a sound control method and sound based on human position detection. The position detection device of the sound is configured as a human position detection hardware, which is configured to detect the position of the target human in real time and obtain detection information. The processor device stores the correlation model of any of the above embodiments and is configured to perform similarity calculation on the detection information and position information according to the correlation model, and call the associated logical time sequence to control the power amplifier device. Through the pre-calibrated sound control logic, the logical time sequence is called according to the similarity calculation of the detection information and the position information.

[0005] The prior art determines the current interaction control effect of the sound through the complexity of the voice instruction, and performs similarity calculation on the position information of the sound through the human position and the logical time sequence to obtain the matching control process. However, these processing methods are inclined to isolated event processing of voice control, and the implementation process cannot be adjusted according to the running state of each sound and the specific scene, resulting in problems such as lack of coherence and dependence on fixed scenes in processing decisions, which reduces the efficiency of sound control. SUMMARY

[0006] In order to solve the above technical problems, the technical scheme adopted by the present application is: a control method of an intelligent sound system, comprising: S1, acquiring a playing signal of a sound, and based on environmental parameters collected in a current scene, control instructions and sound real-time running parameters, and in a hierarchical structure manner, extracting control variables under each layer structure in a three-layer structure of an input layer-state layer-output layer, and mapping the control variables with the corresponding scene to establish scene mapping data.

[0007] S2, calling the execution process of each control instruction in the scene mapping data, the execution process including execution duration, execution time and execution scene, connecting the execution process according to sound state changes to obtain a state transition graph of the current scene.

[0008] S3, confirming the personnel position under any state node in the state transition graph, clustering the space according to the position where the personnel is located, configuring the position cluster of the personnel position, and calculating the transition probability under the corresponding position cluster.

[0009] S4, associating the transition probability with the placement position of each sound to determine the sound control result.

[0010] S5, converting the sound control result into the control sequence of each state node in the state transition graph, and determining the control strategy of each position according to the control duration at each control time.

[0011] The beneficial effects of the present application are: firstly, the present application extracts control variables by a three-layer structure of an input layer-state layer-output layer, establishes scene mapping data, constructs a state transition graph according to sound state changes based on control instruction execution processes, generates a position cluster by spatial clustering of personnel positions, calculates the transition probability associated with the position cluster, associates the sound placement position with the transition probability to determine the sound control result, and converts the control result into the state node control sequence to determine the control strategy of each position in combination with the control duration. The scenes are associated according to the state, position, control instruction and control strategy, the transition probability of each scene is quantified, the precision of the control decision is improved, and the data backtracking problem caused by parameter configuration errors is reduced.

[0012] Secondly, the present application collects environmental parameters, control instructions and sound running parameters as initial data sets, establishes a three-layer control network of input-state-output, combines control variables in multiple scenes in a structured form, provides a data basis for subsequent processing, and then identifies frequent item sets of control variables through rule matching, processes each control variable in the frequent item set, so that the extracted control variables meet the characteristics of association and frequent occurrence, are more suitable for adjustment requirements in each scene, and improve the accuracy of scene mapping.

[0013] Thirdly, the application checks the control variable corresponding to the control instruction, merges the control variables of similar states, triggers the conditional execution time sequence, and generates a state transition graph by connecting the nodes with state changes; then, the average deviation of the clustering cluster is used as the difference threshold to determine the checking form of the control variable in the state layer, so that the node of the current data is simplified and the state validity is ensured; secondly, combined with the parameter changes of personnel position and volume, the checking of the overall scene is more biased towards the specific situation of sound in multiple environments, avoiding invalid parameters from being considered as state changes in the corresponding scene, and improving the accuracy of scene adaptation.

[0014] Fourthly, the application determines the position cluster corresponding to each node and groups them; the transition probability from the current state node to the next node is calculated by combining the position cluster; the optimal control direction is determined by combining the personnel position cluster and the sound coverage range; the control decision ensures technical feasibility and user adaptability, and improves the effect of sound control. BRIEF DESCRIPTION OF DRAWINGS

[0015] The application will be further described below in conjunction with the drawings and examples.

[0016] Figure 1 It is a flowchart of a control method of an intelligent sound system.

[0017] Figure 2 It is a flowchart of step S1 of a control method of an intelligent sound system.

[0018] Figure 3 It is a flowchart of step S2 of a control method of an intelligent sound system.

[0019] Figure 4 It is a flowchart of step S3 of a control method of an intelligent sound system.

[0020] Figure 5 It is a flowchart of step S4 of a control method of an intelligent sound system. DETAILED DESCRIPTION

[0021] The embodiments of the application will be described in detail below. The embodiments described below are exemplary and are only used to explain the application, and cannot be understood as a limitation of the application. If the specific technology or condition is not specified in the embodiments, the technology or condition described in the literature in the art or according to the product manual is used.

[0022] Reference Figure 1A control method for an intelligent speaker system includes: S1, acquiring the playback signal of the speaker, and based on the environmental parameters, control commands and real-time operating parameters of the speaker collected in the current scene, and in a hierarchical structure of three layers: input layer, state layer and output layer, extracting the control variables under each layer, and mapping the control variables with the corresponding scene to establish scene mapping data.

[0023] S2 retrieves the execution process of each control instruction in the scene mapping data. The execution process includes the execution duration, execution time, and execution scene. The execution process is connected according to the changes in the sound state to obtain the state transition diagram of the current scene.

[0024] S3. Confirm the personnel position under any state node in the state transition diagram, perform spatial clustering based on the personnel's position, configure the position cluster of the personnel position, and calculate the transition probability under the corresponding position cluster.

[0025] S4. The transition probability is correlated with the placement position of each speaker to determine the speaker control result.

[0026] S5 converts the audio control results into the control sequence of each state node in the state transition diagram, and determines the control strategy for each position according to the control duration of each control.

[0027] The system uses sensors such as microphones, infrared sensors, and millimeter-wave radar to collect the output signals or environmental conditions of the audio system in real time. The collected conditions are then used to adjust the volume and direction of the sound beam according to the volume beam direction under relative interference. The system employs PID control algorithms or neural network control to achieve precise volume and direction adjustment. When the system detects changes in the position of people or other changes in the sound beam direction, it will adjust the volume beam direction and volume in real time to adjust the volume detected at each position.

[0028] Essentially, it provides feedback processing for the effects of volume and playback direction of multiple speakers in scenarios such as conference rooms and museums, so that the sound played from multiple locations can be as clear and accurate as possible, thereby improving the control and regulation of the speakers in multiple environments.

[0029] In the current solution, the audio system in each scenario is structured into three layers: input, state, and output, according to its control commands. Through these three layers, the control variables for each audio control are extracted, and a control network structure related to the three layers is established.

[0030] Based on the relationships between the various control variables in the control network structure, these control variables in these dimensions are mapped to the corresponding scenarios to form scenario mapping data.

[0031] like Figure 2As shown, the implementation of step S1 also includes: S11, collecting environmental parameters, control commands and real-time audio operating parameters in the current scene as the initial dataset.

[0032] S12 establishes a three-layer structure of input, state, and output for the initial dataset, extracts the control variables for each audio control, and establishes a control network structure related to the three layers.

[0033] S13. Based on the correlation between various control variables in the control network structure, the control network structure is mapped to the specific scenario to generate scenario mapping data.

[0034] The control commands identified at this time include user-initiated commands (such as increasing the volume on the left) and system-automatic commands (such as pointing to the exhibit area triggered by scene linkage); the environmental parameters represent the personnel positions, personnel status, and current sound field output signals identified by multiple sensors; as for the real-time operating parameters, they represent the volume value set on the speakers, beam pointing angle, playback mode, and the clarity detected by the microphone array at each position.

[0035] The established three-layer structure uses the control command type (such as volume adjustment, direction switching), the current scene (such as conference room / museum), and the number of people as the input layer; the real-time parameters of the sound system (volume value, direction angle, current mode) and the dynamic state of the environment (changes in personnel position, noise fluctuations, personnel movement status, etc.) as the state layer; and the sound effect indicators of each location, such as clarity, volume consistency, and no feedback probability, are parameters detected by the microphone array as the output layer, which maps the current control command to the scene.

[0036] In other words, the input layer represents the scene features in the control commands and environmental parameters, the state layer represents the specific volume values ​​and the dynamic changes in the environment and the positions of people in the real-time operating parameters, and the output layer represents the sound effect indicators of each position in the real-time operating parameters, thus explaining the current three-layer structure.

[0037] The extracted control variables will be selected from parameters in the input layer, state layer, and output layer to detect and process multiple dimensions such as scene type, personnel status, command control, and sound field effects. Closed-loop control processing will be carried out on the actual control, dynamic jump, and command feedback of different scenes to finally determine the main control parameters and variables in the current scene in order to improve the control effect.

[0038] In this scenario, if any two control variables have a specific relationship, they will be divided according to the execution duration, execution time, and execution scenario to determine the relationship between each control variable.

[0039] Therefore, step S13 is implemented by: checking the frequent itemsets of any two control variables by rule matching; if the frequent itemsets meet the preset rules, combining the frequent itemsets in the form of three layers: input layer, state layer, and output layer to form a combined frequent itemset.

[0040] Based on the execution duration, execution time, and execution scenario corresponding to the combination of frequent itemsets, the control variables are divided, and the output of the divided control variables is used as the scenario mapping data for the corresponding scenario.

[0041] When identifying relationships between control variables, combinations that do not meet the requirements of the scenario need to be eliminated, such as data like "museum scenario + volume 80dB". Frequent itemsets can be used to find combinations of control variables that frequently appear together. These combinations represent strongly correlated variable groups. Frequent itemsets can be set with a support of 10%, meaning that the current control variable combination appears more than 10% of the total data in the historical data to obtain relatively frequent data. If you want to obtain as many frequent itemsets as possible, you can set its support to 5% to obtain enough data. The output frequent itemsets are represented as combinations of three levels of variables. First, find frequent itemsets at a single level, then frequent itemsets at two levels, and finally frequent itemsets at three levels, and output them as combined frequent itemsets. For example, input layer: 3 people + daytime, state layer: volume 50dB + omnidirectional pointing, output layer: frequent itemsets of the combination with clarity ≥ 89%.

[0042] The selected frequent itemsets must adhere to preset rules. These preset rules can be constructed using regular expressions based on historical data. Each frequent itemset is matched against these rules, and the matching itemsets are considered the available frequent itemsets. These preset rules can be pre-configured in the database for easy filtering of corresponding parameters. Itemsets are then grouped by execution duration, execution time, and execution scenario. For example, frequent itemsets in a meeting room scenario, weekday at 2 PM (execution time), and with an execution duration of 1-2 seconds can be grouped together to clarify the specific usage scenario and execution conditions for each group of variables. When constructing the state transition diagram, the changes in the audio system's state can be verified based on relevant data within this scenario. For instance, how changes in input layer variables affect the output layer's performance through state layer adjustments. For example, in the transition from A to B, adding 2 more people corresponds to a 5dB increase in volume, while maintaining 89% clarity, indicating that the variable adjustment is effective.

[0043] The preset rules can be set based on the following: Input layer: [2-5 people] + [Weekdays 9:00-18:00] → Status layer: [Volume 45-60dB] + [Omnidirectional / Conference table] → Output layer: [Clarity ≥88%]. In this way, data for the corresponding scenario is extracted from the corresponding data and a three-layer structure is formed.

[0044] At the same time, the state transition diagram can identify the parts where the output layer effect is not up to standard. For example, if the command volume is increased by 3dB and the output clarity changes from 85% to 84%, the state transition diagram can be used to quickly locate and determine the direction of the current adjustment.

[0045] Subsequent steps can directly use the probability values ​​of the number of times nodes appear and the number of times edges transition in the state transition graph to provide visualization results.

[0046] In one embodiment of the present invention, the state change of the speaker in step S2 specifically includes adjusting the volume, switching the sound effect mode, etc. The state changes of the speaker are used to connect the divided control variables and other contents to view the state transition process in the corresponding scenario.

[0047] like Figure 3 As shown, the implementation of step S2 includes: S21, when retrieving control commands, checking the control variables corresponding to each control command, and merging control variables with similar states based on the value range of the control variables.

[0048] Since step S1 determines the data granularity based on frequent itemsets, it is necessary to divide the data according to the range of the frequent itemsets. For example, the volume in a conference room is usually in the range of 30-60dB. At this time, it is divided into multiple data such as 30 / 35 / 40...60dB in 5dB intervals. These data are divided according to their value range to reduce the number of nodes when judging the current state transition. This prevents the transition relationship from becoming too complicated due to too many nodes, making it difficult to describe the current control command's regulation and processing in the corresponding scenario.

[0049] When merging control variables in similar states, the main focus is on merging control variables at the state layer. When the data at the input layer passes through relevant control commands, it is equivalent to data that has already been filtered under the same control commands. At the same time, the merged control variables must satisfy the same scenario and the same input characteristics to ensure that the merged data can maintain consistency at the input layer. The output layer data is then combined and described based on the combined data from the state layer to explain the range of values ​​of this merged data.

[0050] Preferably, step S21 is further implemented by: performing clustering processing on the volume, pointing and output index values ​​of the state layer according to the value range of the current control variable in the state layer and the output layer respectively, and using the average deviation of the corresponding clusters after clustering as the difference threshold of the current control variable.

[0051] The control variables are arranged according to their value range. When the consecutive control variables are not greater than the difference threshold, the difference threshold is used to merge the consecutive control variables.

[0052] The deviation threshold needs to be set according to the current control command scenario. Taking the conference room scenario as an example: you can select the control variables corresponding to volume and pointer, and then check the corresponding deviation thresholds for these control variables; the volume can be set to 2dB as the deviation threshold. When the volume difference is ≤2dB, the target area clarity fluctuation is ≤2%. These volumes can be merged as similar control variables.

[0053] Regarding the directional bias, a 3° deviation threshold can be selected. When the angle difference is ≤3°, the overlap rate of the sound coverage area in the target region is ≥95%, indicating no significant directional bias and consistent user perception. In this case, merging is possible. If the control variable corresponding to the sound effect mode is selected, since its value is mostly the index number of anti-echo or standard modes, the different values ​​differ greatly, making merging impossible.

[0054] At this point, the deviation threshold can be set based on the value of the same control command in the state layer. By clustering data, clusters with the same scene and the same input layer features can be locked, and the average deviation of these clusters can be regarded as the deviation threshold at this time. When calculating the deviation threshold, historical data can be introduced to obtain as many deviation values ​​as possible to limit the effectiveness of the current deviation threshold selection.

[0055] Clustering can be performed using the k-means algorithm. The number of clusters is set according to the inflection point rule. The volume and pointer values ​​in the state layer are clustered separately to obtain the difference threshold for the corresponding scenario. The number of clusters k ranges from 2 to 8. The sum of squared errors corresponding to each k value is calculated, and the k value with the steepest decrease rate of the sum of squared errors is selected. The maximum number of iterations is set to 100. Iteration stops when the change in cluster center is <0.1. Finally, the average deviation of the clusters is used as the difference threshold.

[0056] The clustering in the output layer is to group data under the same sound effect index to obtain sound effect index data within a relatively wide range. These data are considered to have index consistency, which facilitates subsequent processing of the state transition process. The clustering method of the output layer is the same as that of the state layer.

[0057] As for the dynamic changes of personnel covered in the state layer, they will be recorded based on the sound effect patterns that tend to follow after personnel position clustering. These dynamic changes of personnel position will not be handled in the way similar states are merged. Instead, the relevant personnel positions will be clustered by looking at the data after the control variables are merged, and the changes in personnel position will be used as trigger conditions to describe the changes in volume, direction and sound effect patterns in different scenarios.

[0058] S22, based on the merged control variables, using the merged control variables as state nodes, obtain the triggering conditions of the current state nodes, sort them according to the execution time of the triggering conditions, connect the state nodes that have changed sound state after the triggering, and obtain the output state transition diagram.

[0059] Each state node corresponds to a unified set of input layer variables, a specific range of volume + a specific range of pointing + sound effect mode, and collected sound effect indicators. When the position of the person changes, it is determined whether the change in position is a trigger condition. If it is a trigger condition, it means that the sound effect mode, volume, and pointing need to be adjusted to adapt to the position of the person in the corresponding scene, so that each scene can adapt to the sound field requirements. If the position of the person does not change, or the position of the person is suitable for the sound effect mode, then the state transition process needs to be recorded according to the changes in volume and pointing.

[0060] When obtaining the triggering condition of the current state node in step S22, the implementation method also includes: viewing the current gathering location of the people, and matching the historical data corresponding to the gathering location with the current sound effect mode. When the sound effect mode does not match the gathering location, the change in the current personnel position is regarded as the triggering condition.

[0061] If the sound effect mode matches the gathering location, the current volume and direction change will be the trigger condition.

[0062] When identifying trigger conditions, first check the current location of people gathering → match the current sound effect mode with the latest historical data → simultaneously check the environmental noise + real-time effect → if the mode does not match or the noise / effect does not meet the standard, then the change in the current location of people is regarded as the trigger condition, and the sound effect mode is adjusted.

[0063] The method for identifying where people gather can be the same as the k-means algorithm described above. In this case, the identified spatial locations are clustered by their spatial distance, and then it is checked whether the clustered locations are the default areas of the corresponding sound effect modes. For example, the closed small areas (such as corners) in the partitioned anti-echo mode, the open large areas (such as conference tables) in the enhanced voice mode, and the areas near the noise source (such as doorways) in the strong noise reduction standard mode. These sound effect modes are all commonly used in the corresponding scenarios. The location and sound effect data of these clustered locations are automatically recorded. If the current clustered area does not conform to the corresponding sound effect mode, or the noise, effect, etc. values ​​do not conform to the values ​​marked in the corresponding sound effect mode, then it is necessary to adjust the sound of the current scene to make the sound effect mode adapt to the scene.

[0064] The sound effect values ​​will vary depending on the sound effect mode and the scene. For example, the standard anti-echo mode has a speech intelligibility of ≥88%, referencing the ITU-TP.800 standard and set at the threshold for clear daily communication; echo attenuation is 20-25dB, based on the ANSI echo suppression standard, to avoid muddiness caused by wall reflections; and a volume range of 45-48dB, which represents the comfortable volume for human ears during communication; finally, the feedback rate is ≤1%. The standard anti-echo mode is often set to omnidirectional coverage, suitable for scenarios without fixed directionality, and is appropriate for people sitting in dispersed locations. If the current location of a person changes significantly, causing a change in the identified seating position, or if the sound effect measured at different locations is inconsistent, it indicates that adjustments to the volume and sound effect mode are needed to determine the current sound effect change. Other scenarios can be extracted from preset parts of the database, which will not be discussed further here.

[0065] For scenarios where the sound effect mode matches the gathering position, the volume and pointer will be recorded if they exceed the deviation threshold. That is, the deviation threshold used in step S21 will be called. Only when the change exceeds this threshold will the relevant change be recorded.

[0066] In this state transition process, taking a conference room scenario as an example, each state node clearly defines the current audio status.

[0067] For example, node A: [volume 45-50dB + omnidirectional + anti-echo mode]; node B: [volume 50-55dB + omnidirectional + anti-echo mode]; node C: [volume 50-55dB + pointing to the conference table + anti-echo mode]; node D: [volume 50-55dB + pointing to the conference table + enhanced speech mode]; at this time, each node has different values ​​for the control variables in the state layer, and these values ​​will show a gradual change process; nodes A, B, C, and D are all in the same scene in the same input layer, and the index values ​​of these data in the output layer are also in the same group, such as the speech intelligibility of nodes A, B, C, and D being in the range of 90%-95%.

[0068] In one embodiment of the present invention, when determining the personnel position under any state node in the state transition diagram, the probability value of the current state node + position cluster to the next state node is calculated, and each state node is bound to the personnel position data during the effective period of that node, so as to ensure that the subsequent calculation process can change according to the deduced triggering conditions.

[0069] like Figure 4As shown, the implementation of step S3 includes: S31, binding each state node with the location of the personnel, determining the location cluster corresponding to each state node, the location cluster in the number of times represents the clustering of personnel locations, such as the current scenario is a conference room scenario, which contains three location clusters, corresponding to the conference table cluster, corner cluster, and door cluster. These obtained location clusters will be labeled according to their location; the state nodes are grouped by the labels of the location clusters.

[0070] S32, extract the historical transition records from the current state node to the next state node, and associate each historical transition record with the corresponding location cluster.

[0071] S33 uses a combination of the current state node and the location cluster label to calculate the transition probability to the next state node.

[0072] The transition probability calculated at this time will take the combination of the current state node and the position cluster as a premise, and calculate the conditional probability of the next state node. That is, the ratio of the number of times the combination will transition to the next state node to the total number of transitions of the combination, which describes the probability value when the current state node transitions.

[0073] In one embodiment of the present invention, when retrieving scene mapping data, in addition to identifying the location and direction of each speaker, it is also necessary to determine the plane corresponding to each speaker and the sound volume at each location when each speaker is distributed and controlled, so that the currently emitted sound can be adapted to the corresponding scene, and when the control command is finally issued, it can meet the needs of the current scene, so that the sound configured by the speakers conforms to the processing logic.

[0074] like Figure 5 As shown, the implementation of step S4 includes: S41, based on the location cluster where each person's location is located, the coverage range of the sound system is collected, a mapping relationship between the placement location of the sound system and the location of the person is established, and a sound system-location cluster mapping table is generated.

[0075] When generating the speaker-location cluster mapping table, the physical information of the speakers is first obtained. Taking a conference room as an example, the installation coordinates, pointing type (omnidirectional / directional), and coverage range of all speakers in the conference room are recorded (e.g., directional speakers have a maximum coverage angle of 60° and an effective coverage distance of 5 meters).

[0076] Then, the sound coverage area is synchronized to the specific spatial coordinates, and the location clusters and corresponding coordinates obtained in step S3 are introduced, such as conference table cluster C1: X∈[1.0, 2.0], Y∈[3.0, 4.0]; corner cluster C2: X∈[0.5, 1.0], Y∈[1.0, 1.5].

[0077] Calculate the extent to which the current sound covers these coordinate ranges, that is, the overlap rate between the sound coverage range and the location cluster coordinates, and determine whether the sound covers the corresponding location. If the overlap rate is ≥70%, it is considered as primary coverage; if the overlap rate is 30%-70%, it is considered as secondary coverage; if the overlap rate is less than 30%, it is considered as supplementary coverage, used to fine-tune the sound field of the current scene. After labeling these data, a sound-location cluster mapping table is formed.

[0078] The overlap rate is calculated using the area overlap rate, which is calculated as: Area overlap = (intersection area of ​​the sound coverage area and the location cluster) / (total area of ​​the location cluster) × 100%. Here, the directional sound is a sector area with the installation coordinates as the vertex, the coverage angle as the apex, and the effective distance as the side length, while the omnidirectional sound is a circular area with the installation coordinates as the center and the effective distance as the radius. The area of ​​the location cluster is the polygonal area formed by the boundary coordinates of the location cluster.

[0079] For example, the main speaker S1 (coordinates 1.5, 3.0, directional, coverage angle 60°): its coverage overlaps with C1 (conference table cluster) by 90% → S1 mainly covers C1.

[0080] Auxiliary speaker S2 (coordinates 0.8, 1.2, directional, coverage angle 45°): coverage area overlaps with C2 (corner cluster) by 85% → S2 mainly covers C2.

[0081] S42, based on the relevant data of the transition probability, screen the feasibility of executing control commands for any element in the speaker-position cluster mapping table, and obtain the optimal control direction.

[0082] When obtaining the optimal control direction, it is necessary to determine the optimal transition probability of each state node in the state transition diagram. When the optimal state transition is met, check its corresponding parameters. For example, if the direction is adjusted to 25°, the parameters corresponding to the transition probability should meet the settings of the audio hardware. When they meet the requirements, it means that the parameter value corresponding to the current control command is a feasible operation, and the corresponding parameter is regarded as the current optimal control direction.

[0083] When determining that the transition probability is optimal and feasible, the transition probability is calculated mainly by the state transition diagram, the Boolean value of the device feasibility, and the probability of frequent user adjustments after setting the transition probability parameters. The combined probability value of these three factors is regarded as the current optimal transition probability, and the parameter corresponding to the maximum optimal transition probability value is selected as the current optimal control direction.

[0084] It should be noted that the Boolean value represents the values ​​of 0 and 1. When the corresponding volume and direction can be adjusted on the current device, it is regarded as 1, otherwise it is 0. As for the probability of frequent user adjustments, it is set based on the ratio of the number of times the user manually adjusts the current parameters after the settings in historical data to the total number of times. At the same time, in order to filter out the parts that the user rarely jumps, it is necessary to subtract the probability value from 1 to obtain the probability value of user adjustment.

[0085] Therefore, when obtaining the optimal control direction, the implementation method includes: comparing the volume and direction corresponding to the current transfer probability with the device parameters of the audio system; when the adjustment of volume and direction is executable, the number of times the user manually adjusts after the current parameter setting is introduced, and the probability of user adjustment is calculated; the maximum value of the product of the user adjustment probability and the transfer probability is regarded as the optimal transfer probability; the parameter corresponding to the optimal transfer probability is used as the optimal control direction of the output.

[0086] S43 outputs the audio control result using the parameter value corresponding to the optimal control direction. The output audio control result represents the adjustment methods for volume, direction, and other data that can be executed in the current scene.

[0087] If there are multiple speakers in the conference room, control command conflicts may occur, resulting in an unbalanced overall sound field. These issues need to be addressed by prioritizing speakers and ensuring sound field consistency.

[0088] Priority represents the primary coverage, secondary coverage, and supplementary coverage marked according to the speaker coverage range; control commands are executed in this order, while checking the consistency of multiple speaker parameters, that is, verifying that the volume difference is ≤5dB and there is no overlapping interference in the direction; non-conflicting parameters are output to the relevant speakers in the optimal control direction to ensure the overall sound field balance during subsequent speaker control.

[0089] In step S5, the order in which the audio control results are actually distributed to the device is checked, and it is marked as the control sequence of the corresponding state node. The implementation of the overall control strategy is explained according to the duration of each execution, so that the sound field always meets the current usage requirements no matter how people move or how the environment changes.

[0090] The implementation of step S5 includes: extracting the average execution time of each parameter in the audio control result, synchronizing the average execution time to the corresponding state node, and using it as the output control time; the average execution time is obtained by dividing the time of a single effective adjustment in the historical data by the number of effective records, interpreting the effective situation under multiple adjustments of the current audio system, and when this average execution time is obtained, the control instructions under the same conditions can be executed by selecting this time length.

[0091] Obtain the control commands for each adjustment and synchronize the control commands as updated data to the database.

[0092] When synchronizing each control command, you can choose to accumulate 100 control commands as a cycle to record the execution time and control sequence of the corresponding control commands, and at the same time synchronize the adjustment status of each audio device to facilitate subsequent data optimization and traceability.

[0093] The control strategy for each location is based on the speaker-location cluster mapping table and the optimal control direction. The speaker control results are mapped to the corresponding speaker location and the location of relevant personnel in the speaker-location cluster mapping table, and the corresponding optimal control method is invoked. This prioritizes the execution of the main coverage cluster, followed by the auxiliary coverage cluster, and finally the supplementary coverage cluster. The control strategy for each location is then determined by combining the control duration. Examples are as follows: 1. Main coverage cluster (e.g., conference room table cluster C1): Optimal control direction is prioritized, volume 50-55dB, pointing towards the conference table, control duration is averaged at 1.5 seconds; 2. Auxiliary coverage cluster (e.g., conference room corner cluster C2): Synchronously adapts to the main coverage cluster parameters, volume 45-50dB, pointing slightly adjusted to the corner without overlapping interference with the main coverage cluster, control duration 1.2 seconds; 3. Supplementary coverage cluster (e.g., conference room door cluster C3): Adjustment is triggered only when noise > 50dB, volume 40-45dB, pointing away from the door, control duration 2 seconds. After all strategies are executed, control commands are synchronized to the database, updated once every 100 commands.

[0094] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention, which are still covered within the protection scope of the present invention.

Claims

1. A control method for an intelligent audio system, characterized in that, include: S1 acquires the playback signal from the speaker, and based on the environmental parameters, control commands, and real-time operating parameters of the speaker collected in the current scene, it extracts the control variables in each layer of the structure in a hierarchical manner, with a three-layer structure of input layer, state layer, and output layer, and uses the control variables to map with the corresponding scene to establish scene mapping data; The input layer consists of the control command type, the current scene, and the number of people; the real-time parameters of the audio system and the dynamic state of the environment; and the output layer consists of the sound effect indicators for each location. S2, retrieve the execution process of each control command in the scene mapping data. The execution process includes the execution duration, execution time, and execution scene. Connect the execution processes according to the changes in the sound state to obtain the state transition diagram of the current scene. S3, confirm the personnel position under any state node in the state transition diagram, perform spatial clustering based on the personnel's position, configure the position cluster of the personnel's position, and calculate the transition probability under the corresponding position cluster; S4. The transition probability is correlated with the placement position of each speaker to determine the speaker control result. The implementation of step S4 includes: S41, based on the location clusters where each person's position is located, the coverage range of the speakers is collected, a mapping relationship between the placement position of the speakers and the positions of the people is established, and a speaker-location cluster mapping table is generated; S42, based on the relevant data of the transition probability, the feasibility of executing control commands for any element in the speaker-location cluster mapping table is screened, and the optimal control direction is obtained; S43, the parameter value corresponding to the optimal control direction is used as the output speaker control result. S5 converts the audio control results into the control sequence of each state node in the state transition diagram, and determines the control strategy for each position according to the control duration of each control.

2. The control method for an intelligent audio system according to claim 1, characterized in that, The implementation of step S1 also includes: S11: Collect environmental parameters, control commands, and real-time audio operating parameters of the current scene as the initial dataset; S12, establish a three-layer structure of input, state and output for the initial dataset, extract the control variables for each audio control, and establish a control network structure related to the three layers; S13. Based on the correlation between various control variables in the control network structure, the control network structure is mapped to the specific scenario to generate scenario mapping data.

3. The control method for an intelligent audio system according to claim 2, characterized in that, The implementation methods of step S13 include: The frequent itemsets of any two control variables are examined by rule matching. If the frequent itemsets meet the preset rules, they are combined into a combined frequent itemset in the form of three layers: input layer, state layer, and output layer. Based on the execution duration, execution time, and execution scenario corresponding to the combination of frequent itemsets, the control variables are divided, and the output of the divided control variables is used as the scenario mapping data for the corresponding scenario.

4. The control method for an intelligent audio system according to claim 1, characterized in that, Step S2 can be implemented in the following ways: S21, when retrieving control commands, check the control variables corresponding to each control command, and merge control variables with similar states based on the value range of the control variables; S22, based on the merged control variables, using the merged control variables as state nodes, obtain the triggering conditions of the current state nodes, sort them according to the execution time of the triggering conditions, connect the state nodes that have changed sound state after the triggering, and obtain the output state transition diagram.

5. The control method for an intelligent audio system according to claim 4, characterized in that, The implementation of step S21 also includes: Based on the current value range of the control variable in the state layer and the output layer, the volume, pointer and output index values ​​of the state layer are clustered respectively, and the average deviation of the corresponding clusters after clustering is used as the difference threshold of the current control variable. The control variables are arranged according to their value range. When the consecutive control variables are not greater than the difference threshold, the difference threshold is used to merge the consecutive control variables.

6. The control method for an intelligent audio system according to claim 4, characterized in that, When obtaining the triggering condition for the current state node in step S22, the implementation method also includes: View the current gathering location of people and match the historical data corresponding to the gathering location with the current sound effect mode. When the sound effect mode and the gathering location do not match, the change of the current personnel location is regarded as the trigger condition. If the sound effect mode matches the gathering location, the current volume and direction change will be the trigger condition.

7. The control method for an intelligent audio system according to claim 1, characterized in that, Step S3 can be implemented in the following ways: S31, bind each status node to the personnel location, determine the location cluster corresponding to each status node, and group the status nodes by the label of the location cluster; S32, extract the historical transition records from the current state node to the next state node, and associate the location cluster corresponding to each historical transition record; S33 uses a combination of the current state node and the location cluster label to calculate the transition probability to the next state node.

8. The control method for an intelligent audio system according to claim 1, characterized in that, The methods for obtaining the optimal control direction include: The current transfer probability is compared with the speaker's device parameters for volume and direction. When volume and direction adjustment is feasible, the number of times the user manually adjusts the current parameters is set is taken into account, and the probability of user adjustment is calculated. The maximum value of the product of the user adjustment probability and the transfer probability is taken as the optimal transfer probability. The parameter corresponding to the optimal transfer probability is taken as the optimal control direction of the output.

9. The control method for an intelligent audio system according to claim 1, characterized in that, Step S5 can be implemented in the following ways: Extract the average execution time of each parameter in the audio control results, synchronize the average execution time to the corresponding state node, and use it as the output control time; Obtain the control commands for each adjustment and synchronize the control commands as updated data to the database.

Images