Self-adaptive acoustic analysis method and system

Abstract

The invention relates to the technical field of acoustic analysis, in particular to a self-adaptive acoustic analysis method and system, and the method comprises the following steps: obtaining a sound pressure amplitude screening channel, extracting a main sound point time index, judging a fluctuation direction to generate a path identifier, and completing channel alignment and path binding to generate an analysis result. According to the invention, by acquiring the maximum sound pressure amplitude of each channel and dynamically screening the effective channel, the capability of capturing significant sound in a complex sound source environment is improved, and by combining directional extraction and fluctuation trend judgment of main sound point time index differences in a continuous period, dynamic perception and track identification of sound source feature changes are realized. The stability of path identification is enhanced by using direction consistency and time sequence alignment operation, accurate binding and real-time adaptive sound source tracking of multiple channels are completed, the identification accuracy and path continuity in a sound source dynamic change scene are effectively improved, meanwhile, redundant channel interference is avoided, and the accuracy of path identification is improved. And the processing efficiency and the adaptive capacity of the acoustic model are enhanced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of acoustic analysis, in particular to an adaptive acoustic analysis method and system. BACKGROUND

[0002] The technical field of acoustic analysis mainly involves the core matters of acquisition, analysis, identification and parameter extraction of sound signals, and is widely used in scenes such as speech recognition, environmental sound detection, voiceprint recognition and fault sound diagnosis. Usually, the collected acoustic signals are identified and analyzed through methods such as establishing a sound data model, analyzing sound characteristic parameters, matching existing sound patterns and classifying and identifying. Among them, the traditional adaptive acoustic analysis method refers to an analysis method that adjusts model parameters to adapt to the change of sound source characteristics when facing different acoustic environments or changes of sound sources. It usually uses the method of extracting acoustic characteristics based on mel-frequency cepstral coefficients, training sound categories using Gaussian mixture model, and dynamically adjusting model parameters according to real-time sampling values to improve the accuracy of sound feature recognition in variable acoustic scenes.

[0003] Traditional acoustic analysis relies on fixed feature parameters such as mel-frequency cepstral coefficients and Gaussian mixture model training, which leads to the problem that feature extraction cannot timely reflect the dynamic changes of acoustic scenes in the case of changes in sound source position or inter-channel signal interference. Although the model parameters can be adaptively adjusted, they lack tracking of the time change direction of the main response point, making it difficult to accurately locate the actual sound occurrence path. Especially in the case of multiple sound source overlap or severe environmental noise fluctuation, the model may cause mismatch or path jump, thereby causing classification errors or recognition lag. For example, a valid sound source channel may not be fully recognized due to signal instantaneous attenuation, which may lead to inaccurate judgment of the entire path and affect the overall analysis accuracy. SUMMARY

[0004] The purpose of the present application is to solve the problems existing in the prior art and to provide an adaptive acoustic analysis method.

[0005] In order to achieve the above purpose, the present application adopts the following technical scheme: an adaptive acoustic analysis method, comprising the following steps:

[0006] S1: obtaining the maximum sound pressure amplitude of each sound source channel in each period of the acoustic acquisition stage, screening the channel number whose sound pressure amplitude exceeds the threshold value of the sound pressure level, and enabling the adjacent channel number to generate a set of available sound source channels;

[0007] S2: based on the set of available sound source channels, obtaining the time index position corresponding to the maximum sound pressure value in each channel of the current period, extracting the main response point time index of the current period, and combining the main response point time index of the previous period to determine the change direction of the time difference, and generating a main response point time fluctuation sequence;

[0008] S3: according to the current main response point time difference direction mark in the main response point time fluctuation sequence and the last two periods, combining the perception time difference to judge whether the main response point difference direction is continuous and consistent, synchronously determining the current classification path, generating a period path switching mark sequence;

[0009] S4: based on the period path switching mark sequence, matching all the channel numbers currently in the enabled state, performing main response point time sequence alignment, screening the direction consistent channel numbers, generating a path binding channel list;

[0010] S5: according to the path binding channel list, detecting whether the state between the current path mark and the last period path mark is consistent, and recording the current path and channel number, generating an adaptive acoustic analysis result.

[0011] As a further scheme of the application, the available sound source channel set includes a channel number set, an adjacency relationship judgment result, and a sound pressure amplitude screening record, the main response point time fluctuation sequence includes a main response point index difference value, a direction change mark, and a fluctuation trend label, the period path switching mark sequence includes a path number mapping, a direction continuity mark, and a difference direction judgment result, the path binding channel list includes a direction consistency channel number, a main response point time alignment information, and a current path corresponding channel mapping, and the adaptive acoustic analysis result includes a path state record, a channel binding relationship, and a path continuity mark.

[0012] As a further scheme of the application, the acquisition step of the available sound source channel set is specifically:

[0013] S111: based on the acoustic acquisition data of each sound source channel in each period, acquiring the maximum sound pressure amplitude value of the channel, constructing a mapping set of the channel number and the corresponding maximum sound pressure amplitude, combining the corresponding channel sound pressure level threshold in the initialization configuration, comparing the maximum sound pressure amplitude of each channel with the corresponding threshold by using a comparison operation, and generating a sound pressure abnormal channel number set;

[0014] S112: based on the sound pressure abnormal channel number set, judging whether each abnormal channel has an adjacent channel number directly connected thereto according to the adjacency matrix structure information between the sound source channels, extracting all the channel numbers adjacent to the abnormal channel numbers, and merging to generate an expanded channel number set;

[0015] S113: according to the expanded channel number set, combining the sound pressure amplitude characteristics of each channel in the current period and the correlation relationship of the adjacent channels, identifying and screening the use validity of each channel number, judging whether the sound source channel enabling condition is met, obtaining a channel number set meeting the availability standard, and generating an available sound source channel set.

[0016] As a further scheme of the present application, the step of obtaining the main sound point time fluctuation sequence is specifically:

[0017] S211: Based on the sound pressure sequence data of each channel in the set of available sound source channels, the time index corresponding to the maximum value in the sound pressure sequence of each channel in the current period is extracted, the index position of the corresponding value in the time sequence is determined by locating the maximum value point, and the channel number and the corresponding time index are recorded to generate a set of channel maximum value time indexes;

[0018] S212: Based on the set of channel maximum value time indexes, the maximum sound pressure value index is used to filter the maximum sound point time index of the channel, and the current main sound point time index value is obtained by calculation;

[0019] S213: According to the current main sound point time index value and the main sound point time index value extracted in the last period, the index value is compared and the difference direction of the index time is judged, the continuous period main sound point index sequence is recorded and updated, and the main sound point time fluctuation sequence is generated.

[0020] As a further scheme of the present application, the step of obtaining the period path switching identification sequence is specifically:

[0021] S311: Based on the main sound point time index recorded in the main sound point time fluctuation sequence in the continuous three periods, the time index positions of the current period and the last two periods are extracted in turn, the three groups of index positions are arranged in time sequence, the change direction of the time index between the previous and subsequent periods is compared, and the continuous direction mark is formed by judging whether the change trend is consistent, and the main sound time offset direction sequence is obtained;

[0022] S312: According to the main sound time offset direction sequence, the time interval between the continuous two periods is detected whether it is within the range combined with the perception time difference, and it is confirmed whether the change direction is consistent, the period which meets the consistent direction is marked with the trend, and the trend classification intensity value sequence is obtained;

[0023] S313: According to the trend classification intensity value sequence and the main sound time offset direction sequence, the direction consistency and trend classification consistency in three periods are compared, whether the current period belongs to the continuous path state or the trend switching state with inconsistent direction is screened, and the path identification content of the current period is updated to generate the period path switching identification sequence.

[0024] As a further scheme of the present application, the step of obtaining the path binding channel list is specifically:

[0025] S411: Based on the path number corresponding to the periodic path switching identification sequence, the path number set in the enabled state is extracted in combination with the path state record in each current period, and the path number is matched with all available channel numbers in the current period, all channel numbers associated with each path are summarized, and a path number associated channel sequence is obtained;

[0026] S412: According to the path number associated channel sequence, the main response point time index record of all associated channels in the current period is extracted, all channel numbers under the path number are arranged in time sequence, the change trend of the main response point index is identified, and the channel historical direction marker information is screened in combination to judge whether the main response direction of the channels under the same path is consistent, the channel numbers with consistent direction are retained, a channel direction matching sequence is obtained;

[0027] S413: Based on the channel direction matching sequence, the channel numbers with consistent direction are aggregated according to the path number, and the screened channel number is recorded in the path corresponding table item, forming the consistency binding mapping of the path number and the channel number, the binding result is structured and arranged to remove the repeated identification, the channel set is output in turn according to the path number, and the path binding channel list is generated.

[0028] As a further scheme of the application, the adaptive acoustic analysis result obtaining step is specifically:

[0029] S511: Based on the channel number listed in the path binding channel list, the channel binding relationship of each path in the current period is read, the path state of each path number in the current period is extracted, and the state of the same path number in the previous period is synchronously searched, the state consistent or inconsistent result of the path number is recorded in sequence, and a path state consistency index sequence is generated;

[0030] S512: According to the path state consistency index sequence, the path number set with consistent state is identified, and the corresponding channel number is read, the synchronization deviation between the main response point time and the path main response point index is compared, whether the main response point time of each channel is in the synchronization range is evaluated, the effective channel list is recorded according to the path number, and a path synchronization channel set is obtained;

[0031] S513: According to the path synchronization channel set, the running state of each path and the bound channel in the current period is read, the path state consistency and channel synchronization data are combined, and the adaptive acoustic analysis result is generated by recording and summarizing according to the path number as the index.

[0032] An adaptive acoustic analysis system, comprising:

[0033] The channel screening mechanism module is used for performing S1: obtaining the maximum sound pressure amplitude of each sound source channel in each period of the acoustic acquisition stage, screening the channel number whose sound pressure amplitude exceeds the sound pressure level threshold, enabling the adjacent channel number, and generating a set of available sound source channels;

[0034] The main response feature extraction module is used for performing S2: based on the set of available sound source channels, obtaining the time index position corresponding to the sound pressure maximum value in each channel of the current period, extracting the main response point time index of the current period, combining the main response point time index of the last period, judging the change direction of the time difference, and generating a main response point time fluctuation sequence;

[0035] The path consistency identification module is used for performing S3: according to the main response point time difference direction mark of the current and last two periods in the main response point time fluctuation sequence, combining the perception time difference to judge whether the main response point difference direction is continuous and consistent, synchronously determining the current classification path, and generating a period path switching identification sequence;

[0036] The path channel binding module is used for performing S4: based on the period path switching identification sequence, matching all channel numbers currently in the enabled state, aligning the main response point time sequence, screening the direction consistent channel number, and generating a path binding channel list;

[0037] The analysis result confirmation module is used for performing S5: according to the path binding channel list, detecting whether the state between the current path identification and the last period path identification is consistent, recording the current path and channel number, and generating an adaptive acoustic analysis result.

[0038] Compared with the prior art, the advantages and positive effects of the present application are that:

[0039] In the present application, by obtaining the maximum sound pressure amplitude of each channel and dynamically screening the effective channel, the capture ability of the significant sound in the complex sound source environment is improved, the directionality extraction and fluctuation trend judgment of the main response point time index difference in the continuous period are combined, the dynamic perception and trajectory recognition of the sound source feature change are realized, the stability of the path recognition is strengthened by using the direction consistency and time sequence alignment operation, the precise binding of multiple channels and the real-time adaptive sound source tracking are completed on the basis of the periodic state recording and path consistency judgment, the recognition accuracy and path continuity in the dynamic change scene of the sound source are effectively improved, the redundant channel interference is avoided, and the processing efficiency and the adaptability of the acoustic model are enhanced. BRIEF DESCRIPTION OF DRAWINGS

[0040] Figure 1 The main step flowchart of the present application is shown in the figure;

[0041] Figure 2 The available sound source channel set acquisition flowchart of the present application is shown in the figure;

[0042] Figure 3 Flow chart for acquiring main response point time fluctuation sequence of the application;

[0043] Figure 4 Flow chart for acquiring cycle path switching identification sequence of the application;

[0044] Figure 5 Flow chart for acquiring path binding channel list of the application;

[0045] Figure 6 Flow chart for acquiring adaptive acoustic analysis result of the application. DETAILED DESCRIPTION

[0046] In order to make the objects, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and do not limit the present application.

[0047] In the description of the present application, it should be understood that the terms "length", "width", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer" and the like indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, and are only used to facilitate the description of the present application and simplify the description, and do not indicate or imply that the device or element referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as a limitation of the present application. In addition, in the description of the present application, the meaning of "a plurality of" is two or more, unless otherwise explicitly and specifically limited.

[0048] Please refer to Figure 1 An adaptive acoustic analysis method, comprising the following steps:

[0049] S1: Obtain the maximum sound pressure amplitude of each sound source channel in each cycle of the acoustic acquisition stage, combine the sound pressure level threshold in the initialization configuration (in line with the environmental noise standard ≥ 65dB(A)), compare the sound pressure amplitude of each channel with the corresponding threshold size in turn, screen the channel number whose sound pressure amplitude exceeds the sound pressure level threshold, and enable the adjacent channel number according to the channel adjacency relationship, to generate a set of available sound source channels;

[0050] S2: Based on the sound pressure sequence data corresponding to the channels in the set of available sound source channels, obtain the time index position corresponding to the maximum sound pressure value in each channel in the current cycle, extract the main response point time index of the current cycle, and combine the main response point time index of the previous cycle to determine the direction of change of the time difference, to generate a main response point time fluctuation sequence;

[0051] S3: According to the current and the last two periods of the main response point time difference direction mark in the main response point time fluctuation sequence, combined with the perception time difference (reference standard set ± 20ms), judge whether the main response point difference direction is continuous and consistent, if the direction does not change for three consecutive periods, mark the current path as a keep state, if there is a same direction deviation for two periods, select the stable direction trend in the past path as the current classification path, generate a period path switching identification sequence;

[0052] S4: Based on the path number corresponding to the period path switching identification sequence, match all the current channel numbers in the enabled state, align the main response point time sequence, select the direction consistent channel number as the current path binding channel set, and generate a path binding channel list;

[0053] S5: According to the channel number in the path binding channel list, detect whether the state between the current path identification and the last period path identification is consistent, and record the current path and the bound channel number, and generate an adaptive acoustic analysis result.

[0054] The available sound source channel set includes channel number set, adjacency relationship judgment result, sound pressure amplitude screening record, the main response point time fluctuation sequence includes main response point index difference, direction change identification, fluctuation trend label, the period path switching identification sequence includes path number mapping, direction continuity mark, difference direction judgment result, the path binding channel list includes direction consistency channel number, main response point time alignment information, current path corresponding channel mapping, the adaptive acoustic analysis result includes path state record, channel binding relationship, path continuity identification.

[0055] Please refer to Figure 2 , S1 step is:

[0056] S111: Based on the acoustic acquisition data of each sound source channel in each period, the maximum sound pressure amplitude value of the channel is obtained, the channel number and the corresponding maximum sound pressure amplitude are constructed into a mapping set, combined with the corresponding channel sound pressure level threshold in the initialization configuration, the comparison operation is used to compare the maximum sound pressure amplitude of each channel with the corresponding threshold, and the sound pressure abnormal channel number set is generated;

[0057] Based on the acoustic acquisition data of each sound source channel in each cycle, sound pressure level amplitude information is extracted. This requires reading the raw acoustic signal data for each channel within the cycle, using an acquisition device with a sampling frequency of 48kHz to digitize the sound waves, and extracting the time-domain sound pressure signal curve for each channel within a specified sampling period. The maximum sound pressure level amplitude is then obtained using a maximum value extraction method. For example, in cycle 1, the maximum sound pressure level amplitude of channel CH01 is 64.7dB, channel CH02 is 69.3dB, channel CH03 is 62.5dB, channel CH04 is 70.6dB, and channel CH05 is 74.3dB. To obtain a clear mapping relationship, an amplitude dataset indexed by channel number needs to be constructed. Simultaneously, the channel threshold parameters set in the initialization configuration need to be introduced. The initialization configuration sets the channel weighted sound pressure level threshold based on environmental noise standards; ordinary channels are set to 65.0dB, and key area channels are set as follows:

[0058] CH04 and CH05 were set to 70.0 dB. The sound pressure levels of each channel were compared to the threshold value according to their channel number to determine if they constituted an abnormal state. CH01, at 64.7 dB, was less than the threshold of 65.0 dB and was therefore excluded. CH02, at 69.3 dB, was greater than the threshold of 65.0 dB and was therefore identified as an abnormal channel. CH04, at 70.6 dB, exceeded the threshold of 70.0 dB and was therefore identified as an abnormal channel. CH05, at 74.3 dB, was also identified as an abnormal channel. After comparing each channel, the set of channel numbers constituting an abnormality was extracted as follows:

[0059] Table 1. Comparison of maximum sound pressure level and threshold for each channel.

[0060] Channel number Maximum sound pressure amplitude (dB) Weighted sound pressure level threshold (dB) Is abnormal CH01 64.7 65.0 No CH02 69.3 65.0 Yes CH03 62.5 65.0 No CH04 70.6 70.0 Yes CH05 74.3 70.0 Yes

[0061] As shown in Table 1, by comparing the maximum sound pressure amplitude with the set threshold item by item, the abnormal channel number set is obtained as {CH02, CH04, CH05}, which is the sound pressure abnormal channel number set.

[0062] S112: Based on the sound pressure anomaly channel number set, according to the adjacency matrix structure information between sound source channels, determine whether each anomaly channel has an adjacent channel number directly connected to it, extract all channel numbers adjacent to the anomaly channel number, and merge them to generate an extended channel number set;

[0063] According to the sound pressure anomaly channel number set {CH02, CH04, CH05} obtained as described above, the channel nodes having associated connections for each anomaly channel are sequentially searched in combination with the channel adjacency relationship matrix configured in the sound source system. For example, the adjacency relationship is stored in a matrix form. If CH02 has adjacency relationship with CH03 and CH05, the positions of CH03 and CH05 in the CH02 row in the matrix are set to 1, and the values of 0 are regarded as no adjacency. After traversing CH02, it is obtained that the adjacent channels of CH02 are CH03 and CH05, the adjacent channels of CH04 are CH06 and CH08, and the adjacent channels of CH05 are CH02 and CH06. In this way, all the adjacent channel numbers are summarized as {CH03, CH05, CH06, CH08, CH02}, which are merged into the original anomaly number set, and after deduplication, the expanded channel number set is {CH02, CH03, CH04, CH05, CH06, CH08}. The expanded set expands the subsequent analysis area to all signal path channels in the adjacent area, and builds a basis for subsequent availability screening. The final output result is the expanded channel number set.

[0064] S113: According to the expanded channel number set, the use effectiveness of each channel is identified and screened in combination with the sound pressure amplitude characteristics of each channel in the current period and the correlation of the adjacent channels, whether the sound source channel enabling condition is satisfied is judged, a channel number set satisfying the availability standard is obtained, and a set of available sound source channels is generated.

[0065] According to the expanded channel number set {CH02, CH03, CH04, CH05, CH06, CH08}, the maximum sound pressure amplitude of each channel in the current period is extracted again, and the corresponding sound pressure level threshold is combined to identify whether it satisfies the availability enabling requirement. Meanwhile, the adjacency structure characteristics are introduced, the adjacency relationship strength square sum value of each channel, the sound source coupling degree array between the channel and the adjacent channel, and the response coefficient array of the corresponding period are extracted. For example, for channel CH05, the extraction results are as follows:

[0066] Table 2 Channel CH05 Association Parameter Table

[0067]

[0068] As shown in Table 2, the CH05 channel has a sound pressure characteristic higher than the threshold value in the current period, the adjacent relationship strength square sum is 4.5, indicating that the connection is high, and the corresponding channel coupling degree and response coefficient are 3 groups of values, respectively quantifying the signal synchronization and intensity response ability between the channel and the adjacent node. Based on the set of such parameters, it is judged whether it reaches the enabled state by passing into the availability identification condition, and the same feature extraction and judgment operation is performed on other channels in the extended number set such as CH03, CH04, etc. Finally, all channel number sets that meet the enablement judgment condition are summarized as the output available sound source channel set.

[0069] Please refer to Figure 3 , S2 step is:

[0070] S211: Based on the sound pressure sequence data of each channel in the available sound source channel set, the time index corresponding to the maximum value in the sound pressure sequence of each channel in the current period is extracted, the index position of the corresponding value in the time sequence is determined by locating the maximum value point, and the channel number and the corresponding time index are recorded to generate the channel extreme value time index set;

[0071] To obtain the sound pressure sequence data corresponding to each channel in the available sound source channel set, the time sequence signal data collected in the acoustic measurement system needs to be extracted in sequence according to the period. Each channel collects sound pressure data at a sampling rate of 48 kHz to form an equally spaced sample set. For channel CH05, the sound pressure data in the period can be represented as the sequence {71.2, 72.5, 73.4, …, 74.3, …}. The system finds the maximum sound pressure amplitude by traversing the sequence and identifying the maximum value through the traversal extreme value process

[0072] 74.3 dB and locates its time index position as sample number 2201, which is the main response time position of the current channel CH05 in the period. All available channels are processed in the same logic and their channel numbers and maximum value index positions are recorded to form a channel number and time index mapping structure, such as {CH01: 2158, CH02: 2190, CH03: 2133, CH05: 2201}. The system records the mapping and stores it in the data buffer module. The average time consumption for each channel extraction is 0.002 seconds, and the total extraction time is less than 0.25 seconds. The following parameter table is established:

[0073] Table 3 Sound pressure extreme value time index extraction table

[0074]

[0075]

[0076] As shown in Table 3, the channel extreme value time index set is obtained by traversing and indexing the maximum sound pressure value of each channel in the current period.

[0077] S212: Based on the channel extreme value time index set, the extreme value time point of the channel where the maximum sound pressure value index is located is screened to extract the periodic main response point time index, using the formula:

[0078]

[0079] The current main response point time index value is obtained by operation; wherein, The time index position of the main response point in the i-th period is represented, and the unit is a sampling point, S ij The maximum sound pressure amplitude value of the j-th channel in the i-th period is represented, and the unit is dB, The average value of the maximum sound pressure values of all available channels in period i is represented, and the unit is dB, ΔT ij The extreme point time offset of the j-th channel in period i is represented, and the unit is a sampling point, defined as The maximum value of all channel time offsets in period i is represented, and the unit is a sampling point, which is used for normalization to eliminate unit influence, D ijk The maximum sound pressure difference value of the j-th channel and its adjacent channel k in period i is represented, and the unit is dB, μ ik The average value of the sound pressure difference values of all adjacent channels of channel j in period i is represented, and the unit is dB, The cumulative value of all adjacent channel numbers k of channel j is represented.

[0080] After calling the channel extreme value time index set, a comprehensive judgment index needs to be constructed based on multiple associated factors to screen out the main response point time index of the current period. First, the maximum sound pressure value of channel CH05 is 74.3 dB, the average maximum sound pressure value in the period is 69.5 dB, and the sound pressure difference is 74.3-69.5=4.8 dB. At the same time, the extreme value time index of the channel in the last period is called from the data recording module, which is 2180 points, and the current index is 2201 points, with an offset ΔT ij =21 points, and the maximum offset in this period is After normalization After squaring, it is 0.36. Then the adjacent channel sound pressure difference values of CH05 are extracted as 5.2, 4.1, and 3.7 dB, and the average disturbance μ ik =4.33 dB, and the deviation square of each item is calculated as: (5.2-4.33) 2 =0.7569, (4.1-4.33) 2 =0.0529, (3.7-4.33) 2 =0.3969, and the sum of the disturbance items is 1.2067. The total denominator square value obtained by combining the normalized time item and the disturbance item is 1.5667, and the square root is 1.2517. Finally, the numerical calculation is as follows:

[0081] Molecule = |74.3-69.5| = 4.8;

[0082] Normalized time term square = (21 / 35) 2 = 0.36;

[0083] Disturbance term square sum = (5.2-4.33) 2 + (4.1-4.33) 2 + (3.7-4.33) 2 = 1.2067;

[0084] Denominator square sum = 0.36 + 1.2067 = 1.5667;

[0085]

[0086] Expression value = 4.8 / 1.2517 ≈ 3.835;

[0087] Table 4 main response point extraction key parameter table

[0088]

[0089] As shown in Table 4, CH05 obtains the maximum normalized index value 3.835 in the expression calculation, and after the same operation is performed on all channels, the time index 2201 with the maximum expression value is selected as the main response time index of the period, and the current main response point time index value is obtained.

[0090] The current main response point time index value represents the time position of the maximum sound pressure value corresponding to the channel with the most dominant sound pressure characteristics in all available channels in the sound source system in a specific period, which is identified in the form of a sample point number, reflecting the main response moment of the energy peak in the sound field in the period. Its essence is to extract a representative time point after comprehensive comparison of the multi-channel sound pressure change trend in the period, and to locate the reference time point with the most main response significance in the current period by introducing sound pressure amplitude difference, channel time stability, and adjacent disturbance consistency. Weighted judgment, so as to locate the reference time point with the most main response significance in the current period. This value not only reflects the time position of the sound wave main response in the period, but also serves as the core reference coordinate for cross-period main response trend analysis.

[0091] The formula operation logic aims to comprehensively evaluate the main response representativeness of each channel in the current period. First, in the numerator part, the sound pressure deviation represents the difference between the maximum sound pressure of the current channel and the average sound pressure of the period. The greater the difference, the stronger the acoustic response of the channel in the current period, and the higher the representativeness, so it needs to be taken as an enhancement factor for the main response point candidate; In the denominator part, first introduce the time offset term For characterizing the stability degree of the current channel in time dimension relative to the reference period, the greater the channel main response time offset is, the worse the stability is, the greater the value should reduce its priority, so it is placed in the denominator to inversely affect the result, and the item is squared to ensure the symbol is positive and to strengthen the suppression of large offset; next, the disturbance term is introduced The term represents the dispersion degree of the sound pressure difference between the channel and its adjacent channel relative to the average of its adjacent channels. This term reflects whether the channel has independent main response characteristics in the neighborhood. The smaller the dispersion is, the more representative it is. The square term reflects the total disturbance intensity under multiple adjacent channels. Then, the time offset term and the disturbance term are summed to form a joint deviation factor. The square root is taken to unify the dimension and perform nonlinear compression processing. Finally, the sound pressure difference is taken as the dominant term and the comprehensive disturbance factor to form a ratio structure. This structure is processed by abs and Operation, the most periodical main response characteristics in all channels are extracted, and the selection of the main response point time index is completed.

[0092] S213: According to the current main response point time index value and the main response point time index value extracted in the last period, the index value is compared and the difference direction of the index time is judged. The main response point index sequence of consecutive periods is recorded and updated to generate the main response point time fluctuation sequence.

[0093] According to the current main response point time index value 2201 and the last period main response point time index 2187, the offset is calculated as 2201-2187=14 points. It is judged that the main response appears a delay trend, the offset direction is marked as +, the time fluctuation of this period is recorded as +14, and the value is added to the time fluctuation vector set. The current period sequence is {+14}. Then, the period main response index is appended as 2192, 2184, 2208, the offset is-9, -8, +24, and the accumulated fluctuation vector sequence becomes {+14, -9, -8, +24}. This sequence describes the drift trend and oscillation amplitude of the main response point on the time axis in consecutive periods, providing basic data for the main response stability trend evaluation. Finally, the main response point time fluctuation sequence is generated.

[0094] Please refer to Figure 4 , the S3 step is:

[0095] S311: Based on the consecutive three period main response point time indexes recorded in the main response point time fluctuation sequence, the time index positions of the current period and the last two periods are extracted in turn. The three groups of index positions are arranged in chronological order. The change direction of the time index between the previous and the next period is compared. The consistency of the change trend is judged to form a continuous direction mark to obtain the main response time offset direction sequence.

[0096] Based on the recorded three consecutive period main response point time indexes in the main response point time fluctuation sequence, the period time indexes are sequentially taken as 2184, 2192, 2208, corresponding to the i-2, i-1 and i periods respectively, the index difference between the i-1 period and the i-2 period is 2192 minus 2184 equal to 8, the index difference between the i period and the i-1 period is 2208 minus 2192 equal to 16, by judging the positive and negative directions of the above two difference values, marked as "positive" direction, and then the result is classified into the direction comparison record set, the consistency of the two direction marks is judged, if the direction marks are the same, it is considered that the direction change remains consistent, otherwise the direction changes, the direction consistency judgment result is finally used for subsequent judgment of path continuity, such as in this example, the directions are all positive, indicating that the main response point index continuously moves backward in three periods, the direction does not change, the main response time offset direction sequence is obtained, as shown in Table 5:

[0097] Table 5 Main response point time index change table

[0098] Period number Main sound point time index Index difference value Direction mark i-2 period 2184 8 Positive i-1 period 2192 16 Positive i period 2208 — —

[0099] S312: According to the main response time offset direction sequence, the time interval between the two consecutive periods is detected within the range combined with the perception time difference, and it is confirmed whether the change direction is consistent, the trend of the period satisfying the consistent direction is marked, and the trend classification intensity value sequence is obtained;

[0100] According to the two consecutive "positive" direction mark values identified in the main response time offset direction sequence, combined with the current sampling rate of 48000Hz, the perception time difference of 20ms is converted into 960 sampling points, and the index difference values of 8 and 16 in paragraph 1 are compared with 960 respectively, by setting the judgment standard as the index difference value must be lower than 960 points and the direction does not change, it is considered that the trend of this period does not change significantly, and then this group of period trend is marked as stable state, if any item exceeds 960 points or the direction changes, it is determined as trend change state, in this example, because the two index difference values are 8 and 16, which are much smaller than 960, and the direction mark remains consistent, therefore it is judged that the period interval meets the trend classification consistency requirement, and the trend classification intensity value sequence is obtained.

[0101] S313: According to the trend classification intensity value sequence and the main response time offset direction sequence, the direction consistency and the trend classification consistency in three periods are compared, whether the current period belongs to the continuous path state or the trend switching state with inconsistent direction is screened, and the path identification content of the current period is updated, the period path switching identification sequence is generated;

[0102] According to the trend classification intensity value sequence and the listed direction marker information obtained in the foregoing, a classified path judgment operation is performed on the current period i and the two preceding periods. First, whether the directions of the three periods are consistent is compared. If two or more directions are "positive" or "negative" and there is no reversal, the direction consistency state is marked as consistent, and then it is determined that the three-period section is "trend consistent". Then, whether the index difference values of the three periods are all within the set sampling point difference threshold range is compared. Here, they are all less than 960, which satisfies the stability criterion. Finally, the identification of the "maintain" path state is formed. If the direction is reversed or the difference value exceeds the threshold, the identification is "switch" state, and the current period path update result is recorded in the period identification field set as a structured field. Finally, the period path switch identification sequence is generated.

[0103] Referring to Figure 5 , the S4 step is:

[0104] S411: Based on the path number corresponding to the period path switch identification sequence, the path number set in the enabled state is extracted combined with the path state record in each current period, and the path number is matched with all available channel numbers in the current period. All channel numbers associated with each path are summarized to obtain the path number associated channel sequence;

[0105] Based on the path number corresponding to the period path switch identification sequence, all path state markers in the current period are called, and the path number set in the "maintain" or "switch" state and effective is screened. The enabled state field of all channels under each path number is extracted in turn. By traversing the channel record running in the current period, the channel number with the state value "enabled" is identified. According to the path number, it is aggregated to form the corresponding structure relationship. In the actual example, the enabled channels associated with the path number P01 are CH05 and CH08, the path number P02 corresponds to CH01 and CH02, and the path number P03 is bound to CH07. According to the matching relationship, the path and channel correspondence table is constructed, and the path channel grouping initial record set is uniformly archived to obtain the path number associated channel sequence, as shown in Table 6:

[0106] Table 6 Path number and channel matching table

[0107] Path number Channel number Main sound point index Direction mark P01 CH05 2208 Positive P01 CH08 2205 Positive P02 CH01 2191 Negative P02 CH02 2189 Negative P03 CH07 2175 Positive

[0108] S412: According to the path number associated channel sequence, the main response point time index record of all associated channels in the current period is extracted. The channel numbers under the path number are sorted in time sequence. The trend of the main response point index is identified, and the channel history direction marker information is screened to determine whether the main response directions of the channels under the same path are consistent. The channel numbers with consistent directions are retained to obtain the channel direction matching sequence.

[0109] According to the combination relationship of path number and channel number in Table 6, the main response point time index and direction marker field of each channel in the current period are extracted, the channel numbers are classified and arranged in the path dimension, the order of the main response point time index in each path is calculated, and whether the directions are consistent is retrieved, for example, the main response indexes of channels CH05 and CH08 in path P01 are 2208 and 2205, and the directions are both "positive", so it is determined that the directions are consistent; the main response indexes of channels CH01 and CH02 in path P02 are 2191 and 2189, and the directions are both "negative", which also belong to consistent directions; if there is a case of inconsistent directions in a path, such as CH09 in P04, the direction is "negative", and the direction of CH10 is "positive", then the path does not meet the screening condition, and CH10 channel is excluded, not counted in the path binding structure record. After screening, sorting, comparison and other action processing, the screening result is written into the structure cache to obtain the channel direction matching sequence.

[0110] S413: Based on the channel direction matching sequence, the channel numbers with consistent directions are aggregated according to the path number, and the screened channel numbers are recorded in the path corresponding table item to form the consistency binding mapping of the path number and the channel number. After the binding result is structured and arranged, the repeated identifiers are removed, the channel set is output according to the path number in turn, and the path binding channel list is generated;

[0111] The channel direction matching sequence is called to bind and confirm the channel number set screened under each path number, and the channel numbers with consistent directions and continuous main response time index distribution are included in the path number binding field, and an index structure is established according to the path number as the field primary key. A one-to-many mapping structure is established under the path field, and the format is P01 binding CH05 and CH08, P02 binding CH01 and CH02, and P03 binding CH07. The uniqueness of each group of channel number combinations is checked, and after removing the repeated numbers and invalid numbers, it is updated to the path binding structure record, and finally the standardized path channel relationship table is formed to generate the path binding channel list.

[0112]

[0113] Please refer to Figure 6 , the S5 step is:

[0114] S511: Based on the channel numbers listed in the path binding channel list, the channel binding relationship of each path in the current period is read, the path state of each path number in the current period is extracted, and the state of the same path number in the previous period is retrieved synchronously. The state consistency or inconsistency result of each path number is recorded in sequence to generate a path state consistency index sequence;

[0115] ​Based on the channel number listed in the path binding channel list, the path number and channel number corresponding structure is obtained, and the path state field of the current period and the last period is extracted respectively, wherein the state field includes direction mark field, main response point index field, binding channel number field and other items. For each path number, the current period state identifier is read and compared with the state identifier recorded in the same path number of the last period at the field level. The direction mark field is expanded according to the value sequence, and the direction mark corresponding to the index P01 path is "positive". If the direction mark is "positive" and the direction mark of the last period is "negative", it is determined that the direction is inconsistent. Similarly, the number level comparison operation is performed on the binding channel number field. For example, the current period path P02 binding channel is CH01, CH02, and the last period is CH01, CH03. Then CH02 is identified as a new channel, CH03 is removed, and the channel structure change is recorded. The numerical value comparison is performed on the main response point index field. If the index offset exceeds the set range, it is determined that the state changes. Otherwise, it is a state continuation. The offset threshold is set to ±20 points based on the average change amplitude of the main response index. According to the state comparison result of multiple path numbers, a state difference matrix is constructed, the path number set with consistent and inconsistent states is divided, the state consistency marking structure of each path between consecutive periods is formed, and finally the path state consistency index sequence is established.

[0116] S512: According to the path state consistency index sequence, the path number set with consistent state is identified, the corresponding channel number is read, the synchronization deviation between the main response point time and the path main response point index is compared, whether the main response point time of each channel is in the synchronization range is evaluated, the valid channel list is recorded according to the path number, and the path synchronization channel set is obtained.

[0117] The path state consistency index sequence is called, the path number whose state field value remains consistent between the last period and the current period is selected, the binding channel number is read and the main response point time index value of the channel in the current period is extracted, the average value of the main response point index of all channels in the path is calculated with the path number as the attribution dimension, and then it is judged whether the main response point index of each channel is within the deviation range. The set offset tolerance range is ±960 points, and the corresponding sampling rate is 20ms time window under the condition of 48kHz. If the channel main response point index deviates from the path main response point average value by more than 960 points, the channel is removed. The path number within the deviation range is retained, and the retained channel number is recombined with the path number as the index field to form the corresponding structure of the path and the synchronization channel. The actual data is shown as follows:

[0118] Table 7 path channel index comparison table

[0119] Path number Channel number Main sound point index Index deviation Synchronization state P01 CH05 2208 +3 Synchronized P01 CH08 2205 0 Synchronized P02 CH01 2180 -12 Synchronized P02 CH02 3100 +932 Rejected

[0120] As shown in Table 7, the CH02 channel under the P02 path is removed due to deviation exceeding the reference threshold, and the remaining channel numbers enter the subsequent analysis link, and finally the path synchronization channel set is obtained.

[0121] S513: According to the path synchronization channel set, read the running state of each path and the bound channel in the current period, combine the path state consistency and channel synchronization data, record and summarize with path number as index, and generate adaptive acoustic analysis results;

[0122] According to the path synchronization channel set, the bound channel record is arranged item by item according to the path number, and the state field set is established for each path number. The state field includes direction mark value, channel main response index average value, channel number, offset interval and other parameters. The direction mark field is extracted in turn, and it is confirmed whether the direction is uniform. If the direction mark of P01 is all "positive", the direction is consistent. The main response index field is extracted, the average value and the maximum and minimum difference of the main response index of each channel are calculated, the index concentration degree is judged, and whether the main response in the path is concentrated is identified. The change degree of the main response index in the continuous period is further extracted, and whether the main response point is stably running is identified. The channel number change record is extracted, compared with the channel record of the last period, and it is confirmed whether the path structure is stable. The path number, direction mark, average index value, channel number, change condition and other structure fields are combined to form a path state record table, which is summarized in turn according to the path number as the primary key index. The record format is as follows:

[0123] Table 8 adaptive acoustic path state table

[0124] Path number Channel number Direction mark Average main sound index Index fluctuation range State change mark P01 2 Positive 2206.5 ±3 Stable P02 1 Negative 2180 ±0 Slightly changed P03 0 / / / No channel

[0125] Referring to Table 8, if the number of channels under the path is 0 or the state changes significantly, it is recorded as an unusable path, and the state structure of the stable path is reserved, and finally the adaptive acoustic analysis results are generated.

[0126] An adaptive acoustic analysis system comprises:

[0127] The channel screening mechanism module is used to execute S1: obtaining the maximum sound pressure amplitude of each sound source channel in each period of the acoustic acquisition stage, screening the channel number whose sound pressure amplitude exceeds the threshold of sound pressure level, and enabling the adjacent channel number, to generate a set of available sound source channels;

[0128] The main response feature extraction module is used to execute S2: based on the set of available sound source channels, obtaining the time index position corresponding to the maximum sound pressure value in each channel in the current period, extracting the main response point time index in the current period, and combining the main response point time index in the last period to judge the change direction of the time difference, and generate the main response point time fluctuation sequence;

[0129] The path consistency identification module is configured to perform S3: according to the current main response point time difference direction mark in the main response point time fluctuation sequence and the previous two periods, in combination with the perception time difference, judging whether the main response point time difference direction is continuous and consistent, synchronously determining the current classification path, and generating a period path switching mark sequence;

[0130] The path channel binding module is configured to perform S4: based on the period path switching mark sequence, matching all the channel numbers currently in an enabled state, performing main response point time sequence alignment, screening the direction consistent channel numbers, and generating a path binding channel list;

[0131] The analysis result confirmation module is configured to perform S5: according to the path binding channel list, detecting whether the state between the current path mark and the previous period path mark is consistent, recording the current path and the channel number, and generating an adaptive acoustic analysis result.

[0132] The above is only a preferred embodiment of the present application, and is not intended to limit the present application in other forms. Any skilled person in the art can use the disclosed technical content to make changes or modifications to equivalent embodiments applied to other fields. However, any simple modification, equivalent change and modification made to the above embodiments without departing from the technical solution content of the present application, in accordance with the technical essence of the present application, still belongs to the protection scope of the technical solution of the present application.

Claims

1. An adaptive acoustic analysis method, characterized in that, Includes the following steps: S1: Obtain the maximum sound pressure amplitude of each sound source channel in each cycle of the acoustic acquisition phase, filter the channel numbers whose sound pressure amplitude exceeds the weighted sound pressure level threshold, enable the adjacent channel numbers, and generate a set of available sound source channels. S2: Based on the available sound source channel set, obtain the time index position corresponding to the maximum sound pressure value in each channel in the current period, extract the time index of the main sound point in the current period, and combine it with the time index of the main sound point in the previous period to determine the direction of time difference change and generate the main sound point time fluctuation sequence. S3: Based on the current and previous two cycles' main sound point time difference direction markers in the main sound point time fluctuation sequence, and combined with the perceived time difference, determine whether the main sound point difference direction is continuous and consistent, synchronously determine the current classification path, and generate a cycle path switching identifier sequence. S4: Based on the periodic path switching identifier sequence, match all channel numbers that are currently in the enabled state, align the main response time order, filter channel numbers with the same direction, and generate a path-bound channel list; S5: Based on the path-bound channel list, detect whether the current path identifier is consistent with the previous cycle path identifier, record the current path and channel number, and generate adaptive acoustic analysis results.

2. The adaptive acoustic analysis method according to claim 1, characterized in that, The available sound source channel set includes a channel number set, adjacency relationship determination results, and sound pressure amplitude filtering records. The main sound point time fluctuation sequence includes the main sound point index difference, direction change identifier, and fluctuation trend label. The periodic path switching identifier sequence includes path number mapping, direction continuity marker, and difference direction judgment results. The path-bound channel list includes direction consistency channel number, main sound point time alignment information, and current path corresponding channel mapping. The adaptive acoustic analysis results include path status records, channel binding relationships, and path continuity identifiers.

3. The adaptive acoustic analysis method according to claim 1, characterized in that, The specific steps for obtaining the set of available sound source channels are as follows: S111: Based on the acoustic acquisition data of each sound source channel in each cycle, obtain the maximum sound pressure amplitude value of the channel, form a mapping set with the channel number and the corresponding maximum sound pressure amplitude, combine the weighted sound pressure level threshold of the corresponding channel in the initial configuration, and use a comparison operation to compare the maximum sound pressure amplitude of each channel with the corresponding threshold to generate a set of sound pressure abnormal channel numbers; S112: Based on the sound pressure anomaly channel number set, according to the adjacency matrix structure information between sound source channels, determine whether each anomaly channel has an adjacent channel number directly connected to it, extract all channel numbers adjacent to the anomaly channel number, and merge them to generate an extended channel number set. S113: Based on the extended channel number set, combined with the sound pressure amplitude characteristics of each channel in the current cycle and the correlation of adjacent channels, the validity of each channel number is identified and screened to determine whether the conditions for enabling the sound source channel are met, thereby obtaining a set of channel numbers that meet the availability criteria and generating a set of available sound source channels.

4. The adaptive acoustic analysis method according to claim 1, characterized in that, The specific steps for obtaining the main response point time fluctuation sequence are as follows: S211: Based on the sound pressure sequence data of each channel in the available sound source channel set, extract the time index corresponding to the maximum value in the sound pressure sequence of each channel in the current period, determine the index position of the corresponding value in the time sequence by locating the maximum value point, and record the channel number and the corresponding time index to generate a channel extreme value time index set; S212: Based on the channel extreme time index set, the extreme time point of the channel where the maximum sound pressure value index is located is used for filtering, the periodic main sound point time index is extracted, and the current main sound point time index value is obtained by calculation; S213: Based on the current main sound point time index value and the main sound point time index value extracted in the previous period, compare the index value size and determine the direction of the index time difference, record and update the continuous period main sound point index sequence, and generate the main sound point time fluctuation sequence.

5. The adaptive acoustic analysis method according to claim 1, characterized in that, The specific steps for obtaining the periodic path switching identifier sequence are as follows: S311: Based on the main sound point time index of the three consecutive cycles recorded in the main sound point time fluctuation sequence, extract the time index position of the current cycle and the previous two cycles in sequence, arrange the three sets of index positions in chronological order, compare the change direction of the time index between the previous and previous cycles, and form a continuous direction mark by judging whether the change trend is consistent, and obtain the main sound time offset direction sequence. S312: Based on the main sound time offset direction sequence, combined with the sensing time difference detection, it is determined whether the time interval between two consecutive cycles is within the range, and it is confirmed whether the direction of change is continuously consistent. For cycles that meet the requirement of continuous consistency of direction, a trend mark is made, and a trend classification intensity value sequence is obtained. S313: Based on the trend classification intensity value sequence and the main response time offset direction sequence, compare the direction consistency and trend classification consistency within three cycles, screen whether the current cycle belongs to a continuous path state or a trend switching state with inconsistent direction, update the path identifier content of the current cycle, and generate a cycle path switching identifier sequence.

6. The adaptive acoustic analysis method according to claim 1, characterized in that, The specific steps for obtaining the path-bound channel list are as follows: S411: Based on the path number corresponding to the periodic path switching identifier sequence, and combined with the path status record in each current period, extract the set of path numbers in the enabled state, and match the path number with all available channel numbers in the current period item by item. Summarize all channel numbers associated with each path to obtain the path number associated channel sequence. S412: Based on the path number associated channel sequence, extract the main sound point time index record of all associated channels in the current period, sort all channel numbers under the path number in chronological order, identify the changing trend of the main sound point index, and screen in combination with the channel historical direction marking information to determine whether the main sound direction of the channels under the same path is consistent, retain the channel numbers with consistent directions, and obtain the channel direction matching sequence. S413: Based on the channel direction matching sequence, aggregate the channel numbers with consistent directions according to the path number, and record the filtered channel numbers in the path corresponding table to form a consistent binding mapping between path number and channel number. After the binding result is structured and the duplicate identifier is removed, the channel set is output sequentially according to the path number to generate a path-bound channel list.

7. The adaptive acoustic analysis method according to claim 1, characterized in that, The specific steps for obtaining the adaptive acoustic analysis results are as follows: S511: Based on the channel numbers listed in the path binding channel list, read the channel binding relationship of each path in the current period, extract the current period path status for each path number, and synchronously retrieve the status under the same path number in the previous period, record the status consistency or inconsistency results of each path number, and generate a path status consistency index sequence. S512: Based on the path status consistency index sequence, identify the set of path numbers with consistent status, read their corresponding channel numbers, compare the synchronization deviation between the main response time and the path main response index, evaluate whether the main response time of each channel is within the synchronization range, re-record the list of valid channels according to the path number, and obtain the set of path synchronization channels. S513: Based on the path synchronization channel set, read the operating status of each path and its bound channel in the current cycle, combine the path status consistency and channel synchronization data, record and summarize the data with the path number as the index, and generate adaptive acoustic analysis results.

8. An adaptive acoustic analysis system, characterized in that, The system is used to implement the adaptive acoustic analysis method according to any one of claims 1-7, comprising: The channel filtering mechanism module is used to execute S1: obtain the maximum sound pressure amplitude of each sound source channel in each cycle of the acoustic acquisition phase, filter the channel numbers whose sound pressure amplitude exceeds the weighted sound pressure level threshold, enable the adjacent channel numbers, and generate a set of available sound source channels. The main sound feature extraction module is used to perform S2: based on the available sound source channel set, obtain the time index position corresponding to the maximum sound pressure value in each channel in the current period, extract the main sound point time index of the current period, and combine it with the main sound point time index of the previous period to determine the direction of time difference change and generate the main sound point time fluctuation sequence. The path consistency identification module is used to execute S3: based on the current and previous two cycles' main sound point time difference direction markers in the main sound point time fluctuation sequence, combined with the perceived time difference, determine whether the main sound point difference direction is continuous and consistent, synchronously determine the current classification path, and generate a periodic path switching identifier sequence; The path channel binding module is used to execute S4: based on the periodic path switching identifier sequence, match all channel numbers that are currently in the enabled state, align the main response time order, filter channel numbers with the same direction, and generate a path binding channel list; The analysis result confirmation module is used to execute S5: based on the path-bound channel list, detect whether the status between the current path identifier and the previous cycle path identifier is consistent, record the current path and channel number, and generate adaptive acoustic analysis results.

Images