Communication information sorting system based on big model historical parameter analysis

Abstract

The invention relates to the technical field of data processing, in particular to a communication information sorting system based on large model historical parameter analysis, which comprises a control terminal, a receiving layer, a sorting layer and an interaction layer, the control terminal is a main control terminal of the system and is used for deciding on and off of the system and on and off of each module in the system; audio frequency appearing at an audio frequency sending end of the interphone is received through the receiving layer, corresponding audio frequency spectrums are obtained based on the received audio frequency, the audio frequency source direction is identified through the audio frequency spectrums, and the audio frequency and the audio frequency spectrums are marked based on the identified audio frequency source direction; according to the method and the device, operations such as segmentation and character information conversion are performed on the corresponding audio frequency spectrum of the audio frequency, so that the purpose of finally reducing the audio frequency frame which does not contain information in the audio frequency is achieved, the audio frequency transmitted by the interphone is clearer and simpler, the contained noise is less, and when a user uses the interphone, the user experience is improved. Based on the system, better communication experience is brought to the user.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a communication information sorting system based on large model historical parameter analysis. Background Art

[0002] Intercom communication is a convenient two-way wireless communication method. It transmits voice signals through a specific frequency band, is easy to operate, and does not require complex settings. It is suitable for short-distance communication, such as construction sites, security patrols, outdoor activities and other scenes. Users can speak by pressing a button and listen by releasing it, realizing instant communication.

[0003] The invention patent with application number 202311025009.3 discloses a communication method for a walkie-talkie, which is characterized in that it is applied to a communication system of a walkie-talkie, the communication system includes a starting walkie-talkie, a transit walkie-talkie and a target terminal, the number of the transit walkie-talkies is N, where N is a positive integer, and the communication method includes: the starting walkie-talkie obtains the device information of the starting walkie-talkie, and sends the device information to the transit walkie-talkie within a first communication range, the device information at least includes location information, and the first communication range is the communication range of the starting walkie-talkie: the transit walkie-talkie receives the device information and forwards the device information to the target terminal, the target terminal exists within the second communication range of at least one of the transit walkie-talkies, and the second communication range is the communication range of the transit walkie-talkie; the target terminal receives the device information.

[0004] The application aims to solve the problem that "in some specific situations, such as when exploring in the wild, when the physical distance between the staff holding the walkie-talkie and the control center exceeds the communication distance of the walkie-talkie, the user cannot communicate with the control center through the walkie-talkie to report his or her location. At this time, the user's location cannot be determined, posing a safety hazard."

[0005] However, during voice communication, the noise in the communication environment will affect the quality of the communication voice, and the noise in the communication environment will also make the communication voice data packets larger, so that the task of transmitting the communication voice by the walkie-talkie increases, and the communication difficulty also increases accordingly;

[0006] To this end, we propose a communication information sorting system based on large model historical parameter analysis. Summary of the invention

[0007] In view of the above-mentioned shortcomings of the prior art, the present invention provides a communication information sorting system based on large-model historical parameter analysis, which solves the technical problems raised in the above-mentioned background technology.

[0008] To achieve the above objectives, the present invention is implemented through the following technical solutions:

[0009] A communication information sorting system based on large model historical parameter analysis includes a control terminal, a receiving layer, a sorting layer and an interaction layer;

[0010] The control terminal is the main control terminal of the system, which is used to start and stop the decision-making system and each module in the system;

[0011] The audio from the intercom audio transmitter is received by the receiving layer, and the corresponding audio spectrum is obtained based on the received audio. The direction of the audio source is identified through the audio spectrum, and the audio and audio spectrum are marked based on the identified audio source direction. The sorting layer receives the audio and audio spectrum, and uses the audio spectrum to segment the audio to obtain several groups of audio frames with different corresponding unit times. The audio frames obtained from each segmentation are further sorted, and the remaining audio frames are fed back to the interaction layer. The interaction layer reorganizes the received audio frames and sends the combined audio to the intercom audio receiver;

[0012] The sorting layer includes a transliteration module, a segmentation module and a sorting module. The transliteration module is used to obtain the audio and the corresponding audio spectrum received in the receiving layer, further select the audio and the audio spectrum with the mark, convert the selected audio into text information, and forward the obtained audio spectrum to the segmentation module. The segmentation module is used to receive the audio spectrum and perform segmentation processing on the audio spectrum. The sorting module is used to receive the audio frame obtained by the segmentation in the segmentation module and the text information conversion result of the audio, and sort the audio frame in combination with the audio frame and the text information conversion result.

[0013] When receiving the audio frame and the text information conversion result of the audio, the sorting module synchronously performs information entropy recognition on each audio frame, generates an audio frame sorting queue based on the information entropy recognition result of the audio frame, and then performs the audio frame sorting operation;

[0014] The recognition logic of the information entropy of the audio frame is expressed as:

[0015]

[0016] Where: H is the information entropy of the audio frame; K is the total amount of eigenvalues ​​in the audio frame; p j is the probability of occurrence of the jth group of eigenvalues; ε j is the weight of the jth group of eigenvalues; θ is the correction;

[0017] Among them, when the audio frame is identifying information entropy, the peak frequency point in the audio frame is used as the boundary, the length of the shortest line segment in the line segments on both sides of the peak frequency point in the audio frame is used as the truncation reference of another line segment, the peak frequency point is used as the starting point, and a line segment equal to the length of the shortest line segment is intercepted on the other line segment. The segmentation accuracy is further set. Based on the segmentation accuracy, the shortest line segment and the intercepted line segment on both sides of the peak frequency point in the audio frame are segmented to obtain several groups of sub-segments. The maximum value of the amplitude values ​​represented by the two ends of the several groups of sub-segments is recorded as the eigenvalue in the audio frame, and the weight ε of the jth group of eigenvalues ​​is j The closer the sub-segment of the eigenvalue source is to the peak frequency point in the audio frame, the greater the weight ε j The larger the value, the greater the weight ε j >0, and

[0018] Furthermore, the receiving layer includes a receiving module, an identification module and a marking module. The receiving module is used to receive the audio appearing at the audio transmitting end of the intercom, and synchronously convert the received audio into an audio spectrum. The identification module is used to obtain the audio spectrum obtained by the audio conversion in the receiving module, and identify the direction of the audio source based on the audio spectrum. The marking module is used to receive the audio source direction identification result in the identification module, mark the audio belonging to the audio source direction and its corresponding audio spectrum, so as to distinguish other audio and corresponding audio spectrum received by the receiving module;

[0019] Among them, the receiving module is integrated by three groups of microphones, the three groups of microphones are distributed in a triangular shape, and the distance between each adjacent group of microphones is equal. The receiving module is installed on the intercom, receives the audio emitted by the user of the intercom audio transmitter in real time, and imports the audio spectrum into the audio spectrum analyzer, and outputs the audio spectrum corresponding to the audio through the audio spectrum analyzer.

[0020] Furthermore, the audio spectrum analyzer is deployed inside the intercom. After the microphone receives the audio sent by the user at the audio sending end of the intercom, it is transmitted to the audio spectrum analyzer in real time. Each group of microphones in the receiving module receives a group of audio, and each group of audio outputs its corresponding audio spectrum through the audio spectrum analyzer.

[0021] The audio source direction recognition logic in the recognition module is expressed as:

[0022]

[0023] Where: SIMM (a, b) is the similarity between audio spectrum a and audio spectrum b; f p1 is the value of the frequency point with the maximum energy in the audio spectrum a; f p2 is the value of the frequency point with the maximum energy in the audio spectrum b; f maxis the value of the frequency point with the largest energy in the two sets of audio spectra; ω1 is the weight; N is the total number of frequency bands obtained by dividing the audio spectrum; E i-a is the energy proportion of the i-th frequency band of the audio spectrum a; E i-b is the energy proportion of the i-th frequency band of the audio spectrum b;

[0024] Among them, based on the above formula, the similarity between the three groups of audio spectra is calculated, and the group with the lowest similarity is selected. max The source audio spectrum is used as a reference for identifying the direction of the audio source. The direction of the audio source is identified by the microphone of the reference source, that is, the direction of the audio source.

[0025] Furthermore, the weight ω1≤0.5, before performing similarity calculation on the audio spectrum a and the audio spectrum b, the two groups of audio spectrum are divided so that the time of each divided sub-audio spectrum is one second;

[0026] Before performing similarity calculation on the audio spectrum a and the audio spectrum b, the two sets of audio spectrum are normalized so that the sum of the energies of the two sets of audio spectrum are both 1;

[0027] The marking content of the audio spectrum and audio corresponding to the audio source direction recognition result in the marking module is a text mark, and the text mark content is: origin, and the other audio and corresponding audio spectrum, that is, the remaining two groups of audio and audio spectrum.

[0028] Furthermore, after receiving the audio spectrum, the segmentation module picks up all valley and peak frequency points in the audio spectrum, and takes the spectrum segment where each three adjacent groups of frequency points are located as a group of audio frames, and among the three groups of adjacent frequency points, there are two groups of valley frequency points and one group of peak frequency points, and the peak frequency point is between the two groups of valley frequency points. The above is recorded as the audio frame determination logic. The segmentation module segments the audio spectrum based on the audio frame determination logic to obtain a plurality of groups of audio frames, and the unit time of the plurality of groups of audio frames is not equal;

[0029] After the segmentation module segments the audio spectrum to obtain a number of groups of audio frames, the segmentation module synchronously obtains the audio obtained in the transliteration module, and segments the audio based on the ratio of the corresponding unit time of the several groups of audio frames to obtain a number of groups of sub-audios. After the several groups of sub-audios are configured in the corresponding audio frames, they are sent to the transliteration module, and text information conversion operations are performed on each group of sub-audios. The text information obtained by the conversion is further bound to the audio frame corresponding to the sub-audio of the text information source, and then returned to the segmentation module. Based on the segmentation module, the bound text information and audio frames are sent to the sorting module.

[0030] Furthermore, 2>θ≥1, and it is subject to: the greater the difference in length of the line segments on both sides of the peak frequency point in the audio frame, the greater the value of the modified θ, and vice versa, the smaller the value of the modified θ;

[0031] When the sorting module receives the audio frame and the text information conversion result of the audio, that is, the text information and the audio frame that are bound to each other, the sorting module runs to generate the audio frame sorting queue, and arranges the audio frames in descending order according to the size of the information entropy corresponding to the audio frame to generate the audio frame sorting queue, and takes the audio frame with the large information entropy in the audio frame sorting queue as the sorting target first;

[0032] The logic representation of the audio frame sorting operation in the sorting module is:

[0033]

[0034] Where: Q t L is the judgment value of whether the audio frame t is sorted as a target; t is the text information converted from the audio frame based on the transliteration module; T0 is the text information converted from the audio based on the transliteration module;

[0035] Wherein, based on the above formula, a determination value of whether each audio frame in the audio frame sorting queue is a sorting target is obtained;

[0036] If the judgment value of whether the audio frame is a sorting target is 1, the audio frame is a sorting target audio frame of the sorting module, and the sorting target audio frame is deleted by the sorting module.

[0037] Furthermore, the interaction layer includes a picking module, a reorganization module and a refresh module. The picking module is used to pick up the remaining audio frames sorted by the sorting module in the sorting layer, further obtain the time domain corresponding to each sorted remaining audio frame in the audio spectrum, and intercept the audio of the same time domain in the audio based on the obtained time domain, which is recorded as sub-audio. The reorganization module is used to receive all the sub-audio intercepted in the picking module, sort each group of sub-audio based on its corresponding time domain to form a new group of audio. The refresh module is used to receive the new audio obtained by the reorganization module, use the new audio as the audio to be transmitted by the intercom audio sending end, transmit it to the intercom audio receiving end, and refresh the system operation.

[0038] Furthermore, when the reorganization module performs the sorting and reorganization operations on each group of sub-audio, each adjacent group of sub-audio performs fade-in and fade-out processing;

[0039] Among them, when two adjacent groups of sub-audios are performing fade-in and fade-out processing, the fade-in and fade-out duration follows: the longer the duration of the sub-audio with the shortest time domain in the two adjacent groups of sub-audios is, the longer the fade-in and fade-out duration is, and vice versa, the shorter the fade-in and fade-out duration is; when two adjacent groups of sub-audios are performing fade-in and fade-out processing, the fade-in and fade-out audio intensity is within the audio intensity threshold composed of the maximum audio intensity and the minimum audio intensity in the two adjacent groups of sub-audios.

[0040] Furthermore, during the operation phase of the refresh module, the audio source direction recognition result recognized by the recognition module of the receiving layer during the current operation of the system is synchronously recorded. When the audio source directions recorded twice in succession are consistent, the system controls the next operation based on the refresh module, the recognition module does not operate, and the previous audio source direction recognition result is applied to the marking module;

[0041] Among them, during each system operation process, the system operation process in which the recognition module is not running is not used as a determination target for whether the directions of the audio sources of two consecutive records are consistent.

[0042] Furthermore, the transliteration module is interactively connected to the segmentation module and the sorting module through the local area network, the transliteration module is interactively connected to the marking module through the local area network, the marking module is interactively connected to the identification module and the receiving module through the local area network, the sorting module is interactively connected to the picking module through the local area network, and the picking module is interactively connected to the reorganization module and the refresh module through the local area network.

[0043] Compared with the known public technology, the technical solution provided by the present invention has the following beneficial effects:

[0044] The present invention provides a communication information sorting system based on large model historical parameter analysis. During operation, the system collects historical usage habits of walkie-talkie users by setting multiple groups of microphones, thereby selecting user audio received by multiple groups of microphones based on the historical usage habits of walkie-talkie users, ensuring that the system can use a group of audio that is the clearest and fits the original intention of the walkie-talkie user audio as the further operation target of the system, thereby performing operations such as segmenting the corresponding audio spectrum of the audio, converting text information, etc., so as to achieve the goal of ultimately eliminating audio frames that do not contain information in the audio, and ultimately making the audio transmitted by the walkie-talkie clearer, more concise and containing less noise, thereby providing users with a better communication experience based on the system when using the walkie-talkie. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0046] Figure 1 It is a structural schematic diagram of a communication information sorting system based on large model historical parameter analysis;

[0047] Figure 2 This is an example diagram of the deployment posture of the receiving module on the intercom in the present invention;

[0048] Figure 3 This is an example diagram of the audio spectrum in the present invention;

[0049] Figure 4 Schematic diagram of the source of eigenvalues ​​used in audio frame segmentation, reorganization logic and audio frame information entropy calculation in the audio spectrum of the present invention. DETAILED DESCRIPTION

[0050] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0051] The present invention will be further described below in conjunction with the embodiments.

[0052] Embodiment 1:

[0053] A communication information sorting system based on large model historical parameter analysis in this embodiment, such as Figure 1 As shown, it includes a control terminal, a receiving layer, a sorting layer and an interaction layer;

[0054] The control terminal is the main control terminal of the system, which is used to start and stop the decision-making system and each module in the system;

[0055] The audio from the intercom audio transmitter is received by the receiving layer, and the corresponding audio spectrum is obtained based on the received audio. The direction of the audio source is identified through the audio spectrum, and the audio and audio spectrum are marked based on the identified audio source direction. The sorting layer receives the audio and audio spectrum, and uses the audio spectrum to segment the audio to obtain several groups of audio frames with different corresponding unit times. The audio frames obtained from each segmentation are further sorted, and the remaining audio frames are fed back to the interaction layer. The interaction layer reorganizes the received audio frames and sends the combined audio to the intercom audio receiver;

[0056] The receiving layer includes a receiving module, an identification module and a marking module. The receiving module is used to receive the audio appearing at the audio transmitting end of the intercom, and synchronously convert the received audio into an audio spectrum. The identification module is used to obtain the audio spectrum obtained by the audio conversion in the receiving module, and identify the direction of the audio source based on the audio spectrum. The marking module is used to receive the audio source direction identification result in the identification module, mark the audio belonging to the audio source direction and its corresponding audio spectrum, so as to distinguish other audio and corresponding audio spectrum received by the receiving module;

[0057] The receiving module is integrated by three groups of microphones, which are distributed in a triangle shape, and the distance between two adjacent groups of microphones is equal. The receiving module is installed on the intercom, receives the audio sent by the user of the intercom audio transmitter in real time, and imports the audio spectrum into the audio spectrum analyzer, and outputs the audio spectrum corresponding to the audio through the audio spectrum analyzer;

[0058] The sorting layer includes a transliteration module, a segmentation module and a sorting module. The transliteration module is used to obtain the audio and the corresponding audio spectrum received in the receiving layer, further select the audio and the audio spectrum with the mark, convert the selected audio into text information, and forward the obtained audio spectrum to the segmentation module. The segmentation module is used to receive the audio spectrum and perform segmentation processing on the audio spectrum. The sorting module is used to receive the audio frame obtained by the segmentation in the segmentation module and the text information conversion result of the audio, and sort the audio frame in combination with the audio frame and the text information conversion result;

[0059] When receiving the audio frame and the text information conversion result of the audio, the sorting module synchronously performs information entropy recognition on each audio frame, generates an audio frame sorting queue based on the information entropy recognition result of the audio frame, and then performs the audio frame sorting operation;

[0060] The recognition logic of the information entropy of the audio frame is expressed as:

[0061]

[0062] Where: H is the information entropy of the audio frame; K is the total amount of eigenvalues ​​in the audio frame; p j is the probability of occurrence of the jth group of eigenvalues; ε jis the weight of the jth group of eigenvalues; θ is the correction;

[0063] 2>θ≥1, and obeys: the greater the difference in length of the line segments on both sides of the peak frequency point in the audio frame, the larger the value of the modified θ is, and vice versa, the smaller the value of the modified θ is;

[0064] When the sorting module receives the audio frame and the audio text information conversion result, that is, the text information and audio frame bound to each other, and the sorting module runs to generate the audio frame sorting queue, the audio frames are arranged in descending order according to the size of the corresponding information entropy of the audio frame to generate the audio frame sorting queue, and the audio frame with the large information entropy in the audio frame sorting queue is prioritized as the sorting target;

[0065] The logic of audio frame sorting operation in the sorting module is expressed as:

[0066]

[0067] Where: Q t L is the judgment value of whether the audio frame t is sorted as a target; t is the text information converted from the audio frame based on the transliteration module; T0 is the text information converted from the audio based on the transliteration module;

[0068] Wherein, based on the above formula, a determination value of whether each audio frame in the audio frame sorting queue is a sorting target is obtained;

[0069] If the judgment value of whether the audio frame is a sorting target is 1, the audio frame is a sorting target audio frame of the sorting module, and the sorting target audio frame is deleted by the sorting module;

[0070] Among them, when the audio frame is identifying information entropy, the peak frequency point in the audio frame is used as the boundary, the length of the shortest line segment in the line segments on both sides of the peak frequency point in the audio frame is used as the truncation reference of another line segment, the peak frequency point is used as the starting point, and a line segment equal to the length of the shortest line segment is intercepted on the other line segment. The segmentation accuracy is further set. Based on the segmentation accuracy, the shortest line segment and the intercepted line segment on both sides of the peak frequency point in the audio frame are segmented to obtain several groups of sub-segments. The maximum value of the amplitude values ​​represented by the two ends of the several groups of sub-segments is recorded as the eigenvalue in the audio frame, and the weight ε of the jth group of eigenvalues ​​is j The closer the sub-segment of the eigenvalue source is to the peak frequency point in the audio frame, the greater the weight ε j The larger the value, the greater the weight ε j >0, and

[0071] The interaction layer includes a picking module, a reorganization module and a refreshing module. The picking module is used to pick up the remaining audio frames sorted by the sorting module in the sorting layer, further obtain the corresponding time domain of each sorted remaining audio frame in the audio spectrum, and intercept the audio of the same time domain in the audio by obtaining the time domain, which is recorded as a sub-audio. The reorganization module is used to receive all the sub-audios intercepted in the picking module, sort each group of sub-audios based on their corresponding time domains to form a new group of audios. The refreshing module is used to receive the new audio obtained by the reorganization module, use the new audio as the audio to be transmitted by the intercom audio transmitter, transmit it to the intercom audio receiver, and refresh the system operation;

[0072] The transliteration module is interactively connected to the segmentation module and the sorting module through the local area network, the transliteration module is interactively connected to the marking module through the local area network, the marking module is interactively connected to the identification module and the receiving module through the local area network, the sorting module is interactively connected to the picking module through the local area network, and the picking module is interactively connected to the reorganization module and the refresh module through the local area network.

[0073] In this embodiment, the control terminal controls the operation of the system, the receiving module operates to receive the audio appearing at the audio transmitting end of the intercom, and synchronously converts the received audio into an audio spectrum, the identification module synchronously obtains the audio spectrum obtained by the audio conversion in the receiving module, identifies the direction of the audio source based on the audio spectrum, the marking module post-operates to receive the audio source direction identification result in the identification module, marks the audio belonging to the audio source direction and its corresponding audio spectrum to distinguish other audio and corresponding audio spectrum received by the receiving module, the transliteration module further obtains the audio and corresponding audio spectrum received in the receiving layer, further selects the marked audio and audio spectrum, converts the selected audio into text information, and forwards the obtained audio spectrum to the segmentation module, which then receives the audio spectrum and performs the corresponding audio spectrum. The audio spectrum is segmented and processed, and the sorting module receives the audio frames obtained by the segmentation module and the text information conversion results of the audio in real time, and sorts the audio frames in combination with the audio frames and the text information conversion results. Finally, the picking module picks up the remaining audio frames sorted by the sorting module in the sorting layer, and further obtains the time domain corresponding to each sorted remaining audio frame in the audio spectrum, and intercepts the audio of the same time domain in the audio based on the obtained time domain, which is recorded as sub-audio. The reorganization module receives all the sub-audio intercepted in the picking module in real time, sorts each group of sub-audio based on its corresponding time domain to form a new group of audio, and receives the new audio obtained by the reorganization module through the refresh module, and uses the new audio as the audio to be transmitted by the intercom audio transmitter, and transmits it to the intercom audio receiver, and refreshes the system operation;

[0074] Through the operation of the system in the above embodiment, the walkie-talkie performs audio spectrum conversion on the audio when the user uses the transmitted audio, thereby providing audio segmentation conditions at one time, and finally sorting the audio based on the segmented audio spectrum, achieving the purpose of audio simplification and clarity, and bringing a better communication experience to the walkie-talkie user.

[0075] See also Figure 2 As shown, the figure further shows the positions of the three groups of microphones used for the integration of the receiving module, based on the three groups of circular shadow areas shown in the figure, which are used to represent the deployment positions of the microphones;

[0076] See also Figure 3 , Figure 4 As shown, Figure 3 The example shows the audio spectrum resulting from the audio conversion;

[0077] based on Figure 4 The arrow in the middle indicates:

[0078] The vertical coordinates (a); (b); (c); (d) in the figure respectively represent: Figure 3 Line spectrum graphics in the audio spectrum; several groups of audio frames obtained by segmentation; dotted lines represent the sorting target audio frames, and solid lines represent the sorting remaining audio frames; the reconstructed audio spectrum after deleting the sorting target audio frames;

[0079] Based on the horizontal arrows and curved arrows in the figure, the recognition process of audio frame information entropy is shown:

[0080] Based on the indication of the curve arrow, the information entropy recognition target is identified by taking a group of the remaining audio frames as an example. For the convenience of display, the group of audio frames is enlarged, and the peak frequency point in the audio frame is further used as the boundary. The length of the shortest line segment on both sides of the peak frequency point in the audio frame is used as the reference for truncation of another line segment. The peak frequency point is used as the starting point, and a line segment equal to the length of the shortest line segment is intercepted on the other line segment. The result is shown in (1). The segmentation accuracy is further set. Based on the segmentation accuracy, the shortest line segment and the intercepted line segment on both sides of the peak frequency point in the audio frame are segmented to obtain several groups of sub-segments. The result is shown in (2).

[0081] Embodiment 2:

[0082] In terms of specific implementation, based on Example 1, this example refers to Figure 1 A communication information sorting system based on large model historical parameter analysis in Example 1 is further described in detail:

[0083] The audio spectrum analyzer is deployed inside the intercom. After the microphone receives the audio sent by the user at the audio transmitter of the intercom, it is transmitted to the audio spectrum analyzer in real time. Each group of microphones in the receiving module receives a group of audio, and each group of audio outputs its corresponding audio spectrum through the audio spectrum analyzer.

[0084] The audio source direction recognition logic in the recognition module is expressed as:

[0085]

[0086] Where: SIMM (a, b) is the similarity between audio spectrum a and audio spectrum b; f p1 is the value of the frequency point with the maximum energy in the audio spectrum a; f p2 is the value of the frequency point with the maximum energy in the audio spectrum b; f max is the value of the frequency point with the largest energy in the two sets of audio spectra; ω1 is the weight; N is the total number of frequency bands obtained by dividing the audio spectrum; E i-a is the energy proportion of the i-th frequency band of the audio spectrum a; E i-b is the energy proportion of the i-th frequency band of the audio spectrum b;

[0087] Among them, based on the above formula, the similarity between the three groups of audio spectra is calculated, and the group with the lowest similarity is selected. max The source audio spectrum is used as a reference for identifying the direction of the audio source. The direction of the audio source is identified by the microphone of the source, i.e. the direction of the audio source.

[0088] The weight ω1≤0.5, before performing similarity calculation on the audio spectrum a and the audio spectrum b, the two groups of audio spectrum are divided so that the time of each divided sub-audio spectrum is one second;

[0089] Before performing similarity calculation on the audio spectrum a and the audio spectrum b, the two sets of audio spectrum are normalized so that the sum of the energies of the two sets of audio spectrum is 1;

[0090] In the marking module, the audio spectrum and the audio corresponding to the audio source direction recognition result are marked as text marks, and the text mark content is: origin, other audio and corresponding audio spectrum, that is, the remaining two groups of audio and audio spectrum.

[0091] In this embodiment, through the above-mentioned settings, the audio source direction recognition logic in the identification module is further limited, so that based on this as a historical parameter, the use habits of the intercom users are analyzed, and the audio source direction is analyzed when the user operates the intercom, so that the audio captured by the three sets of microphones integrated in the receiving module can be more appropriately selected by the system, thereby improving the system operation accuracy.

[0092] Embodiment 3:

[0093] In terms of specific implementation, based on Example 1, this example refers to Figure 1 A communication information sorting system based on large model historical parameter analysis in Example 1 is further described in detail:

[0094] After receiving the audio spectrum, the segmentation module picks up all valley and peak frequency points in the audio spectrum, and takes the spectrum segment where each three adjacent groups of frequency points are located as a group of audio frames, and among the three groups of adjacent frequency points, there are two groups of valley frequency points and one group of peak frequency points, and the peak frequency point is between the two groups of valley frequency points. The above is recorded as the audio frame determination logic. The segmentation module segments the audio spectrum based on the audio frame determination logic to obtain several groups of audio frames, and the unit time of the several groups of audio frames is not equal;

[0095] After the segmentation module segments the audio spectrum to obtain several groups of audio frames, the segmentation module synchronously obtains the audio obtained in the transliteration module, and segments the audio based on the ratio of the corresponding unit time of the several groups of audio frames to obtain several groups of sub-audios. After the several groups of sub-audios are configured in the corresponding audio frames, they are sent to the transliteration module, and text information conversion operations are performed on each group of sub-audios. The converted text information is further bound to the audio frame corresponding to the sub-audio of the text information source, and returned to the segmentation module. Based on the segmentation module, the bound text information and audio frames are sent to the sorting module.

[0096] Through the above settings, the operating logic of the system in Example 1 when performing audio frame segmentation on the audio spectrum is limited, ensuring that the segmentation module in the system stably operates and outputs audio frames, providing further operation data support for the operation of the analysis module and the system interaction layer.

[0097] like Figure 1 As shown, when the reorganization module sorts and reorganizes each group of sub-audio, each adjacent group of sub-audio performs fade-in and fade-out processing;

[0098] When two adjacent groups of sub-audios perform fade-in and fade-out processing, the fade-in and fade-out duration is subject to: the longer the duration of the shortest sub-audio in the time domain of the two adjacent groups of sub-audios is, the longer the fade-in and fade-out duration is, and vice versa; when two adjacent groups of sub-audios perform fade-in and fade-out processing, the fade-in and fade-out audio intensity is within the audio intensity threshold composed of the maximum audio intensity and the minimum audio intensity of the two adjacent groups of sub-audios;

[0099] During the refresh module operation phase, the audio source direction recognition result of the recognition module of the receiving layer during the current operation of the system is synchronously recorded. When the audio source directions recorded twice in succession are consistent, the system controls the next operation based on the refresh module, the recognition module does not run, and the previous audio source direction recognition result is applied to the marking module;

[0100] Among them, during each system operation process, the system operation process in which the recognition module is not running is not used as a determination target for whether the directions of the audio sources of two consecutive records are consistent.

[0101] Through the above settings, further continuous operation logic is provided for the system in Example 1, ensuring that the system can stably and continuously serve the intercom communication.

[0102] In summary, during operation, the system in the above embodiment collects the historical usage habits of the walkie-talkie user by setting multiple groups of microphones, and selects the user audio received by the multiple groups of microphones based on the historical usage habits of the walkie-talkie user, ensuring that the system can use the clearest set of audio that fits the original intention of the walkie-talkie user's audio as the further operation target of the system, thereby performing operations such as segmenting the corresponding audio spectrum of the audio and converting text information to achieve the goal of ultimately eliminating the audio frames that do not contain information in the audio, ultimately making the audio transmitted by the walkie-talkie clearer, more concise and containing less noise, and then when the user uses the walkie-talkie, the system provides the user with a better communication experience.

[0103] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements will not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A communication information sorting system based on large model historical parameter analysis, characterized in that: It includes control terminal, receiving layer, sorting layer and interaction layer; The control terminal is the main control terminal of the system, which is used to start and stop the decision-making system and each module in the system; The audio from the intercom audio transmitter is received by the receiving layer, and the corresponding audio spectrum is obtained based on the received audio. The direction of the audio source is identified through the audio spectrum, and the audio and audio spectrum are marked based on the identified audio source direction. The sorting layer receives the audio and audio spectrum, and uses the audio spectrum to segment the audio to obtain several groups of audio frames with different corresponding unit times. The audio frames obtained from each segmentation are further sorted, and the remaining audio frames are fed back to the interaction layer. The interaction layer reorganizes the received audio frames and sends the combined audio to the intercom audio receiver; The sorting layer includes a transliteration module, a segmentation module and a sorting module. The transliteration module is used to obtain the audio and the corresponding audio spectrum received in the receiving layer, further select the audio and the audio spectrum with the mark, convert the selected audio into text information, and forward the obtained audio spectrum to the segmentation module. The segmentation module is used to receive the audio spectrum and perform segmentation processing on the audio spectrum. The sorting module is used to receive the audio frame obtained by the segmentation in the segmentation module and the text information conversion result of the audio, and sort the audio frame in combination with the audio frame and the text information conversion result. When receiving the audio frame and the text information conversion result of the audio, the sorting module synchronously performs information entropy recognition on each audio frame, generates an audio frame sorting queue based on the information entropy recognition result of the audio frame, and then performs the audio frame sorting operation; The recognition logic of the information entropy of the audio frame is expressed as: Where: H is the information entropy of the audio frame; K is the total amount of eigenvalues ​​in the audio frame; p j is the probability of occurrence of the jth group of eigenvalues; ε j is the weight of the jth group of eigenvalues; θ is the correction; Among them, when the audio frame is identifying information entropy, the peak frequency point in the audio frame is used as the boundary, the length of the shortest line segment in the line segments on both sides of the peak frequency point in the audio frame is used as the truncation reference of another line segment, the peak frequency point is used as the starting point, and a line segment equal to the length of the shortest line segment is intercepted on the other line segment. The segmentation accuracy is further set. Based on the segmentation accuracy, the shortest line segment and the intercepted line segment on both sides of the peak frequency point in the audio frame are segmented to obtain several groups of sub-segments. The maximum value of the amplitude values ​​represented by the two ends of the several groups of sub-segments is recorded as the eigenvalue in the audio frame, and the weight ε of the jth group of eigenvalues ​​is j The closer the sub-segment of the eigenvalue source is to the peak frequency point in the audio frame, the greater the weight ε j The larger the value, the greater the weight ε j >0, and 2. A communication information sorting system based on large model historical parameter analysis according to claim 1, characterized in that: The receiving layer includes a receiving module, an identification module and a marking module. The receiving module is used to receive the audio appearing at the audio transmitting end of the intercom, and synchronously convert the received audio into an audio spectrum. The identification module is used to obtain the audio spectrum obtained by the audio conversion in the receiving module, and identify the direction of the audio source based on the audio spectrum. The marking module is used to receive the audio source direction identification result in the identification module, mark the audio belonging to the audio source direction and its corresponding audio spectrum, so as to distinguish other audio and corresponding audio spectrum received by the receiving module; Among them, the receiving module is integrated by three groups of microphones, the three groups of microphones are distributed in a triangular shape, and the distance between each adjacent group of microphones is equal. The receiving module is installed on the intercom, receives the audio emitted by the user of the intercom audio transmitter in real time, and imports the audio spectrum into the audio spectrum analyzer, and outputs the audio spectrum corresponding to the audio through the audio spectrum analyzer.

3. A communication information sorting system based on large model historical parameter analysis according to claim 2, characterized in that: The audio spectrum analyzer is deployed inside the intercom. After the microphone receives the audio sent by the user at the audio sending end of the intercom, it is transmitted to the audio spectrum analyzer in real time. Each group of microphones in the receiving module receives a group of audio, and each group of audio outputs its corresponding audio spectrum through the audio spectrum analyzer. The audio source direction recognition logic in the recognition module is expressed as: Where: SIMM (a, b) is the similarity between audio spectrum a and audio spectrum b; f p1 is the value of the frequency point with the maximum energy in the audio spectrum a; f p2 is the value of the frequency point with the maximum energy in the audio spectrum b; f max is the value of the frequency point with the largest energy in the two sets of audio spectra; ω1 is the weight; N is the total number of frequency bands obtained by dividing the audio spectrum; E i-a is the energy proportion of the i-th frequency band of the audio spectrum a; E i-b is the energy proportion of the i-th frequency band of the audio spectrum b; Among them, based on the above formula, the similarity between the three groups of audio spectra is calculated, and the group with the lowest similarity is selected. max The source audio spectrum is used as a reference for identifying the direction of the audio source. The direction of the audio source is identified by the microphone of the reference source, that is, the direction of the audio source.

4. A communication information sorting system based on large model historical parameter analysis according to claim 2 or 3, characterized in that: The weight ω1≤0.5, before performing similarity calculation on the audio spectrum a and the audio spectrum b, the two groups of audio spectrum are divided so that the time of each divided sub-audio spectrum is one second; Before performing similarity calculation on the audio spectrum a and the audio spectrum b, the two sets of audio spectrum are normalized so that the sum of the energies of the two sets of audio spectrum are both 1; The marking content of the audio spectrum and audio corresponding to the audio source direction recognition result in the marking module is a text mark, and the text mark content is: origin, and the other audio and corresponding audio spectrum, that is, the remaining two groups of audio and audio spectrum.

5. A communication information sorting system based on large model historical parameter analysis according to claim 1, characterized in that: After receiving the audio spectrum, the segmentation module picks up all valley and peak frequency points in the audio spectrum, and takes the spectrum segment where each three adjacent groups of frequency points are located as a group of audio frames, and among the three groups of adjacent frequency points, there are two groups of valley frequency points and one group of peak frequency points, and the peak frequency point is between the two groups of valley frequency points. The above is recorded as the audio frame determination logic. The segmentation module segments the audio spectrum based on the audio frame determination logic to obtain a plurality of groups of audio frames, and the unit time of the plurality of groups of audio frames is not equal; After the segmentation module segments the audio spectrum to obtain a number of groups of audio frames, the segmentation module synchronously obtains the audio obtained in the transliteration module, and segments the audio based on the ratio of the corresponding unit time of the several groups of audio frames to obtain a number of groups of sub-audios. After the several groups of sub-audios are configured in the corresponding audio frames, they are sent to the transliteration module, and text information conversion operations are performed on each group of sub-audios. The text information obtained by the conversion is further bound to the audio frame corresponding to the sub-audio of the text information source, and then returned to the segmentation module. Based on the segmentation module, the bound text information and audio frames are sent to the sorting module.

6. A communication information sorting system based on large model historical parameter analysis according to claim 1, characterized in that: 2>θ≥1, and obeys: the larger the difference in length of the line segments on both sides of the peak frequency point in the audio frame, the larger the value of the modified θ is, and vice versa, the smaller the value of the modified θ is; When the sorting module receives the audio frame and the text information conversion result of the audio, that is, the text information and the audio frame that are bound to each other, the sorting module runs to generate the audio frame sorting queue, and arranges the audio frames in descending order according to the size of the information entropy corresponding to the audio frame to generate the audio frame sorting queue, and takes the audio frame with the large information entropy in the audio frame sorting queue as the sorting target first; The logic representation of the audio frame sorting operation in the sorting module is: Where: Q t L is the judgment value of whether the audio frame t is sorted as a target; t The text information converted from the audio frame based on the transliteration module; T0 is the text information converted from audio based on the transliteration module; Wherein, based on the above formula, a determination value of whether each audio frame in the audio frame sorting queue is a sorting target is obtained; If the judgment value of whether the audio frame is a sorting target is 1, the audio frame is a sorting target audio frame of the sorting module, and the sorting target audio frame is deleted by the sorting module.

7. A communication information sorting system based on large model historical parameter analysis according to claim 1, characterized in that: The interaction layer includes a picking module, a reorganization module and a refreshing module. The picking module is used to pick up the remaining audio frames sorted by the sorting module in the sorting layer, further obtain the time domain corresponding to each sorted remaining audio frame in the audio spectrum, and intercept the audio of the same time domain in the audio based on the acquired time domain, which is recorded as sub-audio. The reorganization module is used to receive all the sub-audio intercepted in the picking module, sort each group of sub-audio based on its corresponding time domain to form a new group of audio. The refreshing module is used to receive the new audio obtained by the reorganization module, use the new audio as the audio to be transmitted by the intercom audio transmitting end, transmit it to the intercom audio receiving end, and refresh the system operation.

8. A communication information sorting system based on large model historical parameter analysis according to claim 7, characterized in that: When the reorganization module performs the sorting and reorganization operations on each group of sub-audio, each adjacent group of sub-audio performs fade-in and fade-out processing; Among them, when two adjacent groups of sub-audios are performing fade-in and fade-out processing, the fade-in and fade-out duration follows: the longer the duration of the sub-audio with the shortest time domain in the two adjacent groups of sub-audios is, the longer the fade-in and fade-out duration is, and vice versa, the shorter the fade-in and fade-out duration is; when two adjacent groups of sub-audios are performing fade-in and fade-out processing, the fade-in and fade-out audio intensity is within the audio intensity threshold composed of the maximum audio intensity and the minimum audio intensity in the two adjacent groups of sub-audios.

9. A communication information sorting system based on large model historical parameter analysis according to claim 7, characterized in that: During the refresh module operation phase, the audio source direction recognition result recognized by the recognition module of the receiving layer during the current operation of the system is synchronously recorded. When the audio source directions recorded twice in succession are consistent, the system controls the next operation based on the refresh module, the recognition module does not operate, and the previous audio source direction recognition result is applied to the marking module; Among them, during each system operation process, the system operation process in which the recognition module is not running is not used as a determination target for whether the directions of the audio sources of two consecutive records are consistent.

10. A communication information sorting system based on large model historical parameter analysis according to claim 1, characterized in that: The transliteration module is interactively connected to a segmentation module and a sorting module through a local area network, the transliteration module is interactively connected to a marking module through a local area network, the marking module is interactively connected to an identification module and a receiving module through a local area network, the sorting module is interactively connected to a picking module through a local area network, and the picking module is interactively connected to a reorganization module and a refresh module through a local area network.

Images