Conference summary automatic generation method and system driven by intelligent conference terminal

Abstract

The invention relates to the technical field of intelligent conference terminals, and discloses a conference summary automatic generation method and system driven by an intelligent conference terminal. The method comprises the following steps: acquiring a participating identity list, collecting sound data and extracting voiceprint features; calculating identity features and generating a unique list serial number; converting the sound data into a text conference record and associating the text conference record with identity information; converting the conference record into an actual code by adopting a continuous Chinese coding algorithm; performing dynamic shift encryption on the code based on the list serial number to generate an encryption record; and storing the identity list, the serial number, the encryption record and the corresponding relation to form a conference summary. The system supports dynamic coding adjustment of newly added participants in the conference process, and optimizes the encryption effect through weighted voiceprint feature calculation, data volume ratio inverse correlation adjustment offset value and the like. And during decryption, the original conference record is restored through a reverse algorithm, and automatic generation and efficient sharing of the summary are realized while the content security is ensured.

Description

Technical Field

[0001] The present application relates to the technical field of intelligent conference terminals, and in particular to a method and system for automatically generating meeting minutes driven by an intelligent conference terminal. Background Art

[0002] As the digitalization process of enterprises accelerates and cross-regional collaboration increases, there is an urgent need for real-time processing, accurate retention, and rapid sharing of meeting content. Intelligent transformation can realize the automatic generation of meeting minutes and support the rapid implementation of decisions. The current intelligent transformation of conference scenarios has seen the gradual application of technologies such as smart conference terminals, voice recognition, and natural language processing. Terminals can achieve real-time speech transcription and distinguish between multiple speakers. Some systems can automatically extract meeting key points and generate preliminary minutes frameworks. Furthermore, cloud platforms support the storage and cross-device sharing of meeting data, and AI algorithm optimization improves the accuracy of minutes generation. Existing intelligent conferencing solutions primarily focus on data transmission and storage, without modifying the meeting content itself. This leaves sensitive information in its original form, making it vulnerable to leakage if the terminal or platform is compromised, and the encryption effectiveness is limited. Summary of the Invention

[0003] In order to improve the encryption effect of meeting content, the present application provides a method and system for automatically generating meeting minutes driven by an intelligent conference terminal.

[0004] In the first aspect, the present application provides a method for automatically generating meeting minutes driven by an intelligent conference terminal, which adopts the following technical solutions: A method for automatically generating meeting minutes driven by an intelligent conference terminal comprises the following steps: Get the list of identities attending the meeting; collecting voice data of the meeting in real time, and identifying voiceprint information corresponding to the identity list from the voice data, the voiceprint information including fundamental frequency information and resonance information; Calculating identity features based on the fundamental frequency information and the resonance information, and sorting the identity features to obtain list numbers corresponding to the identity lists, wherein the list numbers are non-repeated; Recognizing the sound data as meeting record data containing text, and associating the identity list with the meeting record data; Converting the text content in the conference record data into actual codes according to a Chinese coding algorithm with a continuous coding range; Based on the association relationship, the identity list is traversed, and the actual code of the meeting record data is shifted according to the list serial number corresponding to the identity list to obtain an encrypted code; Matching a corresponding encrypted record from the Chinese encoding algorithm according to the encryption code; Obtaining a content correspondence between the list number and the encrypted record according to the association relationship; The identity list, the list serial number, the encrypted record and the corresponding relationship between the content are stored as meeting minutes data.

[0005] By adopting the above technical solution, a list of meeting attendees is obtained. Combined with real-time meeting audio data, voiceprint information containing fundamental frequency and resonance information is identified. Identity features are calculated and sorted to obtain a unique list number, laying the foundation for identity association for encryption. After associating the meeting record data identified by the audio data with the list of identities, it is converted into an actual code using a continuous Chinese encoding algorithm. The actual code is then shifted based on the list number to obtain an encrypted code. The encrypted record is matched and a correspondence between the list number and the encrypted record content is established. Finally, the correspondence between the identity list, list number, encrypted record, and content is stored as meeting minutes data. This entire process ensures that the meeting content exists in an encrypted, coded form, rather than in original text. It can only be decrypted by matching the identity and serial number. If the data is obtained by others, each character in the converted text will be offset, resulting in incorrect content. This prevents the leakage of sensitive information when the terminal or platform is compromised, significantly improving the encryption of meeting content.

[0006] Optionally, the method further comprises the following steps: Obtaining the meeting minutes data; Extracting the identity list, the list serial number, the encrypted record and the corresponding relationship between the content from the meeting minutes data; Matching the encryption code according to the encryption record and the inverse algorithm of the Chinese encoding algorithm; Matching the list number with the encryption code according to the corresponding relationship between the encryption record and the content; Reverse shifting the encrypted code according to the corresponding list number to obtain the actual code; Matching the meeting record data according to the actual encoding and the inverse algorithm of the Chinese encoding algorithm; Converting the content correspondence into the association relationship; Output the identity list, the association relationship and the meeting record data.

[0007] By employing this technical solution, we extract elements such as the list of identities and their serial numbers. Using the inverse of the Chinese encoding algorithm, we restore the encrypted records to their encrypted codes. Then, by reverse-shifting the list serial numbers, we obtain the actual codes, ultimately matching the original meeting record data and converting the associated relationships. This process strictly follows the reverse operation of the forward encryption path, ensuring accurate and reliable decryption, fully restoring the original meeting content and identity associations, and achieving secure restoration of meeting minutes data.

[0008] Optionally, the method further comprises the following steps: If the identity feature values ​​of any two corresponding identity lists are the same, then compare the values ​​of the fundamental frequency information corresponding to the two identity features, and increase the value of the identity feature with the larger fundamental frequency information value; If the identity feature values ​​of any two corresponding identity lists are the same and the corresponding fundamental frequency information values ​​are also the same, then the values ​​of the resonance information corresponding to the two features are compared and the value of the identity feature with the larger resonance information value is increased.

[0009] By adopting this technical solution, when identity feature values ​​are identical, the fundamental frequency information values ​​are first compared, and the identity feature value with the larger fundamental frequency value is boosted. If the fundamental frequencies remain the same, the resonance information values ​​are then compared, and the one with the larger resonance value is boosted. This differentiated processing of identity feature values ​​ensures unique identification, ensuring a one-to-one correspondence between identity features and attendees, and helping to ensure the uniqueness and accuracy of the list numbering.

[0010] Optionally, the method further comprises the following steps: Calculating identity features using a weighted average method based on the fundamental frequency information and the resonance information; The weight of the fundamental frequency information is adjusted in a positive correlation with the number of the identity lists; the smaller the number of the identity lists, the smaller the weight of the fundamental frequency information; the larger the number of the identity lists, the greater the weight of the fundamental frequency information; The weight of the resonance information is adjusted in a positive correlation according to the total data volume of the sound data; the smaller the total data volume of the sound data, the smaller the weight of the resonance information; the larger the total data volume of the sound data, the greater the weight of the resonance information.

[0011] By adopting the above technical solution, the fundamental frequency information weight is increased when the number of participants increases, thereby improving the individual feature differentiation; the increase in the total amount of sound data can increase the resonance information weight and enhance the feature stability in complex environments. The two weights can be adaptively adjusted according to the actual situation of the meeting.

[0012] Optionally, the method further comprises the following steps: According to the proportion of the text content corresponding to the identity list in the meeting record data, the shift offset value of the identity list is adjusted in an anti-correlation manner; the smaller the proportion of the data volume, the larger the shift offset value of the identity list; the larger the proportion of the data volume, the smaller the shift offset value of the identity list.

[0013] By adopting the above technical solution, shift = list number + shift offset value; the shift offset value is adjusted inversely according to the proportion of data volume corresponding to the identity. The more people speak, the less encrypted offset is, realizing dynamic differentiated encryption and reducing the probability and amount of offset exceeding the range; it is beneficial to improve the overall accuracy of the meeting minutes.

[0014] Optionally, the method further comprises the following steps: If the identity list is newly added during the process of generating the meeting record data, it is sorted by time; Generating a meeting minutes data corresponding to the current M identity lists within a first period of time; During the second period of time, another meeting minutes data corresponding to the newly added N identity lists is generated, wherein the newly added list serial number corresponds to the sorting position + M.

[0015] By adopting the above technical solution, a time-sharing sorting mechanism is used to realize dynamic personnel management: when adding new participants, the meeting minutes data of the current M identity list is generated first, and then the newly added N identities are processed separately, and their sorting positions are automatically connected to the original sequence; this helps to reduce the interference of dynamic adjustments of participants on the encryption process, ensure that the data of new and old participants are independently encrypted and the serial numbers are continuous, and improve the scalability of the system and the flexibility of meeting content encryption.

[0016] Optionally, the method further comprises the following steps: If N new identity lists are added during the process of generating the meeting record data, then the lists are sorted by time; Generating the list serial numbers of the current M identity lists within a first time period, wherein the sorting positions of the M identity lists are magnified N+1 times; Generate the list serial numbers of N newly added identity lists within the second period of time; Insert the current M identity lists into the newly added N identity lists, re-sort them according to the corresponding list serial numbers, and then generate the meeting minutes data.

[0017] By adopting the above technical solution, the identity list corresponding to the newly added participants is processed through time-sharing sorting and serial number reservation mechanism; first, the current M list serial numbers are amplified N times to reserve space, and then N new serial numbers are generated, and finally inserted and reorganized; this helps to reduce the interference of dynamic adjustments of participants on the encryption process, ensure that the data of new and old participants are independently encrypted and the serial numbers are continuous, and improve the scalability of the system and the flexibility of conference content encryption.

[0018] Optionally, the method further comprises the following steps: Obtaining an even-odd attribute according to the total amount of the list numbers; If the parity attribute is an odd number, the shift is shifted along the first direction; otherwise, the shift is shifted along the second direction.

[0019] By adopting the above technical solution, the shift direction is distinguished by the odd or even attribute of the total number of list numbers, with odd numbers using the first direction and even numbers using the second direction. This increases the randomness and complexity of the encryption code shift, avoids the regularity of a single direction, improves the anti-cracking ability of the conference content encryption, and enhances information security.

[0020] In a second aspect, the present application provides a system for automatically generating meeting minutes driven by an intelligent conference terminal, which adopts the following technical solutions: A system for automatically generating meeting minutes driven by an intelligent conference terminal includes a processor, wherein the processor executes the steps of any one of the above-mentioned methods for automatically generating meeting minutes driven by an intelligent conference terminal.

[0021] In summary, the present application includes at least one of the following beneficial technical effects: improving the encryption effect of meeting content, generating a unique list serial number through voiceprint recognition, and combining Chinese-coded shift encryption to make the meeting content exist in encrypted form to avoid leakage of original sensitive information. Realizing the safe and efficient automatic generation and restoration of meeting minutes, the forward encryption and reverse decryption processes correspond to each other, ensuring accurate decryption, and completely restoring the meeting content and identity associations. Optimizing identity feature recognition, ensuring the uniqueness of identity features by resolving feature conflicts and dynamically adjusting weights, and improving the accuracy of encrypted basic data. Dynamically adjusting encryption strategies, such as adjusting the offset value according to the proportion of data volume, and determining the shift direction according to the parity of the total number of serial numbers, to enhance encryption complexity and anti-cracking capabilities. Supporting dynamic management of meeting participants, two time-sharing sorting mechanisms to cope with new personnel, ensuring the stability of the encryption process, and improving the scalability and flexibility of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 The present invention is a step-by-step diagram of a method for automatically generating meeting minutes driven by an intelligent conference terminal.

[0023] Figure 2It is a step diagram for decrypting encrypted stored meeting minutes data to extract and output the original identity list, the relationship between identities and meeting minutes, and the meeting minutes data.

[0024] Figure 3 is a step diagram when the identity feature values ​​are the same. DETAILED DESCRIPTION

[0025] Embodiments of the present application are described in detail below, examples of which are illustrated in the accompanying drawings.

[0026] Throughout this specification, reference to the terms "certain embodiments," "one embodiment," "some embodiments," "illustrative embodiments," "examples," "specific examples," or "some examples" means that the specific features, structures, materials, or characteristics described in conjunction with the embodiments or examples are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0027] The present application embodiment discloses a method for automatically generating meeting minutes driven by an intelligent conference terminal, referring to Figure 1 , including the following steps: Manually enter attendee identity information, including name and title, through the smart conference terminal's interactive interface. Alternatively, automatically synchronize attendee lists through the terminal's integration with external platforms like the enterprise OA system and Party building system, such as importing attendee information for government meetings from the "Three Major Issues and One Important System." The list of identities is stored in an encrypted local partition on the terminal and synchronized to the cloud-based conference service platform, serving as baseline data for subsequent voiceprint matching.

[0028] The intelligent conference terminal uses an omnidirectional microphone array (with a pickup range of 15 meters, supporting clear capture of multiple people speaking in enclosed spaces) to collect conference audio data in real time. The collected audio signal is pre-processed by the terminal processing module for noise reduction and echo cancellation. The small voice model and voiceprint recognition engine trained based on the Guangming large model are then used to extract the voiceprint information corresponding to the identity list: Fundamental frequency information: The fundamental frequency of the speaker's voice is extracted through speech spectrum analysis, reflecting the vibration characteristics of the vocal cords. For example, the fundamental frequency of men is about 85-180Hz, and that of women is about 165-255Hz. Resonance information: Analyze the resonance peak parameters in the sound spectrum, such as the first resonance peak F1 and the second resonance peak F2, to reflect the structural characteristics of the vocal tract.

[0029] The terminal compares the extracted voiceprint information with the voiceprint templates pre-stored in the identity list to establish the correspondence between voiceprints and identities; the voiceprint templates are templates temporarily generated for first-time participants.

[0030] Calculate the identity eigenvalue based on the extracted fundamental frequency information F0 and resonance information (F1, F2), and use a weighted algorithm to calculate the identity eigenvalue: Identity eigenvalue = ω1×F0 + ω2×(F1 + F2) / 2; where the weights ω1 and ω2 are dynamically adjusted by the Guangming large model according to the meeting scenario. For example, when there are multiple people in the meeting, ω1 is increased to strengthen individual differences. Sort the identity eigenvalues of all participants in ascending order to generate non-repeating list numbers, such as 1, 2, 3... N, and the numbers are mapped one by one with the identity list to ensure the uniqueness of subsequent encrypted associations.

[0031] The terminal transmits the collected voice data to the speech-to-text service in real time, calls a high-precision speech recognition model and a preset industry-specific vocabulary library, such as "futures contract" in a financial meeting and "drug ingredients" in a medical meeting, etc., to convert the voice into meeting record data in text form. At the same time, each text record is associated with the corresponding identity list through the voiceprint recognition result, such as marking "[Number 3]: Discuss the annual budget adjustment plan", and the association relationship is stored in the terminal memory.

[0032] Adopt a Chinese coding algorithm with a continuous coding range, such as GB2312 coding, which covers 6,763 common Chinese characters, and convert each character in the meeting record data into the corresponding binary actual code. For example: the character "会" corresponds to the GB2312 code 0xA1A2, which is converted into the actual code "1010000110100010"; professional vocabulary (such as "三重一大") is preferentially matched with the coding in the terminal-built professional vocabulary library to ensure the accuracy of term conversion.

[0033] Based on the previous association relationship, the terminal traverses the meeting record data and performs a shift operation on the actual code of each character according to the following rules: Encrypted code = Actual code + List number × Offset coefficient; where the offset coefficient can be dynamically set according to the meeting confidentiality level. For example, the offset coefficient for a top-secret meeting is 3, and for an ordinary meeting is 1; the shift direction is determined by the terminal communication module in combination with the transmission protocol of the APN encrypted channel, such as determining the direction based on the key characteristics of the SM2 encryption algorithm. For example, the character "议" corresponding to the speaker with number 2 has an actual code of 0xA2B3, and after shifting, the encrypted code 0xA2B5 is generated.

[0034] Query the Chinese coding algorithm table according to the encrypted code to obtain the corresponding text content (i.e., the encrypted record). For example, the encrypted code 0xA2B5 corresponds to the text "屹", which is different from the original text "议". At the same time, based on the association relationship between the identity and the record, the terminal establishes a mapping between the list serial number and the encrypted record, such as "serial number 2 → encrypted paragraph P2", to ensure traceability to the original speaker during decryption.

[0035] The terminal packages the identity list, list serial number, encrypted record, and content correspondence relationship into a meeting minutes data packet, generates a data check value using the SM3 encryption algorithm, and uploads it to the cloud storage through an exclusive channel isolated by the air gap technology. The cloud storage is encrypted using the SM2 algorithm. The terminal supports the "burn after reading" mode. Users can choose to automatically destroy the original voice and coding data cached locally after the meeting, and only retain the encrypted meeting minutes data packet to further improve data security.

[0036] By integrating the hardware capabilities of the intelligent meeting terminal (such as omnidirectional sound pickup and voiceprint collection) with the algorithm advantages of the Guangming large model, such as professional vocabulary matching and dynamic encryption strategies, the full-process encrypted generation of meeting minutes is achieved: the meeting content exists in the form of encrypted codes. Unauthorized access can only obtain scrambled text, such as "议" becoming "屹". It needs to be decrypted by reverse shift in combination with the identity list and list serial number to solve the risk of original content leakage; relying on the professional vocabulary and high-precision speech recognition to ensure the industry adaptability and accuracy of the meeting record, providing a reliable basis for the restoration of the encrypted content; the compatibility with enterprise OA, Party building systems, etc., supports the synchronization of identity lists and the cross-platform transfer of meeting minutes, meeting the high-efficiency collaboration needs of multi-scenario meetings.

[0037] Refer to Figure 2 , the method further includes the following steps: The cached data can be retrieved through the local storage module of the intelligent meeting terminal, or downloaded from the cloud meeting service platform through the APN encrypted channel. The SM2 encryption algorithm is used to verify the terminal permissions during the download process to ensure that only authorized devices can obtain the data. After extracting the identity list, list serial number, encrypted record, and content correspondence relationship from the meeting minutes data, the data structure is parsed relying on the natural language processing ability of the Guangming large model.

[0038] When matching the encrypted code according to the reverse algorithm of the encrypted record and the Chinese coding algorithm, the system will call the preset professional vocabulary library to assist in verification; for example, in a financial meeting, the "期获合约" that appears in the "encrypted record" will preferentially match the reverse coding rule of "期货合约" in the professional vocabulary library to correct the coding matching deviation and improve the accuracy of encrypted code restoration.

[0039] In the process of corresponding the list serial numbers with the encrypted codes according to the encrypted records and content correspondence, secondary verification is carried out in combination with the historical association data of voiceprint recognition (the mapping log of the voiceprints and serial numbers of the participants cached locally on the terminal), to avoid the mismatch between the serial numbers and the codes caused by data transmission errors.

[0040] When reverse-shifting the encrypted codes according to the corresponding list serial numbers, the reverse operation is strictly carried out in accordance with the shifting rules of forward encryption: if the forward encryption is "actual code + list serial number × offset coefficient", then the reverse shift is "encrypted code - list serial number × offset coefficient", and the shifting direction is the same as that of the forward direction. For example, if the forward shift is along the first direction, the reverse shift is also along the first direction. In this process, the terminal processing module will call a high-performance processor for parallel computing to ensure the efficient completion of the reverse shift of a large number of encrypted codes.

[0041] When matching the meeting record data according to the reverse algorithm of the actual code and the Chinese coding algorithm, the professional thesaurus is also linked for semantic correction. For example, the "huiyi" corresponding to the actual code is corrected to "会议" to ensure that the restored text content conforms to the industry expression habits.

[0042] After converting the content correspondence into an association relationship, the system will automatically complete the association format of the identity list and the meeting record data by referring to the preset templates in the meeting service, such as the formal meeting minutes template, such as marking "[Name 3 (Serial number 3)]: Speech content", and finally output through the terminal interaction interface, or exported as formats such as Word and PDF at one key. The exported file is encrypted by the SM3 algorithm to prevent being tampered with during the transmission process. If the "burn after reading" mode is enabled for the meeting, the terminal will automatically delete the temporary data generated during the decryption process after the output, and only retain the original encrypted meeting minutes data.

[0043] This decryption process fully reuses the hardware computing power (such as high-performance processors) and software capabilities (such as professional thesauruses, large model algorithms) of the intelligent meeting terminal. Through strict symmetry with the forward encryption steps, the accurate restoration of the meeting minutes data is achieved, which not only ensures information security but also ensures the professionalism and integrity of the decrypted content, meeting the accuracy and efficiency requirements of different industries for the decryption of meeting data.

[0044] In the actual meeting scenario, there may be special situations where the identity feature values are the same.针对这一情况,本方法设置了严谨的处理机制,方法还包括如下步骤: After calculating the identity feature values, if the system finds that any two identity lists have the same identity feature value, the smart conference terminal immediately initiates the identity feature differentiation process. The terminal first compares the fundamental frequency information corresponding to the two identity features. For example, the high-precision voice analysis module built into the smart conference terminal can accurately measure the fundamental frequency of sound with an accuracy of ±0.1Hz. For example, if Participant A and Participant B have the same identity feature value and the fundamental frequency comparison reveals that Participant A's fundamental frequency is 120Hz and Participant B's is 125Hz, the system will increase the identity feature value of Participant B, whose fundamental frequency value is higher. The specific increase is determined by a preset adjustment factor. For example, if the adjustment factor is 0.05, Participant B's identity feature value will be increased by 0.05 times, thereby distinguishing the two identity feature values.

[0045] If, after comparing the fundamental frequency information, it is found that the values ​​of the identity features corresponding to any two identity lists are still the same, and the corresponding fundamental frequency information values ​​are also exactly the same, the system will further compare the values ​​of the resonance information corresponding to the two features. Intelligent conference terminals use advanced acoustic algorithms to accurately analyze the sound's resonance peak information, and the measurement error of the first resonance peak (F1) and the second resonance peak (F2) can be controlled within ±5Hz. For example, if two participants have the same fundamental frequency and the resonance information is still compared, if participant C has a first resonance peak F1 value of 800Hz and a second resonance peak F2 value of 1800Hz, and participant D has a first resonance peak F1 value of 820Hz and a second resonance peak F2 value of 1850Hz, the system will increase the value of the identity feature of participant D, whose resonance information value is larger. The increase method is also implemented based on the adjustment factor to ensure that the identity feature values ​​are different.

[0046] By first comparing the fundamental frequency information values ​​and increasing the identity feature value of the one with the larger fundamental frequency value; if the fundamental frequencies remain the same, then comparing the resonance information values ​​and increasing the identity feature value of the one with the larger resonance value, this approach effectively ensures unique identification. Only by ensuring that the identity characteristics of each participant are completely different from those of others can the identity characteristics be matched one-to-one with the participant, thus providing a solid foundation for the subsequent generation of list numbers, effectively ensuring the uniqueness and accuracy of the list numbers, and providing reliable guarantees for the smooth operation of the entire automatic generation and encryption process of meeting minutes.

[0047] Reference Figure 3 , the method further comprises the steps of: When calculating identity features, relying on the high-performance processor built into the smart conference terminal and the algorithm support of the Guangming Big Data model, the weighted average formula identity feature = ω1× fundamental frequency information + ω2× resonance information is used for calculation; where ω1 is the fundamental frequency information weight, ω2 is the resonance information weight, and ω1+ω2=1.

[0048] The weight adjustment mechanism is deeply adapted to the product's actual application scenarios: when the number of identity lists changes, the system adjusts the weight of the fundamental frequency information based on the terminal's judgment of the meeting size, such as whether it is a small departmental meeting or a large industry summit based on the number of identity lists. For example, in a small internal corporate meeting with only three people, the fundamental frequency information weight ω1 might be set to 0.3, relying more on resonance information to distinguish individuals. However, in a large government work meeting with 20 people, ω1 would be increased to 0.7. By strengthening the fundamental frequency, a significant individual voice feature, the fundamental frequency difference between men and women is obvious, improving the accuracy of voiceprint recognition in multi-person scenarios, which is consistent with the product's "speaker differentiation" feature.

[0049] The system dynamically adjusts the weight of resonance information based on the total volume of sound data, calculated from the duration and sampling rate of the audio captured by the terminal's microphone array. For short meetings with sparse speech, resulting in low sound data volumes, such as a 10-minute quick huddle, the resonance information weight ω2 may be as low as 0.2, reducing the interference of ambient noise on feature extraction. However, in a two-hour financial institution investment decision-making meeting, as the volume of sound data increases and continuous speech containing extensive discussion and debate occurs, ω2 increases to 0.6. This leverages the stability of resonance information, which reflects the stable characteristics of the vocal tract structure, in long-duration speech, to enhance the anti-interference capability of voiceprint features in complex environments. Combined with the product's high-sensitivity microphone array, this further ensures accurate identity recognition.

[0050] This adaptive weight adjustment mechanism fully combines the product's hardware performance (such as the microphone array's sound pickup capability and the processor's real-time computing capability) with algorithmic advantages (Guangming Big Data's intelligent judgment of conference scenarios), enabling the calculation of identity features to adapt to the needs of meetings of different sizes and to cope with voice environments with different data volumes. It provides reliable basic feature data for the subsequent generation of list numbers and encryption of meeting content, effectively supporting the product's core goal of providing an efficient, accurate, and secure conference experience.

[0051] The method further comprises the steps of: After generating meeting record data, the intelligent conference terminal uses its built-in data analysis module to calculate the percentage of each participant's (corresponding to the roster) text content in the total meeting record data. For example, if participant A's speech accounts for 30% and participant B's accounts for 5%, the terminal then uses the Guangming Big Data model's dynamic encryption strategy module to adjust the shift offset according to anti-correlation rules, forming an encryption formula: "Shift = roster number + shift offset."

[0052] For example, at a financial institution's investment decision-making meeting, if an analyst (number 5 in the identity list) contributes 40% of the speech (a large amount of data), their shift offset might be adjusted to 1, resulting in a final shift of 5 + 1 = 6. Meanwhile, if an observer (number 8) contributes only 5% of the speech (a small amount of data), their shift offset might be set to 5, resulting in a final shift of 8 + 5 = 13. This adjustment mechanism is linked to the product's professional vocabulary: for content with a large proportion of industry terminology, such as "futures contracts" and "risk exposure," a small offset value reduces encoding conversion errors and ensures the accuracy of professional vocabulary after decryption. For smaller, more scattered speeches, a large offset value enhances encryption strength, in line with the product's "high security" advantage.

[0053] At the same time, the terminal's high-performance processor verifies in real time that the shifted encrypted code is within the valid range of the Chinese encoding algorithm, such as the encoding range of GB2312. An anti-correlation adjustment mechanism reduces the probability of code overflows caused by large offsets: core content with a lot of talk has a small offset, reducing the risk of over-range. Even if the content with few speakers has a large offset, the overall over-range amount is controllable due to its small data volume. This design not only meets the product's performance requirement of "supporting 20 conference rooms concurrently," but also improves the overall security and accuracy of meeting minutes through dynamic differentiated encryption. Together with features such as "burn after listening" and "exclusive security encryption," it builds a multi-layered data protection system.

[0054] The method further comprises the steps of: When temporary participants arrive during a meeting, such as department heads joining an internal company meeting or newly added observers at a government meeting, the smart conference terminal uses real-time synchronization with external platforms such as the company's OA system and Party building system to retrieve the newly added identity list and trigger a time-sharing sorting mechanism. The terminal's processing module first locks onto the meeting data for the first period of time. Based on the current list of M identities, such as the original 10 attendees, it calls the Guangming Big Model's meeting minutes generation engine and combines it with a preset template to generate the first set of meeting minutes data. This data includes the voiceprint information, encrypted records, and content corresponding to each of the M identities. It is then synchronized to cloud storage via an APN encrypted channel. Local caching also supports temporary storage in a "burn after listening" mode.

[0055] After entering the second period, the terminal initiates a separate voiceprint recognition process for the N newly added identities, for example, three new people. This process uses a high-sensitivity microphone array to collect voice data from the newly added individuals, extract fundamental frequency and resonance information, and calculate identity signatures. The newly added identities are then numbered using the "largest original number + 1" rule. For example, if the original numbers were 1-10, the newly added numbers would be 11-13, automatically concatenating the original M numbers + the N newly added numbers. The terminal then generates a second set of meeting minutes based on the newly added individuals' speeches. The encryption process reuses the Chinese encoding algorithm and shifting rules, adjusting the offset based on the proportion of the newly added individuals' speech data to ensure consistent encryption with the first set of minutes.

[0056] This time-sharing sorting mechanism fully adapts to the hardware performance and software architecture of the product: on the one hand, relying on the parallel computing capability of the terminal's high-performance processor, it can simultaneously process the generation of two copies of the minutes data, old and new, without affecting the real-time transcription of the meeting. The performance of supporting concurrent processing of 20 conference rooms can ensure the parallel processing of data in multiple time periods in a single conference room; on the other hand, through the cross-terminal synchronization function of the cloud-based conference service platform, the two copies of the minutes data can be associated and stored in the terminal and external systems (such as OA, party building systems). Users can subsequently merge the two minutes with one click through the terminal interface. During the merging process, the system will automatically verify the continuity of the serial number (1-13 without duplication) and the consistency of the encryption code to ensure the integrity of the meeting records.

[0057] This mechanism effectively copes with the scenario of dynamic personnel adjustments in meetings. It avoids the interference of new personnel on the original encryption process, such as the encryption codes of the original M identities do not need to be recalculated, and ensures the accuracy of identity association through serial number connection. It works together with the product's "automatic transmission of cross-system meeting minutes" and "exclusive security encryption" functions to improve the scalability and flexibility of the system in complex meeting scenarios, and meet the needs of different industries for dynamic meeting management.

[0058] The method further comprises the steps of: When new participants appear temporarily during a meeting, such as risk control experts joining a financial institution's meeting or subject leaders attending an educational institution's seminar, the smart conference terminal can automatically obtain the list of N newly added identities and trigger a time-sharing sorting mechanism through real-time data synchronization with external platforms such as the enterprise OA system and party building system.

[0059] During the first period, the terminal leverages the parallel computing power of its high-performance processor to adjust the sequence numbers of the existing M identities. For example, if the "amplify N+1" strategy is used, for example, if the original sequence numbers were 1-5 (M = 5 participants), and N = 4 participants are added, the sequence numbers are adjusted to 5, 10, 15, 20, and 25. Alternatively, if the "shift N positions back" strategy is used, the original sequence numbers 1-5 are adjusted to 5-9, leaving space for the new participants. During this process, the terminal simultaneously generates meeting record snippets corresponding to the current M identities, performs preliminary encryption using a professional vocabulary, and employs a Chinese encoding algorithm and shift encryption rules. These snippets are then temporarily stored in the cloud via an APN encrypted channel, ensuring that the existing encryption process is not disrupted by the addition of new participants.

[0060] After entering the second period, the terminal initiates the voiceprint collection and processing process for N newly added identities: The omnidirectional microphone array captures the voice data of the newly added individuals, extracts the fundamental frequency and resonance information, and uses the Guangming Large Model voice model to calculate the identity feature value and generate a sequence number, such as 1-4, which is inserted into the reserved slots. After the newly added individuals' speech is transcribed into text in real time, dynamic differential encryption is performed according to their sequence numbers, and the shift offset value is adjusted according to the data volume ratio to form an independent encrypted record segment.

[0061] Ultimately, the terminal reorders the serial numbers of the M original identities and the N newly added identities from smallest to largest, for example, merging them into 1-9. It then calls the template engine in the conference service to integrate the encrypted record fragments of the two time periods into the complete meeting minutes data. During the integration process, the system verifies data integrity using the SM3 hash algorithm to ensure that the old and new serial numbers are continuous and non-repeating. For example, 5, 10, 15... are reorganized with 1-4 to become 1-9, and the shift rules of the encryption code remain consistent. The generated complete minutes can be exported to formats such as Word and PDF with one click, or transmitted to OA, Party building, and other systems through the "Industry and Digital Integration Platform." It also supports the activation of the "burn after listening" function to destroy temporary data.

[0062] This mechanism fully utilizes the real-time processing capabilities of the terminal and the collaborative storage advantages of the cloud. When responding to dynamic adjustments in participants, it not only ensures the stability of the original encrypted data without recalculating historical codes, but also enables seamless access for new personnel through serial number reservation. It is deeply adapted to the product's "high concurrency processing" and "exclusive security encryption" features, significantly improving the system's scalability and encryption flexibility in complex conference scenarios.

[0063] The method further comprises the steps of: Before performing the encoding shift encryption operation, the intelligent conference terminal will first count the total number of the list serial numbers of the current participants, that is, the sum of M+N. M is the initial number of participants, and N is the number of new participants. The parity attribute is judged through the logical operation of the terminal processing module. For example, in the regular meeting of an enterprise department, if 5 people initially participate in the meeting (list serial numbers 1-5) and there are no new participants in the middle, the total is 5 (odd), then the first direction offset is triggered (such as the encoding value increasing direction); if 8 people initially participate in the government agency meeting and 2 new people are added in the middle, the total is 10 (even), then the second direction offset is started, such as the encoding value decreasing direction.

[0064] This mechanism is deeply integrated with the encryption system of the product: the shift rules in the first direction and the second direction are both designed based on the effective range of the Chinese encoding algorithm (such as GB2312), and cooperate with the SM2 and SM3 algorithm systems in the exclusive security encryption to form a multi-layer encryption logic. For example, when the total is odd, the actual encoding of the word "会议" is mapped to "讳仪" after the first direction offset; when the total is even, it is mapped to "佘议" after the second direction offset. The encryption results in the two directions have no regular association, greatly increasing the difficulty of unauthorized cracking.

[0065] At the same time, the high-performance processor of the terminal will real-time verify whether the encrypted encoding after the shift exceeds the encoding range, and combine with the dynamic differential encryption strategy. Those who speak more have a smaller offset value, ensuring that the shifts in different directions are all executed within the safe range. This dynamic direction adjustment based on the parity of the total number takes full advantage of the real-time computing power of the Guangming large model and complements the "burn after reading" function - the former enhances the anti-cracking property of the encryption process, and the latter eliminates the leakage risk from the source of the data life cycle, jointly building a protection system that meets the requirements of high-security level meetings and effectively supporting the application of the product in confidential scenarios such as finance and government.

[0066] The embodiment of the present application also discloses a meeting minutes automatic generation system driven by an intelligent conference terminal, including a processor, and the processor executes the steps of the meeting minutes automatic generation method driven by the intelligent conference terminal as described in any one of the above.

[0067] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and should not be construed as limitations to the present application. Those of ordinary skill in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present application.

Claims

1. A method for automatically generating meeting minutes driven by an intelligent conference terminal, characterized in that: The steps include: Get the list of identities attending the meeting; collecting voice data of the meeting in real time, and identifying voiceprint information corresponding to the identity list from the voice data, the voiceprint information including fundamental frequency information and resonance information; Calculating identity features based on the fundamental frequency information and the resonance information, and sorting the identity features to obtain list numbers corresponding to the identity lists, wherein the list numbers are non-repeated; wherein the identity features are calculated using a weighted average method based on the fundamental frequency information and the resonance information; The weight of the fundamental frequency information is adjusted in a positive correlation with the number of the identity lists; the smaller the number of the identity lists, the smaller the weight of the fundamental frequency information; the larger the number of the identity lists, the greater the weight of the fundamental frequency information; The weight of the resonance information is adjusted in a positive correlation according to the total data volume of the sound data; the smaller the total data volume of the sound data, the smaller the weight of the resonance information; and the larger the total data volume of the sound data, the larger the weight of the resonance information; Recognizing the sound data as meeting record data containing text, and associating the identity list with the meeting record data; Converting the text content in the conference record data into actual codes according to a Chinese coding algorithm with a continuous coding range; Based on the association relationship, the identity list is traversed, and the actual code of the meeting record data is shifted according to the list serial number corresponding to the identity list to obtain an encrypted code; Matching a corresponding encrypted record from the Chinese encoding algorithm according to the encryption code; Obtaining a content correspondence between the list number and the encrypted record according to the association relationship; The identity list, the list serial number, the encrypted record and the corresponding relationship between the content are stored as meeting minutes data.

2. The method for automatically generating meeting minutes driven by an intelligent conference terminal according to claim 1, characterized in that: The method further comprises the steps of: Obtaining the meeting minutes data; Extracting the identity list, the list serial number, the encrypted record and the corresponding relationship between the content from the meeting minutes data; Matching the encryption code according to the encryption record and the inverse algorithm of the Chinese encoding algorithm; Matching the list number with the encryption code according to the corresponding relationship between the encryption record and the content; Reverse shifting the encrypted code according to the corresponding list number to obtain the actual code; Matching the meeting record data according to the actual encoding and the inverse algorithm of the Chinese encoding algorithm; Converting the content correspondence into the association relationship; Output the identity list, the association relationship and the meeting record data.

3. The method for automatically generating meeting minutes driven by an intelligent conference terminal according to claim 1 or 2, characterized in that: The method further comprises the steps of: If the identity feature values ​​of any two corresponding identity lists are the same, then compare the values ​​of the fundamental frequency information corresponding to the two identity features, and increase the value of the identity feature with the larger fundamental frequency information value; If the identity feature values ​​of any two corresponding identity lists are the same and the corresponding fundamental frequency information values ​​are also the same, then the values ​​of the resonance information corresponding to the two features are compared and the value of the identity feature with the larger resonance information value is increased.

4. The method for automatically generating meeting minutes driven by an intelligent conference terminal according to claim 1, characterized in that: The method further comprises the steps of: According to the proportion of the text content corresponding to the identity list in the meeting record data, the shift offset value of the identity list is adjusted in an anti-correlation manner; the smaller the proportion of the data volume, the larger the shift offset value of the identity list; the larger the proportion of the data volume, the smaller the shift offset value of the identity list.

5. The method for automatically generating meeting minutes driven by an intelligent conference terminal according to claim 1, characterized in that: The method further comprises the steps of: If the identity list is newly added during the process of generating the meeting record data, it is sorted by time; Generating a meeting minutes data corresponding to the current M identity lists within a first period of time; During the second period of time, another meeting minutes data corresponding to the newly added N identity lists is generated, wherein the newly added list serial number corresponds to the sorting position + M.

6. The method for automatically generating meeting minutes driven by an intelligent conference terminal according to claim 1, characterized in that: The method further comprises the steps of: If N new identity lists are added during the process of generating the meeting record data, then the lists are sorted by time; Generating the list serial numbers of the current M identity lists within a first time period, wherein the sorting positions of the M identity lists are magnified N+1 times; Generate the list serial numbers of N newly added identity lists within the second period of time; Insert the current M identity lists into the newly added N identity lists, re-sort them according to the corresponding list serial numbers, and then generate the meeting minutes data.

7. The method for automatically generating meeting minutes driven by an intelligent conference terminal according to claim 1, characterized in that: The method further comprises the steps of: Obtaining an even-odd attribute according to the total amount of the list numbers; If the parity attribute is an odd number, the shift is shifted along the first direction; otherwise, the shift is shifted along the second direction.

8. A system for automatically generating meeting minutes driven by an intelligent conference terminal, characterized in that: The method comprises a processor, wherein the processor executes the steps of the method for automatically generating meeting minutes driven by an intelligent conference terminal as described in any one of claims 1 to 7.

Images