CN120597331B - Intelligent conference terminal-driven conference summary automatic generation method and system - Google Patents
Intelligent conference terminal-driven conference summary automatic generation method and systemInfo
- Publication number
- CN120597331B CN120597331B CN202511093640.6A CN202511093640A CN120597331B CN 120597331 B CN120597331 B CN 120597331B CN 202511093640 A CN202511093640 A CN 202511093640A CN 120597331 B CN120597331 B CN 120597331B
- Authority
- CN
- China
- Prior art keywords
- identity
- list
- data
- conference
- meeting
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F21/00—Security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
- G06F21/60—Protecting data
- G06F21/62—Protecting access to data via a platform, e.g. using keys or access control rules
- G06F21/6218—Protecting access to data via a platform, e.g. using keys or access control rules to a system of files or objects, e.g. local or distributed file system or database
- G06F21/6245—Protecting personal data, e.g. for financial or medical purposes
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F21/00—Security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
- G06F21/60—Protecting data
- G06F21/602—Providing cryptographic facilities or services
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/10—Text processing
- G06F40/12—Use of codes for handling textual entities
- G06F40/126—Character encoding
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/10—Office automation; Time management
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L15/00—Speech recognition
- G10L15/26—Speech to text systems
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L17/00—Speaker identification or verification techniques
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- General Engineering & Computer Science (AREA)
- Bioethics (AREA)
- Business, Economics & Management (AREA)
- Entrepreneurship & Innovation (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- Computer Security & Cryptography (AREA)
- Software Systems (AREA)
- Computer Hardware Design (AREA)
- Computational Linguistics (AREA)
- Multimedia (AREA)
- Acoustics & Sound (AREA)
- Human Computer Interaction (AREA)
- Marketing (AREA)
- Artificial Intelligence (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- Quality & Reliability (AREA)
- Operations Research (AREA)
- Economics (AREA)
- Data Mining & Analysis (AREA)
- Medical Informatics (AREA)
- Databases & Information Systems (AREA)
- Telephonic Communication Services (AREA)
Abstract
The application relates to the technical field of intelligent conference terminals and discloses an automatic conference summary generation method and system driven by an intelligent conference terminal; the method comprises the steps of obtaining a participant identity list, collecting sound data and extracting voiceprint features, calculating the identity features and generating unique list serial numbers, converting the sound data into a text conference record and associating identity information, converting the conference record into actual codes by adopting a continuous Chinese coding algorithm, carrying out dynamic shift encryption on the codes based on the list serial numbers to generate encrypted records, and storing the identity list, the serial numbers, the encrypted records and corresponding relations to form conference descriptions. The system supports dynamic coding adjustment of newly added participants in the conference process, and optimizes the encryption effect by means of weighted voiceprint feature calculation, data volume duty ratio anti-correlation adjustment offset value and the like. And when decryption is carried out, the original conference record is restored through a reverse algorithm, so that the summary automatic generation and high-efficiency sharing are realized while the content safety is ensured.
Description
Technical Field
The application relates to the technical field of intelligent conference terminals, in particular to an automatic conference summary generation method and system driven by an intelligent conference terminal.
Background
Along with acceleration of enterprise digitization process, cross-region cooperation is increased, requirements on real-time processing, accurate retention and rapid sharing of conference content are urgent, intelligent transformation can realize automatic generation of conference summary, and support decision is rapid to land.
The intelligent transformation of the current conference scene is gradually applied to the existing technologies such as intelligent conference terminals, voice recognition, natural language processing and the like. The terminal can realize real-time speech transcription and multi-person speech discrimination, and part of the system can automatically extract conference key points and generate a primary summary frame. Meanwhile, the cloud platform supports conference data storage and cross-terminal sharing, and the AI algorithm is optimized to enable the summary generation accuracy to be improved.
The encryption of the existing conference intelligent scheme aims at the data transmission and storage links, and the conference content is not encrypted and modified. Sensitive information exists in an original form, and once a terminal or a platform is broken, the information is easy to leak, and the encryption effect is limited.
Disclosure of Invention
In order to improve the encryption effect of conference contents, the application provides an automatic generation method and system for conference summary driven by an intelligent conference terminal.
In a first aspect, the present application provides a method for automatically generating a meeting summary driven by an intelligent meeting terminal, which adopts the following technical scheme:
an automatic conference summary generation method driven by an intelligent conference terminal comprises the following steps:
acquiring an identity list of participating in a conference;
collecting voice data of a conference in real time, and identifying voiceprint information corresponding to the identity list from the voice data, wherein the voiceprint information comprises fundamental frequency information and resonance information;
calculating identity characteristics according to the fundamental frequency information and the resonance information, and sequencing the identity characteristics to obtain list numbers corresponding to the identity lists one by one, wherein the list numbers are not repeated;
identifying the voice data into conference record data with text content, and associating the identity list with the conference record data;
according to a Chinese coding algorithm with continuous coding range, converting the text content in the conference record data into actual codes;
Traversing the identity list based on the association relation, and shifting the actual codes of the conference record data according to the list serial numbers corresponding to the identity list to obtain encrypted codes;
matching corresponding encrypted records from the Chinese coding algorithm according to the encrypted codes;
obtaining the content corresponding relation between the list serial number and the encrypted record according to the association relation;
and storing the identity list, the list serial number, the encryption record and the content corresponding relation as meeting summary data.
By adopting the technical scheme, the identity list participating in the conference is acquired, the voice print information containing fundamental frequency information and resonance information is identified by combining the conference voice data acquired in real time, the identity features are calculated and sequenced to obtain the unique list serial number, an identity association foundation is laid for encryption, after the conference record data identified by the voice data are associated with the identity list, a Chinese coding algorithm with continuous coding range is converted into actual coding, the actual coding is shifted to obtain encryption coding based on the list serial number, the encryption record is matched, the content corresponding relation between the list serial number and the encryption record is established, and finally the identity list, the list serial number, the encryption record and the content corresponding relation are stored as conference summary data. The whole process enables conference contents to exist in an encryption coding mode instead of original words, decryption can be achieved only through corresponding identities and serial numbers, if data are acquired by other people, each word in the converted word contents is an offset word, finally, the error contents are obtained, sensitive information leakage when a terminal or a platform is broken is avoided, and encryption effects of the conference contents are greatly improved.
Optionally, the method further comprises the steps of:
Acquiring the meeting summary data;
Extracting the identity list, the list serial number, the encryption record and the content corresponding relation from the conference summary data;
matching the encryption code according to the encryption record and the reverse algorithm of the Chinese coding algorithm;
Corresponding the list serial number to the encryption code according to the corresponding relation between the encryption record and the content;
reversely shifting the encryption code according to the corresponding list sequence number to obtain the actual code;
matching the conference record data according to the actual coding and the reverse algorithm of the Chinese coding algorithm;
Converting the content corresponding relation into the association relation;
And outputting the identity list, the association relation and the conference record data.
By adopting the technical scheme, the elements such as the identity list and the list serial number are extracted, the encrypted record is restored to be encrypted codes by utilizing the reverse algorithm of the Chinese coding algorithm, the actual codes are obtained by reverse shift of the list serial number, and finally the original conference record data are matched and the association relation is converted. The process strictly follows the reverse operation of the forward encryption path, ensures the accuracy and reliability of the decryption process, completely restores the original conference content and identity association, and realizes the safe restoration of the conference summary data.
Optionally, the method further comprises the steps of:
If the identity characteristic values of any two corresponding identity lists are the same, comparing the values of the fundamental frequency information corresponding to the two identity characteristics, and increasing the value of the identity characteristic with larger value of the fundamental frequency information;
If the identity characteristic values of any two corresponding identity lists are the same, and the corresponding values of the fundamental frequency information are the same, comparing the values of the resonance information corresponding to the two characteristics, and increasing the value of the identity characteristic with larger value of the resonance information.
By adopting the technical scheme, when the identity characteristic values are the same, the fundamental frequency information values are compared, the identity characteristic value of the larger fundamental frequency value is improved, and when the fundamental frequency is still the same, the resonance information values are compared, and the resonance value of the larger fundamental frequency value is improved. The uniqueness of the identity identification is ensured through the differentiation processing of the values of the identity features, so that the identity features are in one-to-one correspondence with the participants, and the uniqueness and the accuracy of the list serial numbers are ensured.
Optionally, the method further comprises the steps of:
According to the fundamental frequency information and the resonance information, calculating an identity characteristic by adopting a weighted average method;
The method comprises the steps of adjusting the weight of fundamental frequency information according to positive correlation of the number of identity lists, wherein the smaller the number of the identity lists is, the smaller the weight of the fundamental frequency information is, and the larger the number of the identity lists is, the larger the weight of the fundamental frequency information is;
The weight of the resonance information is adjusted according to positive correlation of the total data amount of the sound data, the weight of the resonance information is smaller as the total data amount of the sound data is smaller, and the weight of the resonance information is larger as the total data amount of the sound data is larger.
By adopting the technical scheme, the fundamental frequency information weight is increased when the number of the participants is increased, the individual characteristic distinction degree is improved, the resonance information weight can be improved by increasing the total data volume of the sound data, the characteristic stability in a complex environment is enhanced, and the two weights can be adaptively adjusted according to the actual condition of the conference.
Optionally, the method further comprises the steps of:
And according to the data volume ratio of the text content corresponding to the identity list in the conference record data, adjusting the shift offset value of the identity list in an opposite relation mode, wherein the smaller the data volume ratio is, the larger the shift offset value of the identity list is, and the larger the data volume ratio is, the smaller the shift offset value of the identity list is.
By adopting the technical scheme, the shift = list serial number + shift offset value is adjusted according to the data volume of the content corresponding to the identity, the smaller the shift of encryption of a plurality of speaking parties is, the dynamic differential encryption is realized, the probability and the quantity of the shift out-of-range are reduced, and the accuracy of the whole meeting summary is improved.
Optionally, the method further comprises the steps of:
If the identity list is newly added in the process of generating the conference record data, time-sharing sequencing is performed;
generating one meeting summary data corresponding to the current M identity lists in a first period of time;
And generating another conference summary data corresponding to the newly added N identity lists in a second period of time, wherein the newly added list number corresponds to the sequencing position +M.
By adopting the technical scheme, the time-sharing sequencing mechanism realizes dynamic personnel management, namely when a participant is newly added, meeting summary data generation of the current M identity lists is finished, then the newly added N identities are processed independently, the sequencing positions of the N identities are automatically connected with the original sequence, the interference of dynamic adjustment of the participant on the encryption flow is reduced, the independent encryption of the data of the new participant and the old participant and the continuous serial numbers are ensured, and the flexibility of system expansibility and meeting content encryption is improved.
Optionally, the method further comprises the steps of:
if N identity lists are newly added in the process of generating the conference record data, sorting in a time-sharing way;
generating list serial numbers of the current M identity lists in a first period of time, wherein the sorting positions of the M identity lists are amplified by n+1 times;
generating the list serial numbers of the newly added N identity lists in the second period of time;
And inserting the current M identity lists into the newly added N identity lists, and re-ordering according to the corresponding list serial numbers to generate the meeting summary data.
By adopting the technical scheme, the identity lists corresponding to newly added participants are processed through a time-sharing sequencing and sequence number reservation mechanism, the sequence numbers of the current M lists are amplified by N times to form the newly added N sequence numbers, and finally the reorganization sequencing is inserted, so that the interference of dynamic adjustment of the participants on the encryption flow is reduced, the data of the new and old participants are ensured to be encrypted independently, the sequence numbers are continuous, and the flexibility of system expansibility and conference content encryption is improved.
Optionally, the method further comprises the steps of:
Obtaining odd-even attributes according to the total amount of the list serial numbers;
If the parity attribute is odd, the shift is offset along a first direction, otherwise the shift is offset along a second direction.
By adopting the technical scheme, the shift directions are distinguished through the odd-even attribute of the total list sequence number, and the odd number is in the first direction and the even number is in the second direction. The randomness and the complexity of encryption code shift are increased, the regularity in a single direction is avoided, the anti-cracking capability of conference content encryption is improved, and the information security is enhanced.
In a second aspect, the application provides an intelligent conference terminal-driven conference summary automatic generation system, which adopts the following technical scheme:
An intelligent conference terminal driven conference summary automatic generation system comprises a processor, wherein the processor executes the steps of the intelligent conference terminal driven conference summary automatic generation method.
In summary, the application has at least one of the following beneficial technical effects that the conference content encryption effect is improved, the unique list serial number is generated through voiceprint recognition, and the conference content exists in an encryption mode by combining Chinese coding shift encryption, so that the leakage of original sensitive information is avoided. The method has the advantages that safe and efficient automatic generation and restoration of meeting summary are realized, forward encryption corresponds to reverse decryption flow, decryption accuracy is guaranteed, and meeting content and identity association are completely restored. Optimizing identity feature recognition, ensuring unique identity features by solving feature conflict and dynamically adjusting weights, and improving accuracy of encrypted basic data. And dynamically adjusting an encryption strategy, such as adjusting an offset value according to the data volume ratio, and determining the shift direction according to the sequence total quantity parity, so as to enhance encryption complexity and anti-cracking capability. The dynamic management of the participants is supported, the two time-sharing sequencing mechanisms are used for coping with newly-added personnel, the stability of the encryption flow is ensured, and the expansibility and the flexibility of the system are improved.
Drawings
Fig. 1 is a step diagram of an intelligent conference terminal driven conference summary automatic generation method.
Fig. 2 is a step diagram of decrypting encrypted stored meeting summary data to extract and output an original identity list, an association relationship between an identity and a meeting record, and meeting record data.
Fig. 3 is a step diagram when identity feature values are the same.
Detailed Description
Embodiments of the present application are described in detail below, examples of which are illustrated in the accompanying drawings.
In the description of the present specification, reference to the terms "certain embodiments," "one embodiment," "some embodiments," "an exemplary embodiment," "an example," "a particular example," or "some examples" means that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the present application. In this specification, schematic representations of the above terms do not necessarily refer to the same embodiments or examples. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.
The embodiment of the application discloses an automatic conference summary generation method driven by an intelligent conference terminal, which comprises the following steps with reference to fig. 1:
The identity information of the participants is manually input through the interactive interface of the intelligent conference terminal, including names, job titles and the like, or the lists of the participants are automatically synchronized through the docking function of the terminal and external platforms such as an enterprise OA system, a party building system and the like, for example, the participant information of the government conference is imported from a triple-large system. The identity list is stored in a local encryption partition of the terminal and is synchronized to a cloud conference service platform to serve as reference data for subsequent voiceprint matching.
The intelligent conference terminal collects conference sound data in real time through an omnidirectional pickup microphone array (the pickup range covers 15 meters, a plurality of people in a closed place are supported to be clearly collected), collected audio signals are subjected to noise reduction and echo elimination pretreatment through a terminal processing module, a small voice model and a voiceprint recognition engine based on the training of a large voice model are called, and voiceprint information corresponding to an identity list is extracted:
Fundamental frequency information, namely extracting the fundamental frequency of the sound of a speaker through voice spectrum analysis, reflecting the vibration characteristics of a vocal cord, such as about 85-180Hz of male fundamental frequency and about 165-255Hz of female fundamental frequency;
and resonance information, namely analyzing formant parameters in the sound frequency spectrum, such as a first formant F1 and a second formant F2, and reflecting the vocal tract structural characteristics.
The terminal compares the extracted voiceprint information with a voiceprint template pre-stored in an identity list to establish a correspondence between voiceprints and identities, wherein the voiceprint template is a template temporarily generated by a first participant.
And calculating an identity characteristic value based on the extracted fundamental frequency information F0 and resonance information (F1 and F2), and calculating the identity characteristic value by adopting a weighting algorithm, wherein the identity characteristic value=ω1×F0+ω2× (F1+F2)/2, and the weights ω1 and ω2 are dynamically adjusted by a light model according to conference scenes, such as increasing ω1 during a multi-person conference to strengthen individual differences. And (3) carrying out ascending order sequencing on the identity characteristic values of all the participants to generate non-repeated list serial numbers, wherein the serial numbers are mapped with the identity lists one by one, and the uniqueness of subsequent encryption association is ensured.
The terminal transmits the collected sound data to a voice transcription service in real time, and invokes a high-precision voice recognition model and a preset industry professional word stock, such as futures contracts in financial conferences, medicine components in medical conferences and the like, so as to convert the voice into conference record data in a text form. And meanwhile, each text record is associated with a corresponding identity list through a voiceprint recognition result, for example, a label of [ serial number 3]: discussion annual budget adjustment scheme "is marked, and the association relationship is stored in a terminal memory.
And a Chinese coding algorithm with continuous coding range, such as GB2312 coding, which covers 6763 common Chinese characters, is adopted to convert each character in the conference record data into a corresponding binary practical code. For example, the GB2312 code corresponding to the word "Cone" is converted into 0xA1A2 and converted into the actual code "1010000110100010", and the professional vocabulary (such as "Sanyi Dada") is preferentially matched with the professional word stock code built in the terminal, so that the accuracy of term conversion is ensured.
Based on the association relation, the terminal traverses the conference record data, and performs a shift operation on the actual code of each text according to the following rule, namely, encryption code = actual code + list serial number x offset coefficient, wherein the offset coefficient can be dynamically set according to the conference security level, such as 3 is the offset coefficient of the conference, 1 is the common conference, and the shift direction is determined by a terminal communication module in combination with the transmission protocol of an APN encryption channel, such as the key feature determination direction based on an SM2 encryption algorithm. For example, the word "conference" corresponding to the speaker of sequence number2 actually encodes 0xA2B3, and the encryption code 0xA2B5 is generated after shifting.
And reversely inquiring the Chinese coding algorithm table according to the encryption coding to obtain the corresponding text content (namely the encryption record). For example, the encryption code 0xA2B5 corresponds to the word "stand" and differs from the original word "proposal". Meanwhile, the terminal establishes mapping between the list serial number and the encrypted record, such as 'serial number 2- & gt encrypted paragraph P2', based on the association relation between the identity and the record, so as to ensure that the decryption can be traced back to the original speaker.
The terminal packages the identity list, the list serial number, the encryption record and the content corresponding relation into a conference summary data packet, generates a data check value by adopting an SM3 encryption algorithm, uploads the data check value to a cloud storage through a dedicated channel isolated by a gatekeeper technology, and the cloud storage is encrypted by adopting an SM2 algorithm. The terminal supports a burn after hearing mode, a user can automatically destroy locally cached original voice and coded data after a conference is finished, only encrypted summary data packets are reserved, and the data security is further improved.
The full-flow encryption generation of meeting summary is realized by integrating the hardware capability (such as omnidirectional pickup and voiceprint collection) of the intelligent meeting terminal and the algorithm advantages of a bright large model, such as professional word stock matching and dynamic encryption strategy, wherein meeting content exists in an encryption coding mode, unauthorized access can only acquire disordered characters, such as ' meeting ' is changed into ' meeting, decryption can be carried out through reverse shift only by combining an identity list and a list serial number, the risk of original content leakage is solved, professional word stock and high-precision voice recognition are relied, industry suitability and accuracy of meeting records are ensured, a reliable basis is provided for restoring encrypted content, compatibility with enterprise OA, party-built systems and the like is provided, synchronization of the identity list and cross-platform circulation is supported, and the high-efficiency cooperation requirement of multi-scene meetings is met.
Referring to fig. 2, the method further comprises the steps of:
The cached data can be called through a local storage module of the intelligent conference terminal, or downloaded from a cloud conference service platform through an APN encryption channel, and the terminal authority is verified by adopting an SM2 encryption algorithm in the downloading process, so that only authorized equipment can be ensured to be acquired. And after the corresponding relation of the identity list, the list serial number, the encryption record and the content is extracted from the conference summary data, analyzing the data structure by depending on the natural language processing capability of the Guangming big model.
When the encryption record and the reverse algorithm of the Chinese coding algorithm are matched and encrypted, the system can call a preset professional word stock to assist in verification, for example, a 'term acquisition contract' appearing in the 'encryption record' in a financial conference can be preferentially matched with the reverse coding rule of a 'futures contract' in the professional word stock, the matching deviation of the codes is corrected, and the accuracy of restoring the encrypted codes is improved.
And in the link of corresponding list serial numbers and encryption codes according to the corresponding relation between the encryption record and the content, carrying out secondary verification by combining historical associated data (the voice print of the participant and the serial number mapping log of the local cache of the terminal) of voice print identification, and avoiding mismatch between serial numbers and codes caused by data transmission errors.
When the encryption codes are reversely shifted according to the corresponding list serial numbers, reverse operation is strictly followed by a shifting rule of forward encryption, namely, if the forward encryption is 'actual coding+list serial numbers multiplied by offset coefficients', the reverse shift is 'encryption coding-list serial numbers multiplied by offset coefficients', the shifting direction is consistent with the forward direction, and if the forward direction is shifted along the first direction, the reverse direction is also shifted back along the first direction. In the process, the terminal processing module can call the high-performance processor to perform parallel calculation, so that the efficient completion of reverse shift of a large number of encryption codes is ensured.
When the reverse algorithm of the actual coding algorithm and the Chinese coding algorithm is matched with the conference record data, the professional word stock is linked to carry out semantic modification, for example, a conference instrument corresponding to the actual coding is modified to be a conference, and the restored text content is ensured to accord with the industry expression habit.
After the corresponding relation of the content is converted into the association relation, the system can refer to a preset template in the conference service, such as a formal conference summary template, automatically complement the association format of the identity list and the conference record data, such as marking the speaking content, namely 'name 3 (serial number 3'), and finally output the content through a terminal interaction interface, or export the content into formats such as Word and PDF through one key, and the exported file is encrypted by adopting an SM3 algorithm to prevent tampering in the transmission process. If the meeting starts the burn after hearing mode, the terminal automatically clears the temporary data generated in the decryption process after the meeting is output, and only the original encrypted meeting summary data is reserved.
The decryption process completely multiplexes hardware computing power (such as a high-performance processor) and software capability (such as a professional word stock and a large model algorithm) of the intelligent conference terminal, and realizes accurate restoration of conference summary data through strict symmetry with a forward encryption step, so that information safety is ensured, the professionality and the integrity of decrypted content are ensured, and the requirements of different industries on accuracy and high efficiency of conference data decryption are met.
In an actual conference scenario, special cases may occur where the identity feature values are the same. The method sets a strict processing mechanism aiming at the situation, and the method further comprises the following steps:
After the system calculates the identity characteristic value, if the identity characteristic values corresponding to any two identity lists are found to be the same, at the moment, the intelligent conference terminal immediately starts the identity characteristic differentiation processing flow. The terminal will compare the values of the fundamental frequency information corresponding to the two identity features. Taking a high-precision voice analysis module built in the intelligent conference terminal as an example, the module can accurately measure the fundamental frequency of sound, and the precision of the module can reach +/-0.1 Hz. Assuming that the identity characteristic values of the participant a and the participant B are the same, when the fundamental frequency information is compared, if the fundamental frequency information of the participant a is found to be 120Hz, and the fundamental frequency information of the participant B is found to be 125Hz, the system will increase the value of the identity characteristic of the participant B with larger value of the fundamental frequency information, the specific increase amplitude can be determined according to the preset adjustment factor of the system, for example, the adjustment factor is set to be 0.05, and the identity characteristic value of the participant B will be increased by 0.05 times on the original basis, thereby realizing the distinction of the identity characteristic values of the participant a and the participant B.
If the values of the identity features corresponding to any two identity lists are found to be the same after the fundamental frequency information is compared, and the values of the corresponding fundamental frequency information are completely consistent, the system further compares the values of the resonance information corresponding to the two features. The intelligent conference terminal can accurately analyze the formant information of sound through an advanced acoustic algorithm, and the measurement error of the first formant (F1) and the second formant (F2) can be controlled within +/-5 Hz. For example, when the fundamental frequencies of two participants are the same, the resonance information is continuously compared, if the first formant F1 value of the participant C is 800Hz, the second formant F2 value is 1800Hz, the first formant F1 value of the participant D is 820Hz, the second formant F2 value is 1850Hz, the system will increase the value of the identity feature of the participant D with larger value of the resonance information, and the increase mode is executed according to the adjustment factor, so as to ensure that the difference of the identity feature values occurs.
The identity characteristic value of the larger fundamental frequency value is improved by comparing the fundamental frequency information values, and if the fundamental frequency is still the same, the identity characteristic value of the larger resonance value is improved by comparing the resonance information values, so that the uniqueness of the identity identification can be effectively ensured. Only if the identity characteristic of each consultant is ensured to be completely different from that of other people, the identity characteristic and the consultant can be in one-to-one correspondence, so that a solid foundation is provided for the subsequent generation of the list serial number, the uniqueness and the accuracy of the list serial number are effectively ensured, and a reliable guarantee is provided for the automatic generation of the whole conference era and the smooth development of the encryption process.
Referring to fig. 3, the method further comprises the steps of:
When the identity characteristic is calculated, the identity characteristic=ω1×fundamental frequency information+ω2×resonance information of a weighted average formula is adopted to calculate according to a high-performance processor built in the intelligent conference terminal and algorithm support of a Guangming large model, wherein ω1 is a fundamental frequency information weight, ω2 is a resonance information weight, and ω1+ω2=1.
And the weight adjustment mechanism is deeply adapted to the actual application scene of the product, wherein when the number of the identity lists is changed, the system can be used for positively adjusting the weight of the fundamental frequency information based on the judgment of the terminal on the conference scale, such as whether the number of the identity lists is a small-scale department example meeting or a large-scale industry peak meeting, for example. For example, in a small enterprise internal conference of only 3 people, the weight ω1 of the fundamental frequency information may be set to 0.3, and the individuals are distinguished by more depending on resonance information, while in a large working conference of a government agency of 20 people, ω1 is raised to 0.7, by strengthening the remarkable individual voice characteristic of the fundamental frequency, the fundamental frequency difference between men and women is obvious, the voiceprint recognition accuracy in a multi-person scene is improved, and the functional characteristic of the product of distinguishing speakers is matched.
For the weight adjustment of the resonance information, the system can be dynamically adjusted according to the total data volume of the sound data, wherein the total data volume is calculated by the audio time length and the sampling rate acquired by the terminal microphone array. When meeting time is short and speaking sparsely, such as a rapid meeting of 10 minutes, the resonance information weight omega 2 is possibly as low as 0.2, so that interference of environmental noise on feature extraction is reduced, while in a financial institution investment decision meeting of 2 hours, along with the increase of the sound data, continuous voices with a large amount of discussion and debate are included, omega 2 is increased to 0.6, stability of resonance information reflecting stable features of a sound channel structure in long-term voices is utilized, anti-interference capability of sound-print features in a complex environment is enhanced, and the high-sensitivity microphone array of a product is matched, so that accuracy of identity recognition is further ensured.
The weight self-adaptive adjusting mechanism fully combines the hardware performance (such as the pickup capacity of a microphone array and the real-time computing capacity of a processor) and the algorithm advantages (intelligent judgment of a large model for meeting scenes), so that the calculation of the identity features can adapt to the requirements of meetings of different scales, can also cope with the voice environments under different data volumes, provides reliable basic feature data for the generation of sequence numbers of follow-up lists and the encryption of meeting contents, and effectively supports the core targets of the product in efficient, accurate and safe meeting experience.
The method further comprises the following steps:
After meeting record data are generated, the intelligent meeting terminal can count the proportion of the text content of each participant (corresponding to the identity list) in the total meeting record data through the built-in data analysis module, for example, the speaking content of the participant A accounts for 30% and the participant B accounts for 5%. Based on the duty ratio, the terminal calls a dynamic encryption strategy module of the Ming-Zhuang model, adjusts the shift offset value according to the anti-correlation rule, and forms an encryption formula of 'shift=list serial number+shift offset value'.
For example, in the investment decision-making of a financial institution, if the speech content of an analyst (serial number 5 in the identity list) is 40% (larger data amount), the shift offset value may be adjusted to 1 and finally 5+1=6, while the speech content of an auditor (serial number 8) is only 5% (smaller data amount), the offset value may be set to 5 and 8+5=13. The adjusting mechanism is linked with the professional word stock function of the product, and for industry term dense content with large duty ratio, such as futures contract and risk exposure, the code conversion error is reduced through a small offset value, the accuracy of decrypted professional words is ensured, and for scattered speaking with small duty ratio, the encryption strength is enhanced through a large offset value, so that the product advantage of high safety is met.
Meanwhile, a high-performance processor of the terminal can check whether the shifted encryption codes are in the effective range of a Chinese coding algorithm or not in real time, such as a coding section of GB2312, and an anti-correlation adjustment mechanism can reduce the coding overflow probability caused by large offset, namely core content with more speech has small offset and reduces the risk of over-range, and even if the content with less speech has large offset, the total amount of the whole over-range is controllable due to small data quantity. The design not only adapts to the performance requirement of products for supporting 20 meeting rooms to process concurrently, but also improves the overall safety and accuracy of meeting summary through dynamic differentiated encryption, and constructs a multi-level data protection system together with functions such as burn after hearing, exclusive security encryption and the like.
The method further comprises the following steps:
When temporary meeting staff appear in the meeting, such as department responsible persons added in the middle of the meeting in the enterprise and crews newly added in the meeting of government authorities, the intelligent meeting terminal acquires the newly added identity list and triggers a time-sharing sequencing mechanism through the real-time synchronization function of an external platform such as an OA system, a party building system and the like of the enterprise. The processing module of the terminal firstly locks conference data in a first period of time, calls a conference summary generation engine of a bright large model based on the current M identity lists, such as 10 people of an original participant, and generates first conference summary data in combination with a preset template, wherein the data comprises voiceprint information, encryption records and content corresponding relations corresponding to the M identities, and is synchronized to a cloud for storage through an APN encryption channel, and meanwhile, local caching supports temporary storage of a burn after hearing mode.
After entering the second period of time, the terminal starts an independent voiceprint recognition process for N newly added identity lists, such as 3 newly added people, and acquires sound data of the newly added people through the high-sensitivity microphone array, extracts fundamental frequency information and resonance information, and calculates identity characteristic values. At this time, the list number of the newly added identity is generated according to the rule of 'original maximum number+1', for example, the original number is 1-10, and the newly added number is 11-13, namely, the sorting position is automatically linked to 'original M number+N newly added number'. The terminal generates second conference summary data based on the speaking content of the added person, and the encryption process multiplexes the Chinese coding algorithm and the shift rule, for example, the shift value is adjusted according to the speaking data volume ratio of the added person, so as to ensure consistency with the encryption standard of the first conference summary.
The time-sharing sequencing mechanism is fully adaptive to the hardware performance and the software architecture of a product, on one hand, the generation of new and old parts of summary data can be processed simultaneously by depending on the parallel computing capability of a high-performance processor of a terminal, the real-time transfer of a conference is not affected, the concurrent processing performance of 20 conference rooms can be guaranteed, the multi-period data parallel processing of a single conference room can be guaranteed, on the other hand, the two parts of summary data can be stored in a terminal and an external system (such as an OA and a party building system) in an associated manner by a cross-terminal synchronous function of a cloud conference service platform, a user can combine the two parts of summary data in a one-key manner through a terminal interface, and the consistency of sequence number continuity (1-13 are not repeated) and encryption codes can be automatically checked in the combining process, so that the integrity of conference records is guaranteed.
The mechanism effectively aims at the scene of dynamic adjustment of personnel in the conference, avoids the interference of newly added personnel on the original encryption flow, ensures the accuracy of identity association through serial number connection if the encryption codes of the original M identities are not needed to be recalculated, and cooperates with the functions of 'cross-system conference summary automatic transmission', 'exclusive safe encryption', and the like of products, thereby improving the expansibility and flexibility of the system in the complex conference scene and meeting the requirements of different industries on dynamic participation management.
The method further comprises the following steps:
When a temporary newly-added meeting staff appears in the meeting, such as a windcontrol expert added in the middle of a financial institution meeting and a discipline head person temporarily arriving at the education institution seminar, the intelligent meeting terminal can automatically acquire N newly-added identity lists through the real-time data synchronization function of an external platform such as an enterprise OA system, a party building system and the like, and trigger a time-sharing sequencing mechanism.
In a first period, the terminal adjusts list serial numbers of M identity lists existing currently according to parallel computing capacity of a high-performance processor, if an 'n+1 times amplified' strategy is adopted, if M=5 participants with original serial numbers of 1-5 are newly added with N=4, the serial numbers are adjusted to be 5, 10, 15, 20 and 25, or a 'backward N-bit' strategy is adopted, the original serial numbers of 1-5 are adjusted to be 5-9, and an insertion space is provided for the newly added people by reserving serial number gaps. In the process, the terminal synchronously generates meeting record fragments corresponding to the current M identities, completes primary encryption by combining a professional word stock, adopts a Chinese coding algorithm and a shift encryption rule, and temporarily stores the meeting record fragments to the cloud through an APN encryption channel, so that the original encryption process is ensured not to be interfered by newly added personnel.
After entering the second period of time, the terminal starts the voiceprint acquisition and processing flow of the newly added N identities, namely capturing voice data of the newly added person through the omnidirectional pickup microphone array, extracting fundamental frequency information and resonance information, calculating identity characteristic values through a Guangming large-model voice small model, and generating a sequence number, such as 1-4 of a reserved vacancy. After the speaking content of the newly added person is transcribed into characters in real time, dynamic differential encryption is carried out according to the serial numbers of the characters, and the shift offset value is adjusted according to the data volume ratio, so that an independent encrypted record segment is formed.
And finally, the terminal reorders the serial numbers of the M original identities and the N newly added identities from small to large, if the serial numbers are combined into 1-9, a template engine in the conference service is called, and the encrypted record fragments of the two periods of time are integrated into complete conference summary data. In the integration process, the system verifies the data integrity through an SM3 hash algorithm, so that the new sequence number and the old sequence number are ensured to be continuous and have no repetition, for example, the new sequence number, the old sequence number and the new sequence number are 1-9 after being recombined with 1-4, and the shifting rule of the encryption codes is kept consistent. The generated complete summary can be exported into formats such as Word and PDF by one key, or transmitted to systems such as OA and party building by a 'business number fusion platform', and the temporary data is destroyed by supporting the function of enabling 'burn after hearing'.
The mechanism fully utilizes the cooperative storage advantages of the real-time processing capacity and the cloud end of the terminal, when meeting participants dynamically adjust, the stability of original encrypted data is guaranteed, historical codes are not required to be recalculated, seamless access of newly added persons is realized through sequence number reservation, and the system is deeply adapted to characteristics such as 'high concurrency processing', 'exclusive safe encryption' and the like, so that the expansibility and encryption flexibility of the system in a complex meeting scene are remarkably improved.
The method further comprises the following steps:
Before the intelligent conference terminal executes the code shift encryption operation, the total list number of the current participants, namely the sum of M+N, M is the initial participant, N is the newly added participant, and the odd-even attribute is judged through the logic operation of the terminal processing module. For example, in an enterprise department regular meeting, if 5 people are initially involved (list number 1-5), no new people are added halfway, the total amount is 5 (odd), a first direction offset (such as the increasing direction of the code value) is triggered, and if 2 people are initially involved in a government agency meeting, the total amount is 10 (even), a second direction offset (such as the decreasing direction of the code value) is started.
The mechanism is deeply integrated with an encryption system of a product, namely, a displacement rule of a first direction and a displacement rule of a second direction are both based on an effective interval design of a Chinese coding algorithm (such as GB 2312), and a SM2 and SM3 algorithm system in exclusive security encryption is matched to form a multi-layer encryption logic. For example, when the total amount is odd, the actual code of the word "meeting" is mapped into "the abstinence instrument" after the first direction deviation, and when the total amount is even, the actual code of the word "meeting" is mapped into "the redundancy instrument" after the second direction deviation, and the encryption results of the two directions are irregularly associated, so that the difficulty of unauthorized cracking is greatly improved.
Meanwhile, a high-performance processor of the terminal can check whether the shifted encryption codes exceed the coding range in real time, and the offset values of a plurality of speaking parties are small by combining a dynamic differential encryption strategy, so that the shifting in different directions is ensured to be executed in a safety interval. The dynamic direction adjustment based on the serial number total parity makes full use of the real-time computing capability of the Guangming big model, and is complementary with the function of burn-after-hearing, wherein the former enhances the anti-cracking performance of the encryption process, the latter avoids the leakage risk from the source of the data life cycle, a protection system meeting the requirements of high-security conference is constructed together, and the application of the product in security scenes such as finance, government and the like is effectively supported.
The embodiment of the application also discloses an automatic generation system of the conference summary driven by the intelligent conference terminal, which comprises a processor, wherein the processor executes the steps of the automatic generation method of the conference summary driven by the intelligent conference terminal.
While embodiments of the present application have been shown and described above, it will be understood that the above embodiments are illustrative and not to be construed as limiting the application, and that variations, modifications, alternatives and variations may be made to the above embodiments by one of ordinary skill in the art within the scope of the application.
Claims (6)
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202511093640.6A CN120597331B (en) | 2025-08-06 | 2025-08-06 | Intelligent conference terminal-driven conference summary automatic generation method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202511093640.6A CN120597331B (en) | 2025-08-06 | 2025-08-06 | Intelligent conference terminal-driven conference summary automatic generation method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN120597331A CN120597331A (en) | 2025-09-05 |
| CN120597331B true CN120597331B (en) | 2026-01-02 |
Family
ID=96893664
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202511093640.6A Active CN120597331B (en) | 2025-08-06 | 2025-08-06 | Intelligent conference terminal-driven conference summary automatic generation method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN120597331B (en) |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112037791A (en) * | 2020-08-12 | 2020-12-04 | 广东电力信息科技有限公司 | Conference summary transcription method, apparatus and storage medium |
| CN115277192A (en) * | 2022-07-27 | 2022-11-01 | 维沃移动通信有限公司 | Information encryption method and device and electronic equipment |
Family Cites Families (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104427292A (en) * | 2013-08-22 | 2015-03-18 | 中兴通讯股份有限公司 | Method and device for extracting a conference summary |
| US9990504B1 (en) * | 2017-12-18 | 2018-06-05 | Northern Trust Corporation | Systems and methods for generating and maintaining immutable digital meeting records within distributed network nodes |
| CN108986826A (en) * | 2018-08-14 | 2018-12-11 | 中国平安人寿保险股份有限公司 | Automatically generate method, electronic device and the readable storage medium storing program for executing of minutes |
| CN109523988B (en) * | 2018-11-26 | 2021-11-05 | 安徽淘云科技股份有限公司 | Text deduction method and device |
| CN115242568A (en) * | 2022-07-22 | 2022-10-25 | 思必驰科技股份有限公司 | Conference transcription method, system and electronic device for dynamically updating voiceprint database |
| CN118098204A (en) * | 2024-02-22 | 2024-05-28 | 山东浪潮成方数字服务有限公司 | Intelligent conference minutes method, device and medium based on dynamic voiceprint |
| CN119696866B (en) * | 2024-12-10 | 2025-06-27 | 数字大理建设运营有限公司 | USB flash disk encryption system for data element circulation |
-
2025
- 2025-08-06 CN CN202511093640.6A patent/CN120597331B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112037791A (en) * | 2020-08-12 | 2020-12-04 | 广东电力信息科技有限公司 | Conference summary transcription method, apparatus and storage medium |
| CN115277192A (en) * | 2022-07-27 | 2022-11-01 | 维沃移动通信有限公司 | Information encryption method and device and electronic equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN120597331A (en) | 2025-09-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US6687671B2 (en) | Method and apparatus for automatic collection and summarization of meeting information | |
| CN107221331A (en) | A kind of personal identification method and equipment based on vocal print | |
| CN106790054A (en) | Interactive authentication system and method based on recognition of face and Application on Voiceprint Recognition | |
| CN109873907A (en) | Call processing method, device, computer equipment and storage medium | |
| CN109192216A (en) | A kind of Application on Voiceprint Recognition training dataset emulation acquisition methods and its acquisition device | |
| CN109584887A (en) | A kind of method and apparatus that voiceprint extracts model generation, voiceprint extraction | |
| CN112131365A (en) | Data processing method, device, equipment and medium | |
| CN114764690A (en) | Method, device and system for intelligently conducting conference summary | |
| CN113971956A (en) | Information processing method and device, electronic equipment and readable storage medium | |
| CN120597331B (en) | Intelligent conference terminal-driven conference summary automatic generation method and system | |
| Wang et al. | Interference quality assessment of speech communication based on deep learning | |
| KR20220166465A (en) | Meeting minutes creating system and method using multi-channel receiver | |
| CN110996036B (en) | Remote online conference management system based on AI intelligent technology | |
| CN121188817A (en) | A method and system for monitoring and inputting clinical nursing patient information | |
| CN115294987A (en) | Conference record generation method, device, equipment and storage medium | |
| CN116631109A (en) | Voiceprint-based identification method and device, electronic equipment, storage medium | |
| CN113611328B (en) | A method and device for voiceprint recognition speech evaluation | |
| CN121585789A (en) | A method, device, and medium for offline local area network-based AI digital human conference proxy | |
| US12189740B2 (en) | Multi-biometric authentication system | |
| Hunter et al. | MEETING: A corpus of French meeting-style conversations | |
| CN115296820A (en) | A voice-aware hash authentication method based on smart contracts | |
| CN114724072A (en) | Method, device, device and storage medium for intelligent question answering | |
| CN112784240A (en) | Unified identity authentication platform and face identity recognition method thereof | |
| CN111613228A (en) | Identity and content recognition system based on voiceprint codes | |
| CN115147205B (en) | A digital currency ledger synchronization system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |