Remote conference encryption method
By preprocessing and feature extraction of audio authentication data in remote meetings, and combining audio authentication models for authentication, the problem of poor recognition effect of low-quality sound data in the existing technology is solved, and a more accurate identity is achieved Verification and security management ensure the confidentiality and security of the meeting content.
Patent Information
- Application Number
- CN202411440265.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-15
- Publication Date
- 2025-01-10
AI Technical Summary
When identifying the identity of participants, the commonly used audio identification model has poorer recognition of low-quality sound data, which can easily lead to the risk of identity misjudgment and information leakage.
By denoising, framing, pre-emphasis and feature extraction of the initial audio authentication data, combining the Mel frequency cepspectral coefficient, fundamental frequency characteristics and linear prediction coefficients, the audio authentication model is used for authentication, and based on the identity information Assign open permissions to the meeting content to ensure that only verified participants can participate in the meeting.
It improves the accuracy and security of voice recognition, ensures that the meeting content is only accessible within the scope of authorization, reduces the risk of identity impersonation and information leakage, and enhances the security and management efficiency of remote meetings.
Smart Images
Figure CN119299189A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of network security, and in particular to a remote conference encryption method. Background Art
[0002] Remote conferencing refers to a form of meeting in which participants are not in the same location and communicate and collaborate through technical means such as the Internet or telephone. This type of meeting has become increasingly common in the modern work environment, especially driven by globalization and digitalization. Remote conferencing simulates the experience of traditional face-to-face meetings and provides a flexible and efficient way of communication. However, the security of remote conferencing, especially in sensitive legal matters, faces significant challenges. Existing technologies have failed to fully address issues such as the security of data transmission, participant identity authentication, and cross-platform compatibility of systems.
[0003] When identifying the identities of participants, current remote conference encryption methods generally obtain the audio identification data of participants through cameras and microphones, compare the collected data with the preset stored audio identification data one by one through a specific algorithm, and confirm the identity information of the participants based on the comparison and matching results. However, commonly used audio identification models often require high-quality sound data samples. Affected by real-world environmental elements, high-quality sound data samples are difficult to obtain. The collected sound data samples are often low-quality and contain a lot of noise. Common audio identification models have poor recognition effects on them, which can easily lead to misjudgment of the identities of participants and pose a risk of information leakage. Summary of the invention
[0004] In view of this, the present invention proposes a remote conference encryption method, which improves the accuracy of sound recognition, enhances security and implements real-time monitoring and management through audio identification data collection and preprocessing, audio identification data recognition, identity authentication and permission allocation, and conference content open permission allocation, and solves the problems of data transmission security and participant identity authentication risks.
[0005] The technical solution of the present invention is implemented as follows: The present invention provides a remote conference encryption method, comprising the following steps:
[0006] S1, collecting initial audio identification data of the participant to be verified, and preprocessing the initial audio identification data to obtain first audio identification data;
[0007] S2, identifying the first audio identification data according to the audio identification model to obtain the identity information of the participant to be verified;
[0008] S3, based on the identity information of the participant to be verified, confirm the conference content opening permission level of the participant to be verified, complete the identity verification of the participant to be verified, and mark the corresponding participant to be verified as a formal participant;
[0009] S4, when the identity authentication of all participants to be verified is completed, different conference content opening permissions are allocated based on the conference content opening permission levels of all formal participants, and the remote conference is started according to the different conference content opening permissions.
[0010] On the basis of the above technical solution, preferably, step S1 includes:
[0011] S11, collecting initial audio identification data of the participant to be verified, performing noise reduction and interference removal processing on the collected initial audio identification data, and obtaining initial noise-removed audio identification data;
[0012] S12, performing frame processing on the initial noise-removed audio identification data to obtain each frame signal of the initial noise-removed audio identification data, and performing pre-emphasis processing on each frame signal of the initial noise-removed audio identification data to obtain the original audio identification data;
[0013] S13, extracting sound features of the first audio frequency identification data, normalizing the extracted sound features, and combining the processed sound features to form first audio frequency identification data;
[0014] S14, performing quality assessment on the first audio identification data, screening out unqualified first audio identification data, and obtaining first audio identification data.
[0015] Based on the above technical solution, preferably, step S11 includes:
[0016] The initial audio identification data is subjected to noise reduction processing using a spectrum subtraction method or an adaptive filtering algorithm, and the initial audio identification data is subjected to interference removal processing using a voice activity detection algorithm, so as to remove background noise and interference of the initial audio identification data.
[0017] On the basis of the above technical solution, preferably, step S12 includes:
[0018] The initial de-noised frequency identification data is divided into frames of fixed length, and adjacent frames are overlapped to ensure the continuity of information. Each frame signal of the initial de-noised frequency identification data is pre-emphasized to increase the energy of the high-frequency part of the initial de-noised frequency identification data. A Hamming window function is applied to the pre-emphasized signal to reduce spectrum leakage.
[0019] On the basis of the above technical solution, preferably, step S13 includes:
[0020] The Mel-frequency cepstral coefficients, fundamental frequency features and linear prediction coefficients of each frame of the meta-audio identification data are calculated, the Mel-frequency cepstral coefficients, fundamental frequency features and linear prediction coefficients are normalized, and the processed feature vectors are combined to form the first meta-audio identification data.
[0021] Based on the above technical solution, preferably, step S14 includes:
[0022] Calculating a signal-to-noise ratio of the first audio identification data, and comparing the signal-to-noise ratio of the first audio identification data with a preset signal-to-noise ratio threshold;
[0023] When the signal-to-noise ratio of the first audio identification data is greater than or equal to a preset signal-to-noise ratio threshold, retaining the first audio identification data;
[0024] When the signal-to-noise ratio of the first audio identification data is less than a preset signal-to-noise ratio threshold, removing the first audio identification data;
[0025] All first-element audio identification data that pass the quality assessment are combined to form final first audio identification data.
[0026] Based on the above technical solution, preferably, step S2 includes:
[0027] The first audio identification data includes Mel frequency cepstrum coefficients, fundamental frequency features and feature vectors corresponding to linear prediction coefficients. The mathematical expression of the first audio identification data is:
[0028] S={M T ,L T ,J T};
[0029] Among them, S is the first audio identification data, M is the feature vector corresponding to the Mel-frequency cepstrum coefficient, L is the feature vector corresponding to the linear prediction coefficient, and J is the feature vector corresponding to the fundamental frequency feature.
[0030] On the basis of the above technical solution, preferably, the calculation formula of the audio identification model is:
[0031]
[0032] Among them, I i is the identity information of the i-th participant to be verified output by the audio identification model, X j is the jth convolution kernel of the audio identification model, M, L, and J are the feature vectors corresponding to the Mel frequency cepstrum coefficient, linear prediction coefficient, and fundamental frequency feature, respectively, n is the total number of convolution kernels, and α M , α L , α Jare the weight coefficients of the eigenvectors corresponding to the Mel frequency cepstrum coefficients, linear prediction coefficients and fundamental frequency features, β j is the bias of the jth convolution kernel.
[0033] On the basis of the above technical solution, preferably, step S3 includes:
[0034] According to the identity information of the participant to be verified, the preset permission database is queried, the conference content opening permission level corresponding to the identity information of the participant to be verified is extracted from the permission database, and the extracted conference content opening permission level is compared with the preset permission level threshold.
[0035] When the extracted conference content opening permission level is higher than or equal to the preset threshold, the participant to be verified is marked as a formal participant; when the extracted conference content opening permission level is lower than the preset threshold, the current participant to be verified is denied access to the meeting;
[0036] The verification results are recorded, including a list of formal conference participants who have passed the verification and the corresponding conference content opening permission levels.
[0037] Based on the above technical solution, preferably, step S4 includes:
[0038] When the identity verification of all participants to be verified is completed, the meeting content access permission levels of all formal participants are counted, and the minimum permission requirements for the meeting are set according to the confidentiality level of the meeting content. The meeting content access permission level of each formal participant is compared with the minimum permission requirements for the meeting;
[0039] Full meeting content access permissions are granted to formal participants whose meeting content access permissions are higher than or equal to the minimum permission requirements of the meeting.
[0040] For formal participants whose meeting content access permission level is lower than the minimum permission requirement of the meeting, partial meeting content access permission will be granted according to the meeting content access permission level of the formal participant;
[0041] Generate a conference access control list, which includes the identity information of each official participant and the corresponding conference content disclosure authority;
[0042] Start the remote conference system and control the access rights of each official participant to the conference content according to the conference access control list;
[0043] Start a remote meeting and monitor and manage the information access behavior of formal participants in real time during the meeting.
[0044] The remote conference encryption method of the present invention has the following beneficial effects compared with the prior art:
[0045] (1) By pre-processing and identifying the initial audio identification data of the participants, their identity information is accurately obtained, and the level of conference content access permission is confirmed based on the identity information, ensuring that only verified formal participants can participate in the meeting. Different conference content access permissions are allocated according to the conference content access permission level, thereby enhancing the security and confidentiality of the meeting and ensuring that the conference content is only accessible within the authorized scope;
[0046] (2) The audio identification model is used to achieve efficient and accurate recognition of sound signals through multi-dimensional feature fusion, convolution kernel processing, weight coefficient optimization and bias adjustment, providing strong support for identity authentication in remote conferencing systems;
[0047] (3) Spectral subtraction or adaptive filtering algorithm is used for noise reduction, and voice activity detection algorithm is used for interference removal to ensure high-quality input of initial audio identification data and improve recognition accuracy. (4) Multidimensional feature vectors are constructed by Mel-frequency cepstral coefficients, fundamental frequency features and linear prediction coefficients to enhance the expression ability and recognition accuracy of audio identification data and ensure the high efficiency of sound recognition. BRIEF DESCRIPTION OF THE DRAWINGS
[0048] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0049] Figure 1 The present invention is a flowchart of a remote conference encryption method. DETAILED DESCRIPTION
[0050] The following will be combined with the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0051] See also Figure 1 This embodiment provides a remote conference encryption method, comprising the following steps:
[0052] S1, collecting initial audio identification data of the participant to be verified, and preprocessing the initial audio identification data to obtain first audio identification data;
[0053] S2, identifying the first audio identification data according to the audio identification model to obtain the identity information of the participant to be verified;
[0054] S3, based on the identity information of the participant to be verified, confirm the conference content opening permission level of the participant to be verified, complete the identity verification of the participant to be verified, and mark the corresponding participant to be verified as a formal participant;
[0055] S4, when the identity authentication of all participants to be verified is completed, different conference content opening permissions are allocated based on the conference content opening permission levels of all formal participants, and the remote conference is started according to the different conference content opening permissions.
[0056] Specifically, this embodiment can more accurately identify the identities of conference participants and reduce the risk of identity fraud by collecting and preprocessing audio identification data and combining it with an audio identification model.
[0057] The meeting content access permission level of the participants is determined based on the recognition results, realizing differentiated permission management for different participants.
[0058] Through strict identity authentication and permission allocation mechanisms, it is ensured that only verified official participants can participate in the meeting, which improves the security of remote meetings.
[0059] Different meeting content access permissions are assigned according to the permission levels of participants to effectively prevent unauthorized information access and protect sensitive information.
[0060] Automated identity authentication and permission allocation processes improve the efficiency of organizing and managing remote meetings.
[0061] It can flexibly control the scope of access to conference content according to the permission levels of different participants, achieving more refined information management.
[0062] Step S1 includes:
[0063] S11, collecting initial audio identification data of the participant to be verified, performing noise reduction and interference removal processing on the collected initial audio identification data, and obtaining initial noise-removed audio identification data;
[0064] S12, performing frame processing on the initial noise-removed audio identification data to obtain each frame signal of the initial noise-removed audio identification data, and performing pre-emphasis processing on each frame signal of the initial noise-removed audio identification data to obtain the original audio identification data;
[0065] S13, extracting sound features of the first audio frequency identification data, normalizing the extracted sound features, and combining the processed sound features to form first audio frequency identification data;
[0066] S14, performing quality assessment on the first audio identification data, screening out unqualified first audio identification data, and obtaining first audio identification data.
[0067] Specifically, step S1 of this embodiment eliminates background noise and interference signals in the initial audio identification data through noise reduction and interference removal processing, thereby improving the purity and reliability of the audio identification data.
[0068] The use of frame processing and pre-emphasis processing helps to better capture the time-frequency characteristics of sound and improves the accuracy and effectiveness of subsequent sound feature extraction.
[0069] The extracted sound features are normalized to make the features of different samples comparable, which is beneficial to improving the stability and accuracy of sound recognition.
[0070] Through quality assessment and screening, unqualified audio identification data are eliminated, ensuring that the audio identification data used for subsequent identification is of high quality and representativeness.
[0071] The high-quality first audio identification data obtained after processing provides a more reliable input for the subsequent sound recognition process, which helps to improve the overall recognition accuracy.
[0072] Step S11 includes:
[0073] The initial audio identification data is subjected to noise reduction processing using a spectrum subtraction method or an adaptive filtering algorithm, and the initial audio identification data is subjected to interference removal processing using a voice activity detection algorithm, so as to remove background noise and interference of the initial audio identification data.
[0074] Specifically, step S11 of this embodiment performs noise reduction processing by spectrum subtraction or adaptive filtering algorithm, thereby effectively removing background noise in the initial audio identification data and improving the signal-to-noise ratio and clarity of the audio identification data.
[0075] Using voice activity detection algorithm for interference removal processing can effectively identify and remove interference signals in non-speech parts, retain the real voice information, and improve the purity of audio identification data.
[0076] By removing background noise and interference, the audio identification data is made closer to the ideal speech signal, providing a more reliable input for the sound recognition process.
[0077] The initial audio identification data is processed in a combined manner of multiple algorithms, thereby improving the system's adaptability to audio identification data collected in different environments.
[0078] By effectively removing background noise and interference, the risk of misidentification due to environmental factors is reduced, and the security of the entire remote conference encryption system is improved.
[0079] This processing method can effectively cope with various complex sound collection environments, allowing the system to obtain better audio identification data under different noise backgrounds, thereby improving the practicality and applicability of the system.
[0080] Step S12 includes:
[0081] The initial de-noised frequency identification data is divided into frames of fixed length, and adjacent frames are overlapped to ensure the continuity of information. Each frame signal of the initial de-noised frequency identification data is pre-emphasized to increase the energy of the high-frequency part of the initial de-noised frequency identification data. A Hamming window function is applied to the pre-emphasized signal to reduce spectrum leakage.
[0082] Specifically, step S12 of this embodiment ensures the continuity and integrity of information by dividing the initial noise-removed audio identification data into frames of fixed length and performing overlapping processing. Pre-emphasis processing increases the energy of the high-frequency part and enhances the characteristic performance of the sound. The application of the Hamming window function effectively reduces spectrum leakage and improves the accuracy of spectrum analysis. Step S12 improves the quality and analyzability of the audio identification data, providing a more reliable and accurate data basis for the subsequent feature extraction and recognition process.
[0083] Step S13 includes:
[0084] The Mel-frequency cepstral coefficients, fundamental frequency features and linear prediction coefficients of each frame of the meta-audio identification data are calculated, the Mel-frequency cepstral coefficients, fundamental frequency features and linear prediction coefficients are normalized, and the processed feature vectors are combined to form the first meta-audio identification data.
[0085] Specifically, step S13 of this embodiment comprehensively captures the multi-dimensional characteristics of the sound by calculating the Mel-frequency cepstral coefficients, fundamental frequency features and linear prediction coefficients of each frame of the meta-audio identification data. These features are normalized to make different features comparable and eliminate the influence of dimensional differences. The processed feature vectors are combined to form the first meta-audio identification data, thereby realizing the comprehensive expression of the sound features. Step S13 not only improves the expressiveness and discrimination of the sound features, but also enhances the stability and reliability of the features, and provides high-quality, standardized feature input.
[0086] Step S14 includes:
[0087] Calculating a signal-to-noise ratio of the first audio identification data, and comparing the signal-to-noise ratio of the first audio identification data with a preset signal-to-noise ratio threshold;
[0088] When the signal-to-noise ratio of the first audio identification data is greater than or equal to a preset signal-to-noise ratio threshold, retaining the first audio identification data;
[0089] When the signal-to-noise ratio of the first audio identification data is less than a preset signal-to-noise ratio threshold, removing the first audio identification data;
[0090] All first-element audio identification data that pass the quality assessment are combined to form final first audio identification data.
[0091] Specifically, step S14 of this embodiment achieves strict control of the quality of audio identification data by calculating the signal-to-noise ratio of the first audio identification data and comparing it with a preset threshold. The practice of retaining high signal-to-noise ratio data and eliminating low signal-to-noise ratio data effectively improves the overall quality of the final audio identification data. Step S14 not only ensures that the audio identification data used for subsequent identification has a high signal quality, but also reduces the risk of identification errors introduced by low-quality data. By merging all data that pass the quality assessment, the first audio identification data formed has higher representativeness and credibility, providing reliable and accurate input.
[0092] Step S2 includes:
[0093] The first audio identification data includes Mel frequency cepstrum coefficients, fundamental frequency features and feature vectors corresponding to linear prediction coefficients. The mathematical expression of the first audio identification data is:
[0094] S={M T ,L T ,J T};
[0095] Among them, S is the first audio identification data, M is the feature vector corresponding to the Mel-frequency cepstrum coefficient, L is the feature vector corresponding to the linear prediction coefficient, and J is the feature vector corresponding to the fundamental frequency feature.
[0096] Specifically, the first audio identification data S of this embodiment combines three different sound features (Mel-frequency cepstral coefficients M, linear prediction coefficients L, and fundamental frequency features J) to form a more comprehensive sound representation. This multi-dimensional feature fusion can capture multiple aspects of the sound signal and improve the accuracy and accuracy of sound recognition. Different speakers may perform differently on these three features, and the combined use of these features helps capture the unique sound characteristics of individuals. A single feature may perform poorly in some cases, while multi-feature fusion can compensate for each other and improve the stability of the overall performance. At the same time, this structured representation method makes it simple to add new sound features by simply adding new feature components to the vector.
[0097] By expressing the first audio identification data S as a combination of three feature vectors, namely, Mel-frequency cepstral coefficient M, linear prediction coefficient L and fundamental frequency feature J, a comprehensive and multi-dimensional representation of the sound signal is achieved, capturing different aspects of the sound signal.
[0098] Mel-frequency cepstral coefficients reflect the spectral envelope characteristics of the sound, the linear prediction coefficients reflect the resonance characteristics of the vocal tract, and the fundamental frequency characteristics represent the basic periodic information of the sound. This combination of multi-dimensional features not only improves the accuracy and completeness of the sound representation, but also enhances the ability to distinguish between sound recognition. In this way, the unique characteristics of individual voices can be captured and analyzed more accurately, thereby improving the accuracy and reliability of sound recognition and providing a more comprehensive and reliable data foundation for the identity verification process.
[0099] The calculation formula of the audio identification model is:
[0100]
[0101] Among them, I i is the identity information of the i-th participant to be verified output by the audio identification model, X j is the jth convolution kernel of the audio identification model, M, L, and J are the feature vectors corresponding to the Mel frequency cepstrum coefficient, linear prediction coefficient, and fundamental frequency feature, respectively, n is the total number of convolution kernels, and α M , α L , α J are the weight coefficients of the eigenvectors corresponding to the Mel frequency cepstrum coefficients, linear prediction coefficients and fundamental frequency features, β j is the bias of the jth convolution kernel.
[0102] Specifically, the audio identification model of this embodiment uses the Mel frequency cepstral coefficient (M), linear prediction coefficient (L) and fundamental frequency feature (J) as input feature vectors, and the model can capture the multi-dimensional features of the sound signal. These features respectively reflect the spectrum envelope, vocal tract resonance characteristics and basic periodic information of the sound, providing a comprehensive representation of the sound signal.
[0103] Using multiple convolution kernels to process input features can extract higher-level features. The convolution operation helps capture the local correlation between features and enhance the model's ability to recognize sound features. j By summing the results of (j from 1 to n), the model can extract sound features from different angles and scales, enhancing the comprehensiveness and robustness of feature extraction
[0104] By introducing the weight coefficient α M , α L , α JThe model can weight the importance of Mel-frequency cepstral coefficients (M), linear prediction coefficients (L) and fundamental frequency features (J), so that the model can dynamically adjust the contribution of each feature according to actual conditions and improve the accuracy of recognition.
[0105] The introduction of nonlinearity through the ReLU activation function enables the model to learn and represent more complex feature relationships, improving the model's expressiveness and generalization capabilities.
[0106] Bias β j It plays an adjustment role in the convolution operation, helping the model to better fit the data. The introduction of bias can make the model more adaptable when processing different inputs.
[0107] Finally, the model outputs the identity information I of the i-th participant to be verified i , and accurately identify the identities of the participants. By comprehensively considering multi-dimensional features and convolution operations, the model can effectively distinguish the voice characteristics of different individuals, thereby improving the accuracy of identity authentication.
[0108] This audio identification model achieves efficient and accurate recognition of sound signals through multi-dimensional feature fusion, convolution kernel processing, weight coefficient optimization and bias adjustment, providing strong support for identity authentication in remote conferencing systems.
[0109] Step S3 includes:
[0110] According to the identity information of the participant to be verified, the preset permission database is queried, the conference content opening permission level corresponding to the identity information of the participant to be verified is extracted from the permission database, and the extracted conference content opening permission level is compared with the preset permission level threshold.
[0111] When the extracted conference content opening permission level is higher than or equal to the preset threshold, the participant to be verified is marked as a formal participant; when the extracted conference content opening permission level is lower than the preset threshold, the current participant to be verified is denied access to the meeting;
[0112] The verification results are recorded, including a list of formal conference participants who have passed the verification and the corresponding conference content opening permission levels.
[0113] Specifically, step S3 of this embodiment realizes accurate verification and control of the identity and authority of the participants by querying the preset authority database and comparing the open authority level of the conference content. It not only ensures that only those with sufficient authority can participate in the meeting, but also classifies and manages the participants according to different authority levels. By recording the verification results, including the list of formal participants who have passed the verification and the corresponding authority levels, a reliable basis is provided for conference management and information access control. Step S3 improves the security and confidentiality of remote conferences, effectively prevents unauthorized persons from participating in meetings or accessing sensitive information, and also lays the foundation for differentiated information access management in meetings.
[0114] Step S4 includes:
[0115] When the identity verification of all participants to be verified is completed, the meeting content access permission levels of all formal participants are counted, and the minimum permission requirements for the meeting are set according to the confidentiality level of the meeting content. The meeting content access permission level of each formal participant is compared with the minimum permission requirements for the meeting;
[0116] Full meeting content access permissions are granted to formal participants whose meeting content access permissions are higher than or equal to the minimum permission requirements of the meeting.
[0117] For formal participants whose meeting content access permission level is lower than the minimum permission requirement of the meeting, partial meeting content access permission will be granted according to the meeting content access permission level of the formal participant;
[0118] Generate a conference access control list, which includes the identity information of each official participant and the corresponding conference content disclosure rights;
[0119] Start the remote conference system and control the access rights of each official participant to the conference content according to the conference access control list;
[0120] Start a remote meeting and monitor and manage the information access behavior of formal participants in real time during the meeting.
[0121] Specifically, step S4 of this embodiment achieves refined information access control by counting the authority levels of formal participants and comparing them with the minimum authority requirements of the meeting. According to the authority level of the participants, full or partial meeting content opening authority is flexibly granted, ensuring information security while maximizing meeting participation. A conference access control list is generated and access to conference content is controlled accordingly, providing each participant with a tailored information access range. Real-time monitoring and management of information access behavior during the meeting further enhances the security of the meeting. Step S4 not only effectively protects sensitive information, but also realizes differentiated display of conference content, greatly improving the security, flexibility and efficiency of remote meetings, but also creates a suitable meeting environment for participants with different authority levels, ensuring the smooth progress of the meeting and the effective transmission of information.
[0122] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.
Claims
1. A remote conference encryption method, characterized in that: The following steps are involved: S1, collecting initial audio identification data of the participant to be verified, and preprocessing the initial audio identification data to obtain first audio identification data; S2, identifying the first audio identification data according to the audio identification model to obtain the identity information of the participant to be verified; S3, based on the identity information of the participant to be verified, confirm the conference content opening permission level of the participant to be verified, complete the identity verification of the participant to be verified, and mark the corresponding participant to be verified as a formal participant; S4, when the identity authentication of all participants to be verified is completed, different conference content opening permissions are allocated based on the conference content opening permission levels of all formal participants, and the remote conference is started according to the different conference content opening permissions.
2. A remote conference encryption method as claimed in claim 1, characterized in that: Step S1 includes: S11, collecting initial audio identification data of the participant to be verified, performing noise reduction and interference removal processing on the collected initial audio identification data, and obtaining initial noise-removed audio identification data; S12, performing frame processing on the initial noise-removed audio identification data to obtain each frame signal of the initial noise-removed audio identification data, and performing pre-emphasis processing on each frame signal of the initial noise-removed audio identification data to obtain the original audio identification data; S13, extracting sound features of the first audio frequency identification data, normalizing the extracted sound features, and combining the processed sound features to form first audio frequency identification data; S14, performing quality assessment on the first audio identification data, screening out unqualified first audio identification data, and obtaining first audio identification data.
3. A remote conference encryption method as claimed in claim 2, characterized in that: Step S11 includes: The initial audio identification data is subjected to noise reduction processing using a spectrum subtraction method or an adaptive filtering algorithm, and the initial audio identification data is subjected to interference removal processing using a voice activity detection algorithm, so as to remove background noise and interference of the initial audio identification data.
4. A remote conference encryption method as claimed in claim 3, characterized in that: Step S12 includes: The initial de-noised frequency identification data is divided into frames of fixed length, and adjacent frames are overlapped to ensure the continuity of information. Each frame signal of the initial de-noised frequency identification data is pre-emphasized to increase the energy of the high-frequency part of the initial de-noised frequency identification data. A Hamming window function is applied to the pre-emphasized signal to reduce spectrum leakage.
5. A remote conference encryption method as claimed in claim 4, characterized in that: Step S13 includes: The Mel-frequency cepstral coefficients, fundamental frequency features and linear prediction coefficients of each frame of the meta-audio identification data are calculated, the Mel-frequency cepstral coefficients, fundamental frequency features and linear prediction coefficients are normalized, and the processed feature vectors are combined to form the first meta-audio identification data.
6. A remote conference encryption method as claimed in claim 5, characterized in that: Step S14 includes: Calculating a signal-to-noise ratio of the first audio identification data, and comparing the signal-to-noise ratio of the first audio identification data with a preset signal-to-noise ratio threshold; When the signal-to-noise ratio of the first audio identification data is greater than or equal to a preset signal-to-noise ratio threshold, retaining the first audio identification data; When the signal-to-noise ratio of the first audio identification data is less than a preset signal-to-noise ratio threshold, removing the first audio identification data; All first-element audio identification data that pass the quality assessment are combined to form final first audio identification data.
7. A remote conference encryption method as claimed in claim 1, characterized in that: Step S2 includes: The first audio identification data includes Mel frequency cepstrum coefficients, fundamental frequency features and feature vectors corresponding to linear prediction coefficients. The mathematical expression of the first audio identification data is: S={M T ,L T ,J T }; Among them, S is the first audio identification data, M is the feature vector corresponding to the Mel-frequency cepstrum coefficient, L is the feature vector corresponding to the linear prediction coefficient, and J is the feature vector corresponding to the fundamental frequency feature.
8. A remote conference encryption method as claimed in claim 7, characterized in that: The calculation formula of the audio identification model is: Among them, I i is the identity information of the i-th participant to be verified output by the audio identification model, X j is the jth convolution kernel of the audio identification model, M, L, and J are the feature vectors corresponding to the Mel frequency cepstrum coefficient, linear prediction coefficient, and fundamental frequency feature, respectively, n is the total number of convolution kernels, and α M , α L , α J are the weight coefficients of the eigenvectors corresponding to the Mel frequency cepstrum coefficients, linear prediction coefficients and fundamental frequency features, β j is the bias of the jth convolution kernel.
9. A remote conference encryption method as claimed in claim 1, characterized in that: Step S3 includes: According to the identity information of the participant to be verified, the preset permission database is queried, the conference content opening permission level corresponding to the identity information of the participant to be verified is extracted from the permission database, and the extracted conference content opening permission level is compared with the preset permission level threshold. When the extracted conference content opening permission level is higher than or equal to the preset threshold, the participant to be verified is marked as a formal participant; when the extracted conference content opening permission level is lower than the preset threshold, the current participant to be verified is denied access to the meeting; The verification results are recorded, including a list of formal conference participants who have passed the verification and the corresponding conference content opening permission levels.
10. A remote conference encryption method as claimed in claim 1, characterized in that: Step S4 includes: When the identity verification of all participants to be verified is completed, the meeting content access permission levels of all formal participants are counted, and the minimum permission requirements for the meeting are set according to the confidentiality level of the meeting content. The meeting content access permission level of each formal participant is compared with the minimum permission requirements for the meeting; Full meeting content access permissions are granted to formal participants whose meeting content access permissions are higher than or equal to the minimum permission requirements of the meeting. For formal participants whose meeting content access permission level is lower than the minimum permission requirement of the meeting, partial meeting content access permission will be granted according to the meeting content access permission level of the formal participant; Generate a conference access control list, which includes the identity information of each official participant and the corresponding conference content disclosure rights; Start the remote conference system and control the access rights of each official participant to the conference content according to the conference access control list; Start a remote meeting and monitor and manage the information access behavior of formal participants in real time during the meeting.