Talkback authority management method and system based on voiceprint recognition

Abstract

The invention relates to the technical field of talkback communication, in particular to a voiceprint recognition-based talkback authority management method, which comprises the following steps of: establishing a complete mapping chain; setting a time threshold, and establishing a dual verification mechanism to verify the user identity: verifying the user identity including serial number verification and voiceprint feature verification, and if verification fails, outputting an early warning signal; after the serial number and the voiceprint feature are successfully verified, marking the user job number corresponding to the complete mapping chain to be in an on-duty state; receiving request voice after successful verification of the request user and request voiceprint features corresponding to the request voice; the recognition request voiceprint feature comprises an access request of a response role and a workshop; establishing a one-way communication link sequence from the request user to the multiple response users; verifying the response user state; and calling out the working time length of the same response user in the workshop, and then adjusting the time threshold value of the workshop by adopting the working time length.

Description

Technical Field

[0001] This invention relates to the field of intercom communication technology, and more specifically, to an intercom access control method and system based on voiceprint recognition. Background Technology

[0002] In modern intelligent manufacturing and large-scale industrial production scenarios, intercom systems serve as the core communication carrier for transmitting production scheduling instructions, responding to equipment failures, and facilitating cross-positional collaborative communication. Their operational security, scheduling accuracy, and the effectiveness of access control directly impact factory production efficiency, operational safety, and information confidentiality. However, traditional intercom access control methods are no longer adequate for the complex needs of industrial scenarios, revealing numerous technical shortcomings and application limitations.

[0003] Firstly, the authentication mechanism is inefficient: Before use, existing intercom systems typically establish a binding relationship between the intercom terminal and the corresponding user's employee ID to prevent unauthorized personnel from impersonating others and exceeding their communication privileges. When a user outputs voice, the voiceprint features in the output voice are extracted and then compared one by one with multiple standard voiceprints in the voiceprint standard library. However, in the process of comparing real-time voiceprint features with standard voiceprint features, the voiceprint standard library in factory scenarios needs to cover all positions (such as operators, maintenance technicians, workshop supervisors, and outsourced personnel). The large number of personnel results in a large number of standard voiceprints, and the traditional comparison method requires calculating the similarity between the real-time voiceprint and each standard voiceprint in sequence (such as through algorithms such as Euclidean distance and cosine similarity). Therefore, the authentication process will take too long, especially during peak periods when personnel use the intercom system (such as early shift scheduling, multiple device failure calls), when a large number of users are queuing for authentication, which not only seriously affects the communication response efficiency.

[0004] Secondly, the communication dispatch lacks intelligent adaptation capabilities. When a user initiates an intercom request, it cannot accurately identify the actual status of the responding user (e.g., a maintenance technician is focused on disassembling equipment, or an operator is adjusting production line parameters). It can only blindly initiate a communication link establishment request. If the responding user is in a high-intensity work state, it will not only fail to respond in a timely manner, but also cause the intercom terminal to frequently remind the user to distract the user from their work focus, increasing the risk of operational errors (e.g., accidentally touching the equipment switch during maintenance). If an emergency request (e.g., a call to stop the equipment due to a misjudgment of the status) is missed, it will delay the opportunity to handle the fault and cause the production line to be shut down for a long time.

[0005] In view of this, to address the problems of low verification efficiency and lack of intelligent adaptation capability in communication scheduling of the above-mentioned identity verification mechanisms, we propose a method and system for intercom access control based on voiceprint recognition. Summary of the Invention

[0006] The purpose of this invention is to solve the problems of low verification efficiency caused by the large size of the voiceprint standard library and the traditional comparison method in the traditional intercom system in modern intelligent manufacturing and large-scale industrial production scenarios, as well as the problems of blind calling, delayed response or interference with work focus caused by the inability to accurately identify the actual status of the user in communication scheduling.

[0007] To achieve the above objectives, the present invention provides a method for managing intercom access based on voiceprint recognition, comprising the following steps:

[0008] S1, Perception Permission List, Voiceprint Template Library, establish a complete mapping chain based on the same user ID in them;

[0009] S2. Establish a dual verification mechanism to verify user identity: obtain the intercom serial number of the user's last output voice signal and verify the intercom serial number; after verification, retrieve the corresponding standard voiceprint features in the complete mapping chain, use the voiceprint extraction method to extract the real-time voiceprint features corresponding to the output voice, and verify whether the real-time voiceprint features and the standard voiceprint features match.

[0010] If the serial number and voiceprint feature verification fail, a warning signal is output; if both the serial number and voiceprint feature verification are successful, the user ID corresponding to the complete mapping chain in step S1 is marked as pending; a time threshold is set, and the time interval between the user's last voice output and the current time node is calculated. If the time interval is greater than the time threshold, a reminder signal is output to prompt the user to output verification voice through the intercom terminal.

[0011] After requesting the user to output the requested voice message:

[0012] S3. Receive the request voice after the user's request has been successfully verified in step S2, and the corresponding voiceprint feature of the request voice; identify the access request of the response role and the workshop in the voiceprint feature; match multiple user IDs corresponding to the response role according to the permission list in step S1, and define the user IDs in the waiting state in the complete mapping chain as the response users; receive the time interval corresponding to each response user in step S2, and determine the order of establishing a one-way communication link between the request user and multiple response users according to the time interval.

[0013] After the responding user and the requesting user establish a two-way communication link, if the corresponding time interval for the responding user is greater than a time threshold, the responding user will output a verification voice message through the intercom terminal:

[0014] S4. Verify the status of the responding user using the voiceprint extraction method in step S2; retrieve the working time of the same responding user in the work workshop, and then adjust the time threshold of the work workshop in step S2 using the working time.

[0015] As a further improvement to this technical solution, the permission list in S1 covers the user IDs of all roles in the factory, and each role has a different ID tag, as well as intercom permissions between different types of roles; the voiceprint template library includes the binding relationship between each user ID, standard voiceprint features and intercom serial number, and each user ID corresponds to a unique standard voiceprint feature.

[0016] Establish the complete mapping chain: extract all user IDs from the permission list and all user IDs from the voiceprint template library; based on the principle that the ID characters are completely consistent, compare the same user IDs, retrieve the intercom permissions corresponding to the same user ID in the permission list, as well as the corresponding standard voiceprint features and serial numbers in the voiceprint template library, thereby establishing a complete mapping chain between user ID - standard voiceprint features - intercom permissions - serial numbers.

[0017] As a further improvement to this technical solution, S2 acquires the last voice signal output by the user through the intercom terminal, the time node, and the intercom terminal serial number; defines the time node of the last voice signal output as the silent timing start node, defines the current time node as the permission status verification node; and calculates the time interval between the silent timing start node and the permission status verification node.

[0018] Set a time threshold. If the time interval is greater than the time threshold, the serial number of the intercom terminal corresponding to the last voice output is defined as the output serial number. Send a reminder signal to the corresponding intercom terminal based on the output serial number. Set the time length and define the time node of the output reminder signal as the start time. Set the verification interval through the time length and obtain the serial numbers of all intercom terminals with output voice signals within the verification interval, which are defined as the verification serial numbers.

[0019] Compare the output serial number and the verification serial number:

[0020] If the output serial number is not equal to the verification serial number, the output warning signal will be traced back to the walkie-talkie corresponding to the verification serial number;

[0021] If the output serial number equals the verification serial number, then the serial number verification is considered successful.

[0022] As a further improvement to this technical solution, after the S2 output sequence number and verification sequence number are successfully verified, the standard voiceprint feature corresponding to the verification sequence number is directly retrieved from the complete mapping chain by retrieving the verification sequence number, and the real-time voiceprint feature corresponding to the output speech is extracted using the voiceprint extraction method; the standard voiceprint feature in the complete mapping chain is retrieved based on the verification sequence number to verify whether the real-time voiceprint feature and the standard voiceprint feature match.

[0023] If the real-time voiceprint feature and standard voiceprint feature verification fail, a warning signal will be output again to trace back to the walkie-talkie corresponding to the verification serial number, thereby prompting the user to re-output the voice signal through the walkie-talkie.

[0024] If the real-time voiceprint feature and standard voiceprint feature are successfully verified, the silent timing start node is refreshed, and the user ID corresponding to the complete mapping chain in step S1 is marked as pending.

[0025] 5. The intercom access control method based on voiceprint recognition according to claim 2, characterized in that: in step S2, an upper limit for the number of inputs is set; if multiple user voice outputs fail to be verified, and the number of multiple user voice outputs equals the upper limit for the number of inputs, an alarm signal is output to the administrator.

[0026] As a further improvement to this technical solution, step S2 uses a voiceprint extraction method to extract the real-time voiceprint features corresponding to the output speech. And to verify real-time voiceprint features and standard voiceprint characteristics Does the space match?

[0027] Noise signals from each workshop were collected, and the noise spectrum for each workshop was obtained through Fourier transform. and the noisy speech spectrum of the output speech. ;

[0028] Then, the noise spectrum of each workshop is used sequentially for noise reduction processing of the noisy speech spectrum. The denoised speech spectrum is obtained as follows: ,in For over-subtraction factor, For the first Noise spectrum of each workshop;

[0029] For each denoised speech spectrum, the corresponding real-time speaker features are extracted using a speaker extraction method. This forms a set of real-time speaker features under different noise spectra. ;

[0030] Calculate the real-time voiceprint feature set sequentially Each real-time voiceprint feature Compared with standard voiceprint characteristics The matching degree between them is used to retrieve the real-time voiceprint feature set. Standard voiceprint characteristics The real-time voiceprint feature with the highest matching degree Defined as real-time voiceprint features Set a matching threshold; if the matching degree > the matching threshold, then determine the real-time voiceprint features. Standard voiceprint characteristics match.

[0031] As a further improvement to this technical solution, after S3 establishes a one-way communication link with a certain response, it outputs a prompt signal including a request voice to the corresponding response user through the one-way communication link; then the response user outputs a response voice through the corresponding intercom terminal, and verifies the response user again through the dual verification mechanism described in step S2. After the response user is successfully verified, the response signal in the response voice is displayed.

[0032] If the response signal is "agree", a two-way communication link is established between the responding user and the requesting user, and the establishment of a one-way communication link between the requesting user and other responding users is stopped.

[0033] If the response signal is "reject", then a one-way communication link is established between the next responding user and the requesting user, until the response signal is "agree".

[0034] As a further improvement to this technical solution, after the responding user and the requesting user establish a two-way communication link, S4 receives the user ID corresponding to the output warning signal in step S2. If the user ID is the same as the responding user ID, the work workshop of the requesting user in step S3 is retrieved; the noise spectrum corresponding to the work workshop is matched, and the corresponding noise spectrum is defined as the noise reduction spectrum.

[0035] If, in step S2, when responding to the user's voice verification, the voiceprint extraction method extracts the real-time voiceprint features corresponding to the output voice, it marks the real-time voiceprint features corresponding to the noise reduction spectrum, and then determines whether the real-time voiceprint features are equal to the corresponding real-time voiceprint features. If they are equal, it is determined that the responding user is in a working state; if they are not equal, it is marked that the user ID corresponding to the responding user is in a standby state in the complete mapping chain.

[0036] As a further improvement to this technical solution, in step S4, when it is determined that the responding user is in a working state, the requesting user outputs a request voice as the work start node, and the time difference from the work start node to the current time node is calculated as the working time corresponding to the work workshop.

[0037] Receive the working time of multiple users with the same role in the work workshop, and adjust the time threshold of the work workshop in step S2 based on the working time.

[0038] A voiceprint recognition-based intercom access control system includes:

[0039] The module for constructing and verifying the association system establishes a complete mapping chain.

[0040] The voice permission verification and communication link management module obtains the intercom serial number of the user's last output voice signal and verifies its intercom serial number in the complete mapping chain. After successful verification, it retrieves the corresponding standard voiceprint features in the complete mapping chain, uses the voiceprint extraction method to extract the real-time voiceprint features corresponding to the output voice, and verifies whether the real-time voiceprint features and the standard voiceprint features match.

[0041] If the serial number and voiceprint feature verification fail, an early warning signal will be output; if both the serial number and voiceprint feature are successfully verified, the user ID corresponding to the complete mapping chain in the tag association system construction and adaptability verification module will be in a standby state.

[0042] After requesting the user to output the requested voice message:

[0043] The bidirectional communication link establishment module receives the request voice after the voice permission verification and communication link management module has successfully verified the requesting user, as well as the request voiceprint features corresponding to the request voice; it identifies the access request, which includes the response role and the work workshop, in the request voiceprint features; it matches multiple user IDs corresponding to the response role with the permission list in the association system construction and adaptability verification module, and defines the user IDs in the waiting state in the complete mapping chain as the response users; it sequentially establishes the order of establishing unidirectional communication links between the requesting user and multiple response users.

[0044] The verification duration optimization module verifies the status of the responding user when the responding user outputs verification voice through the intercom terminal after establishing a two-way communication link with the requesting user. It then retrieves the working duration of the same responding user in the workshop and adjusts the time threshold of the workshop in the voice permission verification and communication link management module based on the working duration.

[0045] In addition to the objectives, features, and advantages described above, the present invention has other objectives, features, and advantages. The invention will now be described in further detail with reference to the figures. Attached Figure Description

[0046] Figure 1 This is a schematic diagram of the output reminder signal process in this invention;

[0047] Figure 2 This is a schematic diagram illustrating the working principle of the dual verification mechanism of this invention.

[0048] Figure 3 This is a schematic diagram illustrating the user status and the working principle of establishing a one-way communication link in this invention.

[0049] The meanings of the labels in the diagram are as follows:

[0050] 100. Association system construction and compatibility verification module; 200. Voice permission verification and communication link management module; 300. Two-way communication link establishment module; 400. Verification duration optimization module. Detailed Implementation

[0051] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0052] To accurately differentiate and standardize the intercom permissions of different roles within the factory (such as front-line operators, equipment maintenance technicians, workshop supervisors, and temporary outsourced personnel) regarding equipment operation, area access, and information exchange during the production process, and to effectively bind user identities with intercom permissions to ensure the safe and compliant operation of the intercom system, reference was made to... Figures 1-3 As shown, a method for managing intercom access based on voiceprint recognition includes the following steps:

[0053] Step S1: Obtain the permission list and voiceprint template library; compare the user IDs in the permission list and voiceprint template library, retrieve the same user IDs in the permission list and voiceprint template library, and establish a complete mapping chain between user ID - standard voiceprint feature - intercom permission - serial number.

[0054] The permission list covers user IDs for all roles within the factory (such as frontline operators, equipment maintenance technicians, workshop supervisors, etc.), and each role type has different ID tags (e.g., "GZ" represents operators, "WX" represents maintenance technicians), as well as intercom permissions between different role types (including two core permissions: communication range and function operation); a specific example is as follows:

[0055] Character type Employee ID Label User ID Communication range Function operation permissions front-line operators GZ + number GZ001 Only call the team leader (starting with ZZ) and the repairman (starting with WX). One-on-one calls, fault reporting requests Equipment maintenance technician WX + number WX001 Call all operators (starting with GZ) and maintenance team leaders (WX000) One-to-one / one-to-many calls, maintenance dispatch notifications Workshop Supervisor ZG+number ZG001 All employees of the Huben workshop and the dispatch center (starting with DD) Workshop broadcasts and production instruction push notifications

[0056] The voiceprint template library includes the binding relationship between each user's employee ID, standard voiceprint features, and intercom serial number. Each user's employee ID corresponds to a unique standard voiceprint feature. A specific example is as follows:

[0057] User ID Name Standard voiceprint characteristics Intercom serial number GZ001 Zhang San Male, fundamental frequency 110-150Hz, draws out the word "repair". TP-8000-SN001 WX001 Li Si Male, fundamental frequency 105-145Hz, emphasizes the word "repair" TP-8000-SN002 ZG001 Wang Wu Male, fundamental frequency 120-160Hz, pauses when saying "instruction". TP-8000-SN003

[0058] The working principle of establishing a complete mapping chain in step S1 is as follows: extract all user IDs from the permission list, and denote them as a set. Extract all user IDs from the voiceprint template library again, and denote them as set. ;in These are the number of user IDs in the permission list and the voiceprint template library, respectively.

[0059] Based on the principle that the employee ID characters must be completely identical, the user employee IDs in set A and set B are compared and matched. (i.e., user ID) (If the user belongs to both set A and set B, and their employee ID characters are exactly the same), retrieve the user's employee ID. The system establishes a complete mapping chain between user ID, standard voiceprint features, intercom permissions, and serial numbers by matching the intercom permissions in the permission list with the corresponding standard voiceprint features and serial numbers in the voiceprint template library.

[0060] Step S2: Establish a dual verification mechanism to verify user identity: Obtain the intercom serial number of the user's last output voice signal and verify the intercom serial number; after verification, retrieve the corresponding standard voiceprint features from the complete mapping chain. The speaker extraction method is used to extract the real-time speaker features corresponding to the output speech. Verify real-time voiceprint features and standard voiceprint characteristics Check if the serial number or voiceprint feature verification fails; if the serial number or voiceprint feature verification fails, output a warning signal to the serial number of the intercom terminal that outputs the voice signal.

[0061] If both the serial number and voiceprint feature are successfully verified, the user's identity is successfully verified, and the user's employee number corresponding to the complete mapping chain in step S1 is marked as pending. If the serial number and voiceprint feature verification fails, an early warning signal is output to the serial number of the speaker that outputs the voice signal.

[0062] In step S2, the serial number is first output. and verification serial number After successful verification, real-time voiceprint features will be used again. and standard voiceprint characteristics Verification is performed to form a dual verification mechanism of serial number + voiceprint feature, which effectively solves the problems of cross-use of intercom terminals and single voiceprint verification being easily interfered with in traditional intercom systems.

[0063] Specifically:

[0064] The working principle of the dual verification mechanism in step S2 is as follows: The system obtains the last voice signal output by the user through the intercom terminal, the time point, and the intercom terminal serial number; the time point of the last voice signal output is defined as the start point of the silence timing. Define the current time node as the permission status verification node. ; Calculate the start node of silent timing To the permission status verification node Time interval ;

[0065] Time thresholds can be set by collecting and analyzing historical work data from typical factory positions (such as maintenance technicians, process engineers, and operators), including the average duration of a single operation for different positions (such as average equipment maintenance time and production line parameter debugging cycle). (This can be achieved by collecting data on the duration of multiple work sessions at different positions, then calculating the average duration of these multiple work sessions, and setting a time interval.) For example, if the average time a repairman spends on equipment repair is 30 minutes, then the repairman's time threshold is... =30), if the time interval >Time threshold The serial number of the intercom terminal corresponding to the last voice output is defined as the output serial number. Based on the output sequence number, a reminder signal is sent to the corresponding intercom terminal to prompt the user to output a verification voice through the intercom terminal;

[0066] Set time length Define the time node for outputting the reminder signal as the start time. By time length Set verification interval Obtain the verification interval The serial number of all output voice signals from the intercom terminal is defined as the verification serial number. ;

[0067] Compare the output serial numbers and verification serial number If the output serial number =Verification serial number This indicates that the serial numbers are completely identical, confirming successful serial number verification. This establishes the correspondence between the user and the intercom at the device hardware level, ensuring that the intercom initiating the authorization verification request is legitimate. This prevents situations where intercoms are misused (e.g., outsourced personnel mistakenly using production workshop-specific intercoms) if there is a lack of hardware-level serial number verification in large factories with numerous intercoms distributed across different workshops and positions. By comparing the output serial numbers... and verification serial number Verification is conducted to prevent cross-platform misuse from the source.

[0068] Output serial number Verify serial number After successful verification, retrieve the verification serial number. The verification serial number can be retrieved directly from the complete mapping chain. Corresponding standard voiceprint features The real-time voiceprint features corresponding to the output speech are extracted using a voiceprint extraction method. Based on the verification serial number Retrieve standard voiceprint features from the complete mapping chain Verify real-time voiceprint features and standard voiceprint characteristics Does the pair match?

[0069] The above method directly retrieves the standard voiceprint features corresponding to the user's output voice. To avoid the need to verify voiceprint features, the real-time voiceprint features corresponding to the output speech still need to be verified. The tedious process of comparing each voiceprint feature with other voiceprint templates in the database one by one;

[0070] Since voiceprint template libraries typically contain voiceprint data for all personnel within a factory, comparing them one by one would significantly increase verification time. However, outputting serial numbers... First, narrow down the range of user identities, then retrieve the corresponding standard voiceprint features. Targeted comparison avoids the waste of resources and efficiency loss caused by indiscriminate full comparison, and allows voiceprint verification to focus on the accurate confirmation of legitimate users' identities, thereby significantly improving the overall efficiency of permission verification while ensuring the accuracy of verification.

[0071] If real-time voiceprint features and standard voiceprint characteristics If verification fails, an early warning signal will be output, and the process will be traced back to the verification sequence number. The corresponding walkie-talkie prompts the user to re-output the voice signal through the walkie-talkie terminal.

[0072] It also sets an upper limit on the number of input attempts. If multiple user voice outputs fail to verify, and the number of multiple user voice outputs equals the upper limit of the number of input attempts, an alarm signal will be output to the administrator to indicate that there is an unauthorized verification attempt or abnormal user operation.

[0073] If real-time voiceprint features and standard voiceprint characteristics After successful verification, the silent timer start node will be refreshed. The system marks the user ID corresponding to the complete mapping chain in step S1 as being in a standby state, thereby ensuring that the permissions corresponding to the user ID always match the actual usage status, avoiding the situation where the user is absent from work for a long time, and providing a clear basis for subsequent calls to the corresponding user.

[0074] In step S2, the speaker extraction method is used to extract the real-time speaker features corresponding to the output speech. And to verify real-time voiceprint features and standard voiceprint characteristics The working principle of whether the two match is:

[0075] Considering the varying noise levels across different workshops in the factory, failure to denoise the output speech will significantly reduce speech recognition accuracy and lead to subsequent access requests failing. Therefore, noise signals from each workshop are collected, and the noise spectrum for each workshop is obtained through Fourier transform. and the noisy speech spectrum of the output speech. ;

[0076] Then, the noise spectrum of each workshop is used sequentially for noise reduction processing of the noisy speech spectrum. The denoised speech spectrum is obtained as follows: ,in This is an over-subtraction factor used to balance noise reduction and speech distortion; For the first Noise spectrum of each workshop;

[0077] For each denoised speech spectrum, the corresponding real-time speaker features are extracted using speaker extraction methods (such as Mel-frequency cepstral coefficients (MFCC), linear predictive cepstral coefficients (LPCC), etc.). Ultimately, this forms a set of real-time speakerprint features under different noise spectra. .

[0078] Calculate the real-time voiceprint feature set sequentially Each real-time voiceprint feature Compared with standard voiceprint characteristics The matching degree between them is used to retrieve the real-time voiceprint feature set. Standard voiceprint characteristics The real-time voiceprint feature with the highest matching degree Defined as real-time voiceprint features Set a matching threshold; if the matching degree > the matching threshold, then determine the real-time voiceprint features. Standard voiceprint characteristics match;

[0079] The formula for calculating the matching degree is as follows: Real-time voiceprint features The feature vector is Standard voiceprint characteristics The feature vector is Matching degree ;

[0080] Due to differences in equipment operation and production processes in different workshops, the noise spectrum is unique. The Fourier transform in step S2 can convert the time-domain noise signal into the frequency-domain spectral characteristics, thereby clearly presenting the frequency composition of the noise and providing a precise noise spectrum for subsequent targeted noise reduction.

[0081] Then, the difference between the noise spectrum and the noisy speech spectrum is used to reasonably reduce noise, thereby suppressing noise while preserving the effective voiceprint information of the speech to the greatest extent. In step S2, the voiceprint feature extraction stage, a purer and more distinguishable speech spectrum can be obtained through the voiceprint extraction algorithm. This effectively avoids the problem of decreased speech recognition accuracy caused by the noisy factory environment, further ensuring the reliability of the speech recognition process and making the correspondence between voiceprint and identity more accurate.

[0082] Furthermore, in intercom systems, the user typically outputs a request voice, which usually clearly indicates a corresponding responding user and the specific business collaboration requirements (such as the specific fault information and collaboration goals in "Please have maintenance engineer Li confirm the handling plan for fault code E07 on the robotic arm of production line 3"). Then, a two-way communication link is established between the requesting user and the responding user.

[0083] However, in intercom systems, to avoid situations where the requesting user blindly outputs a voice request to establish a two-way communication link without specifying the responding user's status, leading to the failure of the link establishment, the responding user actively outputs their status (specifically, including working status and standby status), and then the requesting user can choose based on the responding user's status.

[0084] However, due to the dynamic nature of factory operations (e.g., a responding user may receive an emergency maintenance task and enter working status immediately after outputting a standby status, or quickly complete the current task and switch back to standby status while working), and the responding user may forget to update the status information in a timely manner due to focusing on on-site operations, this can lead to inaccurate status information. As a result, the requesting user may select a responding user and initiate a communication request based on incorrect status information (e.g., selecting a responding user that is displayed as standby but has actually entered working status). This will not only still result in the failure to establish a two-way communication link, but will also increase the requesting user's invalid operations due to misleading information (e.g., repeatedly sending requests to users whose status is displayed incorrectly).

[0085] Therefore, to avoid situations where the requesting user blindly calls the responding user, in this invention, S3 involves receiving the request voice after the requesting user has been successfully verified in step S2, as well as the request voiceprint feature corresponding to the request voice. ;

[0086] Automatic speech recognition technology is used to extract the voiceprint features of the request. The access request includes the responding role and the work workshop.

[0087] The response role is the specific job position that needs to respond to the request when the user outputs a request voice through the intercom terminal; for example, when the user (such as a front-line operator, with the employee ID label "GZ") outputs the request voice "call the equipment maintenance technician to check machine tool No. 3", the "equipment maintenance technician" is the response role.

[0088] The work workshop is, for example, when the user is asked to output "Call an equipment repairman to check the equipment in workshop 3", "workshop 3" is the work workshop.

[0089] In step S3, automatic speech recognition technology extracts the voiceprint features of the requester. The working principle of access requests in China is as follows:

[0090] A large amount of voiceprint feature data from known access requests was collected as training samples to train a speech recognition model. Then, a recurrent neural network was used to train the speech recognition model. After training, the input request voiceprint features were... The speech recognition model outputs the requested voiceprint features. The corresponding access request;

[0091] A pre-set keyword library (containing business-related keywords such as response roles and workshops, like "repairman" and "workshop 3") is used to compare access requests with keywords in the pre-set keyword library. If an access request contains a certain type of keyword, the access request corresponding to the voice request is determined. For example, if the access request mentions calling a repairman, then the repairman is determined to be the response role in the access request.

[0092] Extracting voiceprint features After the access request is received, step S3 matches the employee ID tag of the response role according to the permission list in step S1, and then retrieves multiple user employee IDs of the corresponding response role based on the employee ID tag. The user employee ID in the waiting state in the complete mapping chain is defined as the response user, thereby avoiding scheduling users without permissions or in the working state.

[0093] Furthermore, step S3 receives the time interval corresponding to each responding user in step S2. Calculate each time interval With time threshold The time difference between the request and response users is sorted from largest to smallest. One-way communication links are then established between the requesting user and multiple response users in sequence. This allows for priority scheduling of response users with long idle times and high response timeliness, thereby improving request processing efficiency.

[0094] After establishment, a prompt signal including the request voice is output to the corresponding responding user through a one-way communication link. Then, the responding user outputs a response voice through the corresponding intercom terminal. The responding user is verified through the dual verification mechanism in step S2. After the responding user is successfully verified, the response signal in the response voice is extracted using automatic speech recognition technology. If the response signal is "agree", a two-way communication link is established between the responding user and the requesting user, and the establishment of one-way communication links between the requesting user and other responding users is stopped. The two-way communication link facilitates the subsequent transmission of voice data between the responding user and the requesting user.

[0095] If the response signal is "reject", then a one-way communication link is established between the next responding user and the requesting user, until the response signal is "agree".

[0096] The present invention further considers that, in order to prevent users from leaving their posts in step S2, a time threshold is set. This requires users to periodically output voice signals, and then their identity is verified to determine whether they are in a standby state;

[0097] After the requesting user and the responding user establish a two-way communication link, since communication in industrial scenarios often revolves around specific business needs (such as a front-line operator calling an equipment repairman to explain a machine tool malfunction), the communication content has a clear task orientation. Once the key information such as the fault description and repair plan is transmitted, the immediate need for two-way communication is terminated. Therefore, after communication, the requesting user and the responding user may be able to trigger a link disconnection command through a preset command (such as a "communication ended" voice command) or a physical button on the terminal to disconnect the two-way communication link. Then the responding user enters the working state (such as a repairman going to the site to inspect the equipment).

[0098] If step S2 judges that the responding user is in a standby state based solely on historical verification results without dynamically adapting to their actual work scenario, the system will misjudge that the responding user is still in a standby state where they can respond to communication immediately. When other users initiate the same type of request (such as a fault call from another device), step S2 will include the repairman in the responding user list and attempt to establish a communication link. However, the repairman is currently working on-site and cannot respond to new requests in a timely manner. Furthermore, the frequent calls from the intercom may interfere with on-site repair operations, thereby delaying the business processes of fault handling and production scheduling.

[0099] Simultaneously, once the user enters working mode, their work primarily involves highly focused, hands-on tasks (such as equipment disassembly and parameter tuning), making it impossible to periodically output voice signals for verification as in a regular on-duty state. If the time threshold in step S2... If the general settings are still used without scenario-specific adjustments, the user will not output voice messages on time while focused on their work, causing the warning signals to be frequently triggered in step S2. This will disrupt the on-site work rhythm of the technicians and other users, affecting work efficiency and operational safety.

[0100] Therefore, to avoid the above situation, after the responding user and the requesting user establish a two-way communication link, the user ID corresponding to the output warning signal in step S2 is received. If the user ID is the same as the responding user ID, the work workshop of the requesting user in step S3 is retrieved; the noise spectrum corresponding to the work workshop is matched, and the corresponding noise spectrum is defined as the noise reduction spectrum.

[0101] If in step S2, when responding to the user's voice verification, the voiceprint extraction method extracts the real-time voiceprint features corresponding to the output voice. At that time, the real-time speakerprint features corresponding to the denoised spectrum are marked. Then determine the real-time voiceprint features. Is it equal to real-time voiceprint features? If the value is equal to the value, then it is determined that the responding user is in a working state.

[0102] If the values ​​are not equal, the corresponding user ID of the responding user is marked as being in a standby state in the complete mapping chain to avoid misjudgment of the responding user's status; and noise reduction through the corresponding noise reduction spectrum can complete the identity verification without the user frequently outputting voice (i.e., multiple verification failures in step S2) during the responding user's work, allowing the responding user to focus more on on-site operations (such as equipment maintenance and production operations).

[0103] When determining that the responding user is in a working state, define the requesting user's output voice as the work start node, calculate the time difference between the work start node and the current time node, and use it as the corresponding working time of the work workshop. ;

[0104] Receive responses from multiple users with the same role regarding their working hours in the workshop. The time threshold in step S2 is adjusted by using working hours. Time threshold This is used to adjust the frequency of the warning signal output in step S2 when the user is working in the workshop, so as to avoid the user who is busy receiving warnings frequently due to a fixed time threshold (such as the user who is engaged in high-intensity maintenance work for a long time will not be frequently interrupted).

[0105] Time threshold ,in The proportional adjustment factor ensures that the longer the subsequent working time (i.e., the busier the response to users), the higher the time threshold. The larger the value, the lower the frequency of the warning signal output; for The average working time of users with the same role in the workshop.

[0106] A voiceprint recognition-based intercom access control system includes:

[0107] The module for constructing and verifying the association system uses a permission list and a voiceprint template library to establish a complete mapping chain.

[0108] The voice permission verification and communication link management module obtains the intercom serial number of the user's last output voice signal and verifies its intercom serial number in the complete mapping chain. After successful verification, it retrieves the corresponding standard voiceprint features in the complete mapping chain, uses the voiceprint extraction method to extract the real-time voiceprint features corresponding to the output voice, and verifies whether the real-time voiceprint features and the standard voiceprint features match.

[0109] If the serial number and voiceprint feature verification fail, an early warning signal will be output; if both the serial number and voiceprint feature are successfully verified, the user ID corresponding to the complete mapping chain in the tag association system construction and adaptability verification module will be in a standby state.

[0110] After requesting the user to output the requested voice message:

[0111] The two-way communication link establishment module receives the request voice after the voice permission verification and communication link management module has successfully verified the requesting user, as well as the corresponding voiceprint features of the request voice; it identifies the access request, including the response role and work workshop, in the voiceprint features of the request; it matches multiple user IDs corresponding to the response role with the permission list in the association system construction and adaptability verification module, and defines the user IDs in the waiting state in the complete mapping chain as the response users; it sequentially establishes the order of establishing one-way communication links between the requesting user and multiple response users.

[0112] The verification duration optimization module, after establishing a two-way communication link between the responding user and the requesting user, verifies the responding user's status when the responding user outputs verification voice through the intercom terminal if the corresponding time interval of the responding user is greater than the time threshold; it then retrieves the working duration of the same responding user in the work workshop and adjusts the time threshold of the work workshop in the voice permission verification and communication link management module based on the working duration.

[0113] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for managing intercom access based on voiceprint recognition, characterized in that, Includes the following steps: S1, Perception Permission List, Voiceprint Template Library, establish a complete mapping chain based on the same user ID in them; S2. Establish a dual verification mechanism to verify user identity: obtain the intercom serial number of the user's last output voice signal and verify the intercom serial number; after verification, retrieve the corresponding standard voiceprint features in the complete mapping chain, use the voiceprint extraction method to extract the real-time voiceprint features corresponding to the output voice, and verify whether the real-time voiceprint features and the standard voiceprint features match. If the serial number and voiceprint feature verification fail, a warning signal is output; if both the serial number and voiceprint feature verification are successful, the user ID corresponding to the complete mapping chain in step S1 is marked as pending; a time threshold is set, and the time interval between the user's last voice output and the current time node is calculated. If the time interval is greater than the time threshold, a reminder signal is output to prompt the user to output verification voice through the intercom terminal. After requesting the user to output the requested voice message: S3. Receive the request voice after the user has been successfully verified in step S2, and the corresponding voiceprint feature of the request voice. The voiceprint features of the request include access requests for response roles and work workshops; based on the permission list in step S1, multiple user IDs corresponding to the response roles are matched, and user IDs in the waiting state in the complete mapping chain are defined as response users; the time interval corresponding to each response user in step S2 is received, and the order of establishing a one-way communication link between the requesting user and multiple response users is determined based on the time interval. After the responding user and the requesting user establish a two-way communication link, if the corresponding time interval for the responding user is greater than a time threshold, the responding user will output a verification voice message through the intercom terminal: S4. Verify the status of the responding user using the voiceprint extraction method in step S2; retrieve the working time of the same responding user in the work workshop, and then adjust the time threshold of the work workshop in step S2 using the working time.

2. The intercom access control method based on voiceprint recognition according to claim 1, characterized in that: The permission list in S1 covers the user IDs of all roles in the factory, and each role has a different ID tag, as well as intercom permissions between different roles; the voiceprint template library includes the binding relationship between each user ID, standard voiceprint features and intercom serial number, and each user ID corresponds to a unique standard voiceprint feature. Establish the complete mapping chain: extract all user IDs from the permission list and all user IDs from the voiceprint template library; based on the principle that the ID characters are completely consistent, compare the same user IDs, retrieve the intercom permissions corresponding to the same user ID in the permission list, as well as the corresponding standard voiceprint features and serial numbers in the voiceprint template library, thereby establishing a complete mapping chain between user ID - standard voiceprint features - intercom permissions - serial numbers.

3. The intercom access control method based on voiceprint recognition according to claim 2, characterized in that: S2 acquires the last voice signal output by the user through the intercom terminal, the time node, and the intercom terminal serial number; defines the time node of the last voice signal output as the silent timing start node, defines the current time node as the permission status verification node; and calculates the time interval between the silent timing start node and the permission status verification node. Set a time threshold. If the time interval is greater than the time threshold, the serial number of the intercom terminal corresponding to the last voice output is defined as the output serial number. Send a reminder signal to the corresponding intercom terminal based on the output serial number. Set the time length and define the time node of the output reminder signal as the start time. Set the verification interval through the time length and obtain the serial numbers of all intercom terminals with output voice signals within the verification interval, which are defined as the verification serial numbers. Compare the output serial number and the verification serial number: If the output serial number is not equal to the verification serial number, the output warning signal will be traced back to the walkie-talkie corresponding to the verification serial number; If the output serial number equals the verification serial number, then the serial number verification is considered successful.

4. The intercom access control method based on voiceprint recognition according to claim 3, characterized in that: After the S2 output sequence number and verification sequence number are successfully verified, the standard voiceprint feature corresponding to the verification sequence number is directly retrieved from the complete mapping chain by calling up the verification sequence number, and the real-time voiceprint feature corresponding to the output speech is extracted by the voiceprint extraction method. Based on the verification sequence number, retrieve the standard voiceprint features from the complete mapping chain and verify whether the real-time voiceprint features match the standard voiceprint features. If the real-time voiceprint feature and standard voiceprint feature verification fail, a warning signal will be output again to trace back to the walkie-talkie corresponding to the verification serial number, thereby prompting the user to re-output the voice signal through the walkie-talkie. If the real-time voiceprint feature and standard voiceprint feature are successfully verified, the silent timing start node is refreshed, and the user ID corresponding to the complete mapping chain in step S1 is marked as pending.

5. The intercom access control method based on voiceprint recognition according to claim 2, characterized in that: S2 sets an upper limit for the number of input attempts. If multiple user voice outputs fail to verify, and the number of user voice outputs equals the upper limit for the number of input attempts, an alarm signal is sent to the administrator.

6. The intercom access control method based on voiceprint recognition according to claim 4, characterized in that: S2 uses a voiceprint extraction method to extract the real-time voiceprint features corresponding to the output speech. And to verify real-time voiceprint features and standard voiceprint characteristics Does the space match? To address the differentiated noise scenarios in different workshops of the factory, noise signals from each workshop were collected, and the noise spectrum for each workshop was obtained through Fourier transform. and the noisy speech spectrum of the output speech. ; Then, the noise spectrum of each workshop is used sequentially for noise reduction processing of the noisy speech spectrum. The denoised speech spectrum is obtained as follows: ,in For over-subtraction factor, For the first Noise spectrum of each workshop; For each denoised speech spectrum, the corresponding real-time speaker features are extracted using a speaker extraction method. This forms a set of real-time speaker features under different noise spectra. ; Calculate the real-time voiceprint feature set sequentially Each real-time voiceprint feature Compared with standard voiceprint characteristics The matching degree between them is used to retrieve the real-time voiceprint feature set. Standard voiceprint characteristics The real-time voiceprint feature with the highest matching degree Defined as real-time voiceprint features ; Set a matching threshold; if the matching degree is greater than the matching threshold, then determine the real-time voiceprint features. Standard voiceprint characteristics match.

7. The intercom access control method based on voiceprint recognition according to claim 1, characterized in that: After establishing a one-way communication link with a certain responder in step S3, a prompt signal including the request voice is output to the corresponding responder through the one-way communication link; then the responder outputs a response voice through the corresponding intercom terminal, and verifies the responder again through the dual verification mechanism described in step S2. After the responder is successfully verified, the response signal in the response voice is displayed. If the response signal is "agree", a two-way communication link is established between the responding user and the requesting user, and the establishment of a one-way communication link between the requesting user and other responding users is stopped. If the response signal is "reject", then a one-way communication link is established between the next responding user and the requesting user, until the response signal is "agree".

8. The intercom access control method based on voiceprint recognition according to claim 6, characterized in that: After the responding user and the requesting user establish a two-way communication link, S4 receives the user ID corresponding to the output warning signal in step S2. If the user ID is the same as the responding user ID, the work workshop of the requesting user in step S3 is retrieved; the noise spectrum corresponding to the work workshop is matched, and the corresponding noise spectrum is defined as the noise reduction spectrum. If, in step S2, when responding to the user's voice verification, the voiceprint extraction method extracts the real-time voiceprint features corresponding to the output voice, it marks the real-time voiceprint features corresponding to the noise reduction spectrum, and then determines whether the real-time voiceprint features are equal to the corresponding real-time voiceprint features. If they are equal, it is determined that the responding user is in a working state; if they are not equal, it is marked that the user ID corresponding to the responding user is in a standby state in the complete mapping chain.

9. The intercom access control method based on voiceprint recognition according to claim 8, characterized in that: In step S4, when it is determined that the responding user is in a working state, the requesting user outputs a request voice as the work start node, and the time difference from the work start node to the current time node is calculated as the working time corresponding to the work workshop. Receive the working time of multiple users with the same role in the work workshop, and adjust the time threshold of the work workshop in step S2 based on the working time.

10. A voiceprint recognition-based intercom access control system, applied to the voiceprint recognition-based intercom access control method according to any one of claims 1-9, characterized in that, include: The association system construction and compatibility verification module (100) establishes a complete mapping chain using permission lists and voiceprint template libraries; The voice permission verification and communication link management module (200) obtains the intercom serial number of the user's last output voice signal and verifies its intercom serial number in the complete mapping chain. After successful verification, it retrieves the corresponding standard voiceprint feature in the complete mapping chain, uses the voiceprint extraction method to extract the real-time voiceprint feature corresponding to the output voice, and verifies whether the real-time voiceprint feature and the standard voiceprint feature match. If the serial number and voiceprint feature verification fails, an early warning signal will be output; if both the serial number and voiceprint feature are successfully verified, the user ID corresponding to the complete mapping chain in the marker association system construction and adaptability verification module (100) will be in a standby state. After requesting the user to output the requested voice message: The two-way communication link establishment module (300) receives the request voice after the voice permission verification and communication link management module (200) has successfully verified the requesting user, as well as the request voiceprint feature corresponding to the request voice; The voiceprint features of the request include access requests for the response role and the workshop; based on the permission list in the association system construction and compatibility verification module (100), multiple user IDs corresponding to the response role are matched, and user IDs in the waiting state in the complete mapping chain are defined as response users; a one-way communication link is established between the requesting user and multiple response users in sequence; The verification duration optimization module (400) verifies the status of the responding user when the responding user outputs verification voice through the intercom terminal after the responding user and the requesting user establish a two-way communication link. It then retrieves the working duration of the same responding user in the work workshop and adjusts the time threshold of the work workshop in the voice permission verification and communication link management module (100) based on the working duration.

Images