A Power Grid Dispatch Identity Authentication Method and System Based on Voiceprint Recognition

Abstract

This invention proposes a power grid dispatch identity authentication method and system based on voiceprint recognition. It identifies the dispatch voice of the personnel to be identified, and then introduces dynamic password verification to eliminate false dispatch voice and verify the real identity of the on-site personnel, thus solving the problem of deceptive voice.

Description

Technical Field

[0001] This invention belongs to the field of communication-related technology, and in particular relates to a power grid dispatching identity authentication method and system based on voiceprint recognition. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] With the maturation of artificial intelligence technology and the increasing integration of equipment in the new generation of power grid systems, the demand for information-based management in power grid systems will inevitably grow daily. At the same time, the need for "unified centralized dispatching and hierarchical management" in dispatch systems leads to a continuous increase in the amount of data processed by dispatch centers. Although power grid dispatch systems, especially dispatch telephone systems, have established relatively complete and efficient multi-level access control mechanisms that can promptly meet the real-time dispatching needs of daily system operations, various system security risks still exist during implementation.

[0004] The specific problems are analyzed as follows:

[0005] First, there are instances of unauthorized personnel or staff from other departments performing operations beyond their normal authority. The root cause of this problem lies in the fact that the existing dispatch system primarily relies on manual dispatch via telephone, lacking sufficient means of personnel authentication in the dispatch process.

[0006] Secondly, the lack of secondary confirmation and authentication of special dispatch instructions poses a potential threat to power grid security. With the advancement of power grid intelligence, the risk of erroneous dispatch operations is becoming increasingly severe, such as unauthorized dispatch and incorrect dispatch. These risks constantly threaten the normal operation of the power grid; therefore, strengthening power grid risk management and improving the level of power grid dispatch operations are crucial aspects of power grid operational safety. Summary of the Invention

[0007] To overcome the shortcomings of the existing technology, this invention provides a power grid dispatch identity authentication method and system based on voiceprint recognition. Combining the characteristics of power grid operation and control requirements, the method identifies the dispatch voice of the personnel to be identified, and then introduces dynamic password verification to eliminate false dispatch voice and verify the real identity of the on-site personnel, thus solving the problem of deceptive voice.

[0008] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions: The first aspect of the present invention provides a power grid dispatching authentication method based on voiceprint recognition, comprising:

[0009] Obtain the dispatch voice data of the personnel to be verified;

[0010] The acquired scheduling voice data is input into the trained voiceprint recognition model to extract the first voiceprint discriminative feature vector. The extracted first voiceprint discriminative feature vector is then matched with the voiceprint discriminative feature vector in the voiceprint database for voiceprint authentication.

[0011] If the match is successful, a random password text is generated, and the dynamic password voice response from the person to be verified is obtained. The obtained dynamic password voice is input into the speech recognition model and converted into speech text information. The speech text information is then dynamically verified and matched with the random password text. If they match, the identity of the person to be verified is successfully authenticated.

[0012] A second aspect of the present invention provides a power grid dispatching authentication system based on voiceprint recognition, comprising:

[0013] Business application module: Acquire the voice data of the person to be verified;

[0014] Voiceprint authentication module: The acquired scheduling voice data is input into the trained voiceprint recognition model to extract the first voiceprint discriminative feature vector, and the extracted first voiceprint discriminative feature vector is matched with the voiceprint discriminative feature vector in the voiceprint database for voiceprint authentication.

[0015] Dynamic verification module: If the match is successful, a random password text is generated, and the dynamic password voice response from the person to be verified is obtained. The obtained dynamic password voice is input into the speech recognition model and converted into speech text information. The speech text information is dynamically verified and matched with the random password text. If they match, the identity of the person to be verified is successfully authenticated.

[0016] A third aspect of the present invention provides a computer-readable storage medium for storing computer instructions, which, when executed by a processor, perform the steps described in the above method.

[0017] A fourth aspect of the present invention provides an electronic device including a memory and a processor, and computer instructions stored in the memory and running on the processor, wherein the computer instructions, when executed by the processor, perform the steps described in the above method.

[0018] The above one or more technical solutions have the following beneficial effects:

[0019] In this invention, the dispatch voice of the personnel to be identified is recognized in combination with the characteristics of power grid operation and control requirements. Then, dynamic password verification is introduced to eliminate false dispatch voice and verify the real identity of the on-site personnel, thus solving the problem of deceptive voice.

[0020] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0021] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0022] Figure 1 This is a schematic diagram of the voiceprint registration process in Embodiment 1 of the present invention;

[0023] Figure 2 This is a schematic diagram of the voice recognition process for scheduling instructions in Embodiment 1 of the present invention;

[0024] Figure 3 This is a schematic diagram of the dynamic password voice recognition process in Embodiment 1 of the present invention;

[0025] Figure 4 This is a schematic diagram of a power grid dispatching identity authentication system based on voiceprint recognition in Embodiment 2 of the present invention. Detailed Implementation

[0026] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0027] It should be noted that the terminology used herein is for the purpose of describing particular implementations only and is not intended to limit the exemplary implementations of the present invention.

[0028] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0029] Example 1

[0030] This embodiment discloses a power grid dispatching identity authentication method based on voiceprint recognition, characterized in that it includes:

[0031] Step 1: Obtain the dispatch voice data of the person to be verified;

[0032] Step 2: Input the acquired scheduling voice data into the trained voiceprint recognition model to extract the first voiceprint discriminative feature vector, and perform voiceprint authentication matching between the extracted first voiceprint discriminative feature vector and the voiceprint discriminative feature vector in the voiceprint database.

[0033] Step 3: If the match is successful, a random password text is generated, and the dynamic password voice response from the person to be verified is obtained. The obtained dynamic password voice is input into the speech recognition model and converted into speech text information. The speech text information is dynamically verified and matched with the random password text. If they match, the identity of the person to be verified is successfully authenticated.

[0034] In this embodiment, the voiceprint recognition model uses a convolutional neural network NET1. The training of the voiceprint recognition model is as follows: a large amount of speech data from unrelated personnel, i.e., non-scheduling personnel, is collected and spectrograms are extracted as voiceprint features of the samples. Combined with the unique encoding of the identity of the corresponding personnel for each sample, a training sample set for the voiceprint recognition model is constructed. The convolutional neural network NET1 is trained based on the constructed training sample set to obtain the trained voiceprint recognition model.

[0035] In this embodiment, the convolutional neural network NET1 consists of N1 convolutional layers, N2 pooling layers, and 1 softmax layer. The specific values ​​of N1 and N2 can be adaptively adjusted according to the recognition accuracy.

[0036] like Figure 1 As shown, this embodiment also includes voiceprint registration for dispatchers. After removing the last softmax layer of the convolutional neural network NET1, a discriminative feature extractor NET2 is obtained. The dispatcher's voiceprint features are input into the discriminative feature extractor NET2 to obtain two-dimensional voiceprint discriminative features. Gabor wavelets are used to process the features, forming N two-dimensional feature maps. Next, the 2DPCA algorithm is used to reduce the dimensionality of the feature maps and then perform one-dimensional processing to obtain a voiceprint discriminative feature vector. The obtained voiceprint discriminative feature vector and the one-hot encoding of the dispatcher's identity are stored in the voiceprint database.

[0037] Specifically, the encoding format adopts the one-hot encoding format. For example, if there are only 2 samples, one is encoded as 10 and the other as 01; if there are 4 samples, the encodings are 1000, 0100, 0010, and 000 respectively.

[0038] like Figure 2 As shown, in step 2 of this embodiment, voiceprint authentication of the person to be verified is performed, specifically including:

[0039] Step 2-1: Obtain the dispatch command voice of the person to be verified.

[0040] Step 2-2: Preprocess the acquired dispatch instruction speech to generate a spectrogram, input it into the discriminative feature extractor NET2, and process it using Garbor wavelet and 2DPCA algorithms to extract the first voiceprint discriminative feature vector L.x .

[0041] Steps 2-3: The first voiceprint discriminative feature vector L x The voiceprint discriminative feature vectors {L1, L2, ..., L...} of N dispatchers stored in the voiceprint database N By comparing these values, a cosine distance matrix of size N, {S1, S2, ..., S}, was obtained. N The minimum value in the cosine distance matrix is ​​taken as the first similarity score S. k First similarity score S k The corresponding voiceprint discriminative feature vector is L k .

[0042] Steps 2-4: If the first similarity score S k If the threshold value is less than the first preset threshold D, then the person to be verified is a voiceprint discriminative feature vector L. k If the identity is confirmed, proceed to the dynamic verification stage; otherwise, proceed to steps 2-5.

[0043] Steps 2-5: If the first preset threshold D ≤ the first similarity score S k If the threshold value is less than the second preset threshold R, then the first voiceprint discriminative feature vector L... x Perform a second check; if [|S k -S1|+|S k -S2|+...+|S k -S N If |] / (N-1) > the third preset threshold H, then the voiceprint discriminative feature vector L is determined. k Compared with other voiceprint discriminative feature vectors in the same voiceprint database, voiceprint discriminative feature vector L k Compared with the first voiceprint discriminative feature vector L x With more significant similarity, the person to be verified is identified as the voiceprint discriminative feature vector L. k The system identifies the corresponding identity and enters the dynamic verification stage; otherwise, it is judged as unsuccessful, and an audible and visual alarm is triggered, prompting the person to be verified to re-identify.

[0044] It is understandable that the second preset threshold is greater than the first preset threshold.

[0045] Steps 2-6: If the first similarity score S k If the value is greater than or equal to the second preset threshold R, it is determined that the voiceprint verification has failed, and an audio-visual alarm is triggered to prompt the person to be verified to re-identify.

[0046] In this embodiment, the logical judgments in steps 2-5 above can avoid {L1,L2,……,L NSimultaneously, multiple voiceprint discriminative feature vectors and the first voiceprint discriminative feature vector L... x When voiceprint discriminative feature vectors are similar and their cosine distance is almost identical, the system only considers the first similar voiceprint discriminative feature vector, ignoring other highly similar voiceprint discriminative feature vectors, leading to identity verification errors. Using this step, if multiple voiceprint discriminative feature vectors are simultaneously similar to the first voiceprint discriminative feature vector L... x If the identities are similar and the cosine distance is almost the same, the system determines that there is not enough confidence to verify the identity and prompts the user to re-verify.

[0047] like Figure 3 As shown, in step 3 of this embodiment, dynamic verification is performed on the personnel to be verified, specifically as follows:

[0048] Step 3-1: Generate a random password text from the random text library and display it to the person to be verified. The person to be verified reads and records the dynamic password voice within a preset time. If the timeout is exceeded, the identity verification is deemed to have failed.

[0049] Step 3-2: Input the dynamic password voice into the speech recognition model to obtain the dynamic password voice recognition text;

[0050] The speech recognition model uses Tiny-Transducer, which has the following structure: Conv1d: two layers with convolution kernels of 5 and 3 respectively and a stride of 2 for 4x downsampling; LayerNorm: one layer with 40-dimensional PNCC features as input and 256-dimensional representation as output; LSTM: one layer with 256-dimensional input and output and a hidden dim of 128; FSMN: six layers with a left order of 8 and a right order of [2,2,1,2,2,1].

[0051] Step 3-3: Input the random password text into the speech recognition model to obtain the random password recognition text. Input the random password recognition text into the text-to-pinyin model to obtain the random password pinyin text. If there are polyphonic characters in the random password pinyin text, generate several pinyin text segments. Exhaustively enumerate any combination of pinyin for each character in the text segment to form a speech recognition pinyin set. An example is as follows:

[0052] Text paragraph: My surname is Zhai

[0053] Pinyin text segment 1: wo xing zhai

[0054] Pinyin text segment 2: wo xing di

[0055] Among them, the text-to-pinyin model is based on a modern Chinese dictionary, and a data dictionary has been established for each Chinese character and its pinyin.

[0056] Steps 3-4: Input the dynamic password speech recognition text into the text-to-pinyin model to obtain the dynamic password pinyin text.

[0057] Step 3-5: Perform a consistency check between the dynamic password pinyin text and the speech recognition pinyin set. If any text segment in the speech recognition pinyin set is inconsistent with the dynamic password pinyin text, the dynamic verification is deemed to have failed, and the process returns to step 3-1. If any text segment in the speech recognition pinyin set is consistent with the dynamic password pinyin text, the process proceeds to step 3-6.

[0058] Steps 3-6: Input the dynamic password voice of the person to be verified into the voiceprint recognition model to obtain the second voiceprint discriminative feature vector L. y The second voiceprint discriminative feature vector L y N discriminative feature vectors {L1, L2, ..., L...} established with the voiceprint database N By comparing these values, we obtain the second cosine distance matrix {S1, S2, ..., S}. N}, and proceed to steps 3-7.

[0059] Steps 3-7: Transform the second cosine distance matrix {S1, S2, ..., S...} N The minimum value in} is used as the second similarity score S. m Second similarity score S m The corresponding voiceprint discriminative feature vector is L m If the second similarity score S m If the value is less than the first preset threshold D, then the person to be verified is determined to be a voiceprint discriminative feature vector L. m The system identifies the user's identity and compares it with the voiceprint recognition result from the voiceprint authentication stage. If the identities match, the system determines that the dynamic password voice is not deceptive and the user's identity is verified, allowing them to use the dispatch call service. Otherwise, the system determines that the dynamic password voice is deceptive and the identity verification fails.

[0060] By introducing dynamic password verification, fake dispatch voice messages were eliminated, the true identities of on-site personnel were verified, and the problem of deceptive voice messages was solved.

[0061] Implementation of Column 2

[0062] This embodiment provides a power grid dispatching identity authentication system based on voiceprint recognition, including:

[0063] Business application module: Acquire the voice data of the person to be verified;

[0064] Voiceprint authentication module: The acquired scheduling voice data is input into the trained voiceprint recognition model to extract the first voiceprint discriminative feature vector, and the extracted first voiceprint discriminative feature vector is matched with the voiceprint discriminative feature vector in the voiceprint database for voiceprint authentication.

[0065] Dynamic verification module: If the match is successful, a random password text is generated, and the dynamic password voice response from the person to be verified is obtained. The obtained dynamic password voice is input into the speech recognition model and converted into speech text information. The speech text information is dynamically verified and matched with the random password text. If they match, the identity of the person to be verified is successfully authenticated.

[0066] like Figure 4 As shown in this embodiment, a power grid dispatch identity authentication system based on voiceprint recognition includes a business application module, an interface module, an authentication module, and a data module.

[0067] The business application module is mainly used for interaction between users and the system. It directly provides identity authentication services to users. After the system verifies the identity of the collected voice data and verifies the instructions, the module will provide feedback to the user on the verification result and provide dispatch call services. If the verification fails, access will be denied.

[0068] The interface module provides standard interfaces for business application modules and other third-party systems to perform voiceprint registration, voiceprint authentication, dynamic verification, and voiceprint update functions.

[0069] The authentication module includes a voiceprint authentication module and a dynamic authentication module, which provide the core voiceprint processing functions of the system, and realize speech preprocessing, voiceprint recognition, speech recognition, and text-to-pinyin conversion functions.

[0070] The data module includes a voiceprint database, a random text library, and a text pinyin library, which are used to store voiceprint features and text pinyin, as well as generate dynamic passwords, providing data support for the authentication module.

[0071] Example 3

[0072] The purpose of this embodiment is to provide a computing device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the above-described method.

[0073] Example 4

[0074] The purpose of this embodiment is to provide a computer-readable storage medium.

[0075] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the steps of the above method.

[0076] The steps and methods involved in the apparatuses of Embodiments 2, 3, and 4 above correspond to those in Embodiment 1. For specific implementation details, please refer to the relevant description section of Embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media including one or more instruction sets; it should also be understood as including any medium capable of storing, encoding, or carrying an instruction set for execution by a processor and enabling the processor to perform any of the methods in this invention.

[0077] Those skilled in the art will understand that the modules or steps of the present invention described above can be implemented using general-purpose computer devices. Optionally, they can be implemented using computer-executable program code, thereby allowing them to be stored in a storage device for execution by a computer device, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. The present invention is not limited to any particular combination of hardware and software.

[0078] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A power grid dispatching authentication method based on voiceprint recognition, characterized in that, include: Obtain the dispatch voice data of the personnel to be verified; The acquired scheduling voice data is input into the trained voiceprint recognition model to extract the first voiceprint discriminative feature vector. The extracted first voiceprint discriminative feature vector is then matched with the voiceprint discriminative feature vector in the voiceprint database for voiceprint authentication. If the match is successful, a random password text is generated, and the dynamic password voice response from the person to be verified is obtained. The obtained dynamic password voice is input into the speech recognition model and converted into speech text information. The speech text information is dynamically verified and matched with the random password text. If they match, the identity of the person to be verified is successfully authenticated. The extracted first voiceprint discriminative feature vector is matched with the voiceprint discriminative feature vectors in the voiceprint database for voiceprint authentication, specifically including: The cosine distance between the first voiceprint discriminative feature vector and each voiceprint discriminative feature vector in the voiceprint database is calculated to obtain the cosine distance matrix. The minimum value in the cosine distance matrix is ​​used as the first similarity score; If the obtained first similarity score is less than the first preset threshold, then the person to be verified matches the voiceprint discriminative feature vector corresponding to the first similarity score, and the voiceprint authentication is successful. It also includes: if the first similarity score is not less than the first preset threshold and less than the second preset threshold, then a second verification is performed on the first voiceprint discriminative feature vector, specifically: If the sum of the absolute values ​​of the differences between the first similarity score and the cosine distances corresponding to each voiceprint discriminative feature vector in the voiceprint database, and the ratio of the sum of the absolute values ​​...

2. The power grid dispatching authentication method based on voiceprint recognition as described in claim 1, characterized in that, The voiceprint database includes the voiceprint discriminative feature vectors of registered dispatchers and their corresponding unique codes.

3. The power grid dispatching authentication method based on voiceprint recognition as described in claim 1, characterized in that, A random password text is generated, and the dynamic password voice response from the person to be verified is obtained. The obtained dynamic password voice is input into a speech recognition model and converted into speech-to-text information. The speech-to-text information is then dynamically verified and matched with the random password text. Specifically: Dynamic voice commands are input into a speech recognition model and converted into speech-text information; The acquired speech and text information is input into the text-to-speech model to obtain the pinyin text; The consistency between the pinyin text and the pinyin text segments in the speech recognition pinyin set is judged; the speech recognition pinyin set is formed by exhaustively enumerating any combination of pinyin for each character in the random password text based on the polyphony of each character in the random password text.

4. The power grid dispatching authentication method based on voiceprint recognition as described in claim 3, characterized in that, The consistency judgment between the pinyin text and the pinyin text segment in the speech recognition pinyin set is performed. Specifically, the consistency judgment between the pinyin text and the pinyin text segment in the speech recognition pinyin set is performed. If they are consistent, the obtained dynamic password voice is input into the voiceprint recognition model to obtain the second voiceprint discriminative feature vector. The cosine distance between the second voiceprint discriminative feature vector and each voiceprint discriminative feature vector in the voiceprint database is calculated to obtain the second cosine distance matrix. The minimum value in the second cosine distance matrix is ​​used as the second similarity score; If the second similarity score is less than the first preset threshold, the person to be verified is matched with the voiceprint discriminative feature vector corresponding to the second similarity score, and the matched voiceprint discriminative feature vector is compared with the voiceprint discriminative feature vector matched with the first voiceprint discriminative feature vector. If they match, the identity of the person to be verified is verified.

5. A power grid dispatching authentication system based on voiceprint recognition, characterized in that, include: Business application module: Acquire the voice data of the person to be verified; Voiceprint authentication module: The acquired scheduling voice data is input into the trained voiceprint recognition model to extract the first voiceprint discriminative feature vector. The extracted first voiceprint discriminative feature vector is then matched with voiceprint discriminative feature vectors in the voiceprint database for voiceprint authentication. Specifically, this includes: The cosine distance between the first voiceprint discriminative feature vector and each voiceprint discriminative feature vector in the voiceprint database is calculated to obtain the cosine distance matrix. The minimum value in the cosine distance matrix is ​​used as the first similarity score; If the obtained first similarity score is less than the first preset threshold, then the person to be verified matches the voiceprint discriminative feature vector corresponding to the first similarity score, and the voiceprint authentication is successful. If the first similarity score is not less than the first preset threshold and less than the second preset threshold, then the first voiceprint discriminative feature vector is subjected to secondary verification, specifically as follows: If the sum of the absolute values ​​of the differences between the first similarity score and the cosine distances corresponding to each voiceprint discriminative feature vector in the voiceprint database, and the ratio of the sum of the absolute values ​​... Dynamic verification module: If the match is successful, a random password text is generated, and the dynamic password voice response from the person to be verified is obtained. The obtained dynamic password voice is input into the speech recognition model and converted into speech text information. The speech text information is dynamically verified and matched with the random password text. If they match, the identity of the person to be verified is successfully authenticated.

6. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps in the power grid dispatch authentication method based on voiceprint recognition as described in any one of claims 1-4.

7. A processing apparatus, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps in the power grid dispatching authentication method based on voiceprint recognition as described in any one of claims 1-4.

Images