Voiceprint recognition privacy protection method and system based on telephone channel

Abstract

The invention relates to the technical field of telephone communication, in particular to a voiceprint recognition privacy protection method and system based on a telephone channel, and the method comprises the steps: a transmitting end inputs a received voice signal into a pre-trained voiceprint feature extraction model, and obtains a voiceprint feature vector corresponding to the voice signal; the sending end encrypts the voiceprint feature vector according to a preset encryption algorithm to obtain corresponding ciphertext information, and sends the ciphertext information to a receiving end through a telephone channel, so that the receiving end performs decryption through a decryption key based on the received ciphertext information, and sends the decrypted ciphertext information to the receiving end; and outputting an identity judgment result corresponding to the voice signal through the voiceprint recognition model. The method has the advantages that the encryption technology is adopted, voiceprint feature safety is comprehensively protected, voiceprint information leakage is effectively prevented, and privacy safety is comprehensively protected.

Description

Technical Field

[0001] The present invention relates to the technical field of telephone communication, and in particular to a voiceprint recognition privacy protection method and system based on telephone channels. Background Art

[0002] With the continuous evolution of communication technology, the application scope and influence of telephone communication continue to expand, playing an increasingly critical role in people's daily lives. Voiceprint recognition technology, with its unique biometric advantages, shows great application prospects in identity authentication and security protection in the field of telephone communication. However, in the process of implementing voiceprint recognition through telephone channels, there is a problem that cannot be ignored: the user's voiceprint information is very easy to be leaked. This situation seriously threatens the privacy and security of users and poses a major challenge to the protection of personal information.

[0003] At present, the privacy protection measures for voiceprint recognition have obvious deficiencies. Technically, although simple encrypted storage can prevent information from being easily interpreted, the encryption algorithm has the risk of being cracked, and the key management is complex. Differential privacy introduces noise to protect privacy, which will reduce the quality and availability of information. In terms of management, the existing minimum necessary principle is vaguely defined and usually collects too much data.

[0004] In summary, the protection effect of the above methods is not satisfactory and cannot effectively prevent potential privacy leakage risks. The use of user voiceprint information is like being in the "center of the storm" of information security, and the implementation process of protection measures is extremely complicated. It requires large-scale and high-cost upgrades to the existing telephone communication system, which not only increases the difficulty of implementation, but also greatly limits the popularity of these methods in practical applications, making them difficult to be effectively applied in a wide range of telephone communication environments. Summary of the invention

[0005] 1. Technical issues to be resolved

[0006] In view of the above-mentioned shortcomings and deficiencies of the prior art, the present invention provides a voiceprint recognition privacy protection method and system based on telephone channels, which solves the technical problem that the user's voiceprint information is easily leaked during telephone communication, seriously threatening the user's privacy security.

[0007] (II) Technical solution

[0008] In order to achieve the above object, the main technical solutions adopted by the present invention include:

[0009] In a first aspect, an embodiment of the present invention provides a voiceprint recognition privacy protection method based on a telephone channel, comprising:

[0010] S11, the sending end inputs the received voice signal into a pre-trained voiceprint feature extraction model to obtain a voiceprint feature vector corresponding to the voice signal;

[0011] The voiceprint feature extraction model is used to extract voiceprint features from the received speech signal;

[0012] S12, the sending end encrypts the voiceprint feature vector according to a preset encryption algorithm to obtain corresponding ciphertext information, and sends the ciphertext information to the receiving end through the telephone channel, so that the receiving end decrypts the received ciphertext information using a decryption key, and then outputs an identity judgment result corresponding to the voice signal through the voiceprint recognition model;

[0013] The decryption key is a key corresponding to the encryption algorithm.

[0014] Optionally, the S11 includes:

[0015] The sending end pre-processes the received voice signal;

[0016] The sending end inputs the preprocessed speech signal into the pre-trained voiceprint feature extraction model to obtain the voiceprint feature vector corresponding to the speech signal;

[0017] The voiceprint feature extraction model comprises: a feature extraction model obtained by training a convolutional neural network using a first training set;

[0018] The first training set includes: at least one speech signal and a voiceprint feature vector corresponding to each speech signal;

[0019] The voiceprint feature vector is a comprehensive feature that integrates the time domain features and frequency domain features of the speech signal.

[0020] Optionally, the S11 includes:

[0021] The sending end obtains the corresponding process feature vector according to the received voice signal and the preset voice processing algorithm;

[0022] The process feature vector includes a speech formant vector and a filter bank feature vector;

[0023] The sending end obtains the voiceprint feature vector corresponding to the speech signal according to the process feature vector and a pre-set residual convolution network.

[0024] Optionally, the encryption algorithm includes a symmetric encryption algorithm or an asymmetric encryption algorithm;

[0025] Then the S12 includes:

[0026] When a symmetric encryption algorithm is used to encrypt a voiceprint feature vector, the sender encrypts the extracted voiceprint feature vector using a pre-set shared key to obtain ciphertext information corresponding to the voiceprint feature vector;

[0027] The sending end sends the ciphertext information to the receiving end through the telephone channel, so that the receiving end decrypts the ciphertext information according to the preset shared key, and then outputs the identity judgment result corresponding to the voice signal through the voiceprint recognition model;

[0028] And, when an asymmetric encryption algorithm is used to encrypt the voiceprint feature vector, the sending end encrypts the extracted voiceprint feature vector by using the preset receiving end public key to obtain the ciphertext information corresponding to the voiceprint feature vector;

[0029] The sending end sends the ciphertext information to the receiving end through the telephone channel, so that the receiving end decrypts the ciphertext information according to the preset receiving end private key to obtain the decrypted voiceprint feature vector;

[0030] The shared key is a key pre-generated and shared by the sending end and the receiving end in a secure environment;

[0031] The receiving end public key is a key that the receiving end sends to the sending end in advance in a secure environment;

[0032] The decryption key includes a shared key or a receiving end private key;

[0033] The receiving end private key is a key corresponding to the receiving end public key.

[0034] Optionally, the S11 further includes:

[0035] The voiceprint feature vector is saved in a pre-set voiceprint feature information database, and the saving time of each voiceprint feature vector is recorded;

[0036] The voiceprint feature information library includes at least one voiceprint feature vector;

[0037] The saving time is the time when the voiceprint feature vector is saved in the voiceprint feature information database;

[0038] Then, the method further comprises:

[0039] The sending end determines in real time whether the storage time corresponding to each voiceprint feature vector in the voiceprint feature information database exceeds a preset time threshold;

[0040] If it exceeds, the voiceprint feature vector and the storage time corresponding to the voiceprint feature vector will be deleted.

[0041] Optionally, the S12 includes:

[0042] The sending end encrypts the voiceprint feature vector according to a preset encryption algorithm to obtain preliminary process information;

[0043] The sending end performs anonymization processing on the preliminary process information according to a preset hash algorithm to obtain corresponding ciphertext information;

[0044] The sending end sends the ciphertext information to the receiving end through the telephone channel, so that the receiving end decrypts the ciphertext information based on the received ciphertext information using the decryption key, and then outputs the identity judgment result corresponding to the voice signal through the voiceprint recognition model.

[0045] Optionally, the S12 further includes:

[0046] When the sending end obtains the corresponding ciphertext information, the sending end sends a prompt message to prompt the user whether to authorize voiceprint recognition;

[0047] The sending end receives the authorization result sent by the user in real time within a preset time, indicating whether to perform voiceprint recognition authorization;

[0048] When the user authorizes voiceprint recognition, the sending end sends the ciphertext information to the receiving end through the telephone channel.

[0049] In a second aspect, an embodiment of the present invention provides a voiceprint recognition privacy protection system based on a telephone channel, comprising: a transmitting end and a receiving end;

[0050] The transmitting end is used to input the received voice signal into a pre-trained voiceprint feature extraction model to obtain a voiceprint feature vector corresponding to the voice signal;

[0051] And, encrypting the voiceprint feature vector according to a preset encryption algorithm to obtain corresponding ciphertext information, and sending the ciphertext information to a receiving end through a telephone channel;

[0052] The receiving end is used to decrypt the corresponding ciphertext information according to a preset decryption key when receiving the corresponding ciphertext information through the telephone channel, so as to obtain a decrypted voiceprint feature vector;

[0053] And, the decrypted voiceprint feature vector is input into a preset voiceprint recognition model to obtain an identity judgment result corresponding to the voice signal to verify whether the voice signal successfully matches the corresponding user identity.

[0054] Optionally, the voiceprint recognition model includes a GMM-UBM model obtained by training with the second training set;

[0055] Then, the receiving end inputs the decrypted voiceprint feature vector into a preset voiceprint recognition model to obtain the identity judgment result corresponding to the voice signal, including:

[0056] The receiving end inputs the decrypted voiceprint feature vector into the pre-set GMM-UBM model to obtain the identity judgment result corresponding to the voice signal;

[0057] The second training set includes at least one voiceprint feature vector and an identity determination result corresponding to each voiceprint feature vector.

[0058] Optionally, the receiving end inputs the decrypted voiceprint feature vector into a preset voiceprint recognition model to obtain an identity determination result corresponding to the voice signal, including:

[0059] Preprocessing the decrypted voiceprint feature vector; the preprocessing includes reconstruction and noise reduction;

[0060] The preprocessed voiceprint feature vector is used to distinguish people and segment the audio to obtain the corresponding voice feature vector;

[0061] Performing voiceprint analysis on the human voice feature vector to obtain human voiceprint parameters corresponding to the human voice feature vector;

[0062] Calculate the Euclidean distance between the human voiceprint parameters and the preset target voiceprint parameters;

[0063] According to the Euclidean distance between the human voiceprint parameters and the preset target voiceprint parameters, the identity judgment result corresponding to the voice signal is obtained to verify whether the voice signal successfully matches the corresponding user identity.

[0064] (III) Beneficial effects

[0065] The beneficial effects of the present invention are as follows: the voiceprint recognition privacy protection method based on telephone channels of the present invention adopts encryption technology to comprehensively protect the security of voiceprint features, effectively prevent the leakage of voiceprint information, and comprehensively protect privacy security. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1 A simplified flow chart of a voiceprint recognition privacy protection method based on a telephone channel provided by an embodiment of the present invention;

[0067] Figure 2 A flowchart of a voiceprint recognition privacy protection method based on a telephone channel provided by an embodiment of the present invention;

[0068] Figure 3 A schematic diagram of the structure of a voiceprint recognition privacy protection system based on a telephone channel provided by an embodiment of the present invention;

[0069] Figure 4 A flow chart of a voiceprint recognition privacy protection method based on a telephone channel provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0070] In order to better explain the present invention and facilitate understanding, the present invention is described in detail below through specific implementation modes in conjunction with the accompanying drawings.

[0071] The embodiment of the present invention proposes a privacy protection method for voiceprint recognition based on a telephone channel, which adopts encryption technology to fully protect the security of voiceprint features, effectively prevent the leakage of voiceprint information, and fully protect privacy security.

[0072] In order to better understand the above technical solution, exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided to enable a clearer and more thorough understanding of the present invention and to fully convey the scope of the present invention to those skilled in the art.

[0073] Example 1

[0074] This embodiment provides a privacy protection method for voiceprint recognition based on telephone channels. Figure 1 As shown, including:

[0075] The sending end inputs the received voice signal into the pre-trained voiceprint feature extraction model to obtain the voiceprint feature vector corresponding to the voice signal;

[0076] The voiceprint feature extraction model is used to extract voiceprint features from the received speech signal;

[0077] S12, the sending end encrypts the voiceprint feature vector according to a preset encryption algorithm to obtain corresponding ciphertext information, and sends the ciphertext information to the receiving end through the telephone channel, so that the receiving end decrypts the received ciphertext information using a decryption key, and then outputs an identity judgment result corresponding to the voice signal through the voiceprint recognition model;

[0078] The decryption key is a key corresponding to the encryption algorithm.

[0079] The present embodiment provides a privacy protection method for voiceprint recognition in a telephone channel, which adopts an encryption algorithm to fully protect the security of voiceprint features, effectively prevent the leakage of voiceprint information, and fully protect privacy security.

[0080] Example 2

[0081] This embodiment provides a privacy protection method for voiceprint recognition based on telephone channels. Figure 2 As shown, including:

[0082] S11, the sending end inputs the received voice signal into a pre-trained voiceprint feature extraction model to obtain a voiceprint feature vector corresponding to the voice signal;

[0083] The voiceprint feature extraction model is used to extract voiceprint features from a received speech signal.

[0084] Further,

[0085] The sending end pre-processes the received voice signal;

[0086] The sending end inputs the preprocessed speech signal into the pre-trained voiceprint feature extraction model to obtain the voiceprint feature vector corresponding to the speech signal;

[0087] The voiceprint feature extraction model comprises: a feature extraction model obtained by training a convolutional neural network using a first training set;

[0088] The first training set includes: at least one speech signal and a voiceprint feature vector corresponding to each speech signal;

[0089] The voiceprint feature vector is a comprehensive feature that integrates the time domain features and frequency domain features of the speech signal.

[0090] Furthermore, the transmitting end obtains a corresponding process feature vector according to the received speech signal and a preset speech processing algorithm;

[0091] The process feature vector includes a speech formant vector and a filter bank feature vector;

[0092] The sending end obtains the voiceprint feature vector corresponding to the speech signal according to the process feature vector and a pre-set residual convolution network.

[0093] S12, the sending end encrypts the voiceprint feature vector according to a preset encryption algorithm to obtain corresponding ciphertext information, and sends the ciphertext information to the receiving end through the telephone channel, so that the receiving end decrypts the received ciphertext information using a decryption key, and then outputs an identity judgment result corresponding to the voice signal through the voiceprint recognition model;

[0094] The decryption key is a key corresponding to the encryption algorithm.

[0095] Specifically, the sending end encrypts the voiceprint feature vector according to a preset encryption algorithm to obtain preliminary process information;

[0096] The sending end performs anonymization processing on the preliminary process information according to a preset hash algorithm to obtain corresponding ciphertext information;

[0097] The sending end sends the ciphertext information to the receiving end through the telephone channel, so that the receiving end decrypts the ciphertext information based on the received ciphertext information using the decryption key, and then outputs the identity judgment result corresponding to the voice signal through the voiceprint recognition model.

[0098] Furthermore, the encryption algorithm includes a symmetric encryption algorithm or an asymmetric encryption algorithm;

[0099] Then, when a symmetric encryption algorithm is used to encrypt the voiceprint feature vector, the sender encrypts the extracted voiceprint feature vector using a pre-set shared key to obtain the ciphertext information corresponding to the voiceprint feature vector;

[0100] The sending end sends the ciphertext information to the receiving end through the telephone channel, so that the receiving end decrypts the ciphertext information according to the preset shared key, and then outputs the identity judgment result corresponding to the voice signal through the voiceprint recognition model;

[0101] And, when an asymmetric encryption algorithm is used to encrypt the voiceprint feature vector, the sending end encrypts the extracted voiceprint feature vector by using the preset receiving end public key to obtain the ciphertext information corresponding to the voiceprint feature vector;

[0102] The sending end sends the ciphertext information to the receiving end through the telephone channel, so that the receiving end decrypts the ciphertext information according to the preset receiving end private key to obtain the decrypted voiceprint feature vector;

[0103] The shared key is a key pre-generated and shared by the sending end and the receiving end in a secure environment;

[0104] The receiving end public key is a key that the receiving end sends to the sending end in advance in a secure environment;

[0105] The decryption key includes a shared key or a receiving end private key;

[0106] The receiving end private key is a key corresponding to the receiving end public key.

[0107] Furthermore, the S11 further includes:

[0108] The voiceprint feature vector is saved in a pre-set voiceprint feature information database, and the saving time of each voiceprint feature vector is recorded;

[0109] The voiceprint feature information library includes at least one voiceprint feature vector;

[0110] The saving time is the time when the voiceprint feature vector is saved in the voiceprint feature information database;

[0111] Then, the method further comprises:

[0112] The sending end determines in real time whether the storage time corresponding to each voiceprint feature vector in the voiceprint feature information database exceeds a preset time threshold;

[0113] If it exceeds, the voiceprint feature vector and the storage time corresponding to the voiceprint feature vector will be deleted.

[0114] Furthermore, the S12 further includes:

[0115] When the sending end obtains the corresponding ciphertext information, the sending end sends a prompt message to prompt the user whether to authorize voiceprint recognition;

[0116] The sending end receives the authorization result sent by the user in real time within a preset time, indicating whether to perform voiceprint recognition authorization;

[0117] When the user authorizes voiceprint recognition, the sending end sends the ciphertext information to the receiving end through the telephone channel.

[0118] This embodiment provides a privacy protection method for voiceprint recognition based on telephone channels, which uses encryption algorithms to fully protect the security of user voiceprint features. The encryption measures are like a solid line of defense, effectively preventing the leakage of voiceprint information and improving the level of privacy security. Further setting up a variety of privacy protection measures further strengthens the privacy protection effect of voiceprint recognition from different dimensions, and cooperates and synergizes with each other to form a strict privacy protection network to comprehensively resist possible privacy violations.

[0119] Example 3

[0120] This embodiment provides a voiceprint recognition privacy protection system based on telephone channels, and its structure is as follows: Figure 3 As shown, it includes: a sending end and a receiving end, the sending end is connected to the receiving end through telephone information communication;

[0121] The transmitting end is used to input the received voice signal into a pre-trained voiceprint feature extraction model to obtain a voiceprint feature vector corresponding to the voice signal;

[0122] And, encrypting the voiceprint feature vector according to a preset encryption algorithm to obtain corresponding ciphertext information, and sending the ciphertext information to a receiving end through a telephone channel;

[0123] The receiving end is used to decrypt the corresponding ciphertext information according to a preset decryption key when receiving the corresponding ciphertext information through the telephone channel, so as to obtain a decrypted voiceprint feature vector;

[0124] And, the decrypted voiceprint feature vector is input into a preset voiceprint recognition model to obtain an identity judgment result corresponding to the voice signal to verify whether the voice signal successfully matches the corresponding user identity.

[0125] Further, the voiceprint recognition model includes a GMM-UBM model obtained by training with the second training set;

[0126] The receiving end then inputs the decrypted voiceprint feature vector into a preset voiceprint recognition model to obtain the identity judgment result corresponding to the voice signal, including:

[0127] The receiving end inputs the decrypted voiceprint feature vector into the pre-set GMM-UBM model to obtain the identity judgment result corresponding to the voice signal;

[0128] The second training set includes at least one voiceprint feature vector and an identity determination result corresponding to each voiceprint feature vector.

[0129] Furthermore, the receiving end is used to pre-process the decrypted voiceprint feature vector; the pre-processing includes reconstruction and noise reduction;

[0130] The preprocessed voiceprint feature vector is used to distinguish people and segment the audio to obtain the corresponding voice feature vector;

[0131] Performing voiceprint analysis on the human voice feature vector to obtain human voiceprint parameters corresponding to the human voice feature vector;

[0132] Calculate the Euclidean distance between the human voiceprint parameters and the preset target voiceprint parameters;

[0133] According to the Euclidean distance between the human voiceprint parameters and the preset target voiceprint parameters, the identity judgment result corresponding to the voice signal is obtained to verify whether the voice signal successfully matches the corresponding user identity.

[0134] The present embodiment provides a telephone channel-based voiceprint recognition privacy protection system, which does not require large-scale, high-cost transformation of existing telephone communication modes, greatly reduces the implementation threshold. This simple and easy method provides more users with reliable voiceprint recognition privacy protection.

[0135] Example 4

[0136] This embodiment provides a privacy protection method for voiceprint recognition based on telephone channels. Figure 4 As shown, including:

[0137] Voiceprint feature extraction: At the sending end of the telephone channel, the pre-trained convolutional neural network CNN model (voiceprint feature extraction model) is used to accurately extract the local area features of the voice signal at different times and frequencies, and then perform real-time voiceprint feature extraction operations. That is, the voiceprint feature extraction model is used to extract the voiceprint feature vector from the complex voice signal. Through the trained voiceprint feature extraction model, the input voice signal is operated at various layers to extract a representative feature vector from the voice signal. The voiceprint feature vector integrates various deep-level features of the voice signal, can well reflect the voiceprint characteristics, and lays the foundation for subsequent privacy protection and voiceprint recognition.

[0138] Encryption processing: A highly secure encryption processing method is used for the extracted voiceprint feature vector. Symmetric encryption algorithm or asymmetric encryption algorithm can be selected to convert the voiceprint feature vector into ciphertext information. This encryption conversion makes it difficult for the voiceprint feature vector to be stolen and cracked during the transmission process of the telephone channel.

[0139] Transmission: The encrypted ciphertext information is securely transmitted through the telephone channel. In this process, the encrypted voiceprint feature vector is transmitted in the channel in a ciphertext state, effectively avoiding the risk of information leakage during the transmission link.

[0140] Decryption processing: At the receiving end of the telephone channel, a special decryption mechanism is configured to accurately decrypt the received ciphertext information. Through this link, the ciphertext information can be restored to the original voiceprint feature vector, preparing for subsequent voiceprint recognition.

[0141] Voiceprint recognition: Voiceprint recognition is performed on the successfully decrypted voiceprint feature vector with the help of Gaussian mixture model-universal background model (GMM-UBM model) to accurately determine the user's identity and achieve safe and efficient identity authentication.

[0142] Privacy protection strategy: Set up a variety of privacy protection strategies to fully protect the privacy security of users:

[0143] Limit the storage time of voiceprint feature vectors: Strictly stipulate the storage time of voiceprint feature vectors. Once a specific time is exceeded, the system will automatically delete the stored voiceprint feature vectors. This strategy can effectively avoid the risk of privacy leakage caused by long-term retention of voiceprint information, just like setting a "regular cleanup" mechanism for user privacy.

[0144] Anonymization of voiceprint features: The voiceprint feature vector is anonymized and processed through complex algorithms so that the voiceprint feature vector cannot be directly associated with the user's identity. This measure greatly enhances the confidentiality of the user's identity when ensuring voiceprint recognition.

[0145] User authorization mechanism: Voiceprint recognition is only allowed with explicit authorization from the user. Users can grant recognition permissions through specific instructions or operations, thus ensuring that the voiceprint recognition process is completely under the control of the user and preventing unauthorized recognition from infringing on user privacy.

[0146] For example: at the sending end, a symmetric encryption algorithm is used, such as the Advanced Encryption Standard (AES) algorithm. A key, namely the shared key, is generated and shared in a secure environment in advance. When the user conducts voice communication and requires voiceprint recognition, the voiceprint feature vector is extracted and encrypted using the key to obtain the corresponding ciphertext information. At the receiving end, after receiving the ciphertext information, the shared key is also used for decryption, and the original voiceprint feature vector is obtained, and then the voiceprint recognition operation is performed. At the same time, during the entire process, the storage time of the voiceprint feature vector is set to 5 days. After this time, the system automatically deletes the stored voiceprint feature vector; and the voiceprint features are simply anonymized during the storage process, such as masking some non-critical data in the feature vector, and then restoring it at the receiving end; the user can authorize the voiceprint recognition operation by entering a specific password at the sending end.

[0147] For example: At the sending end, an asymmetric encryption algorithm, such as the RSA algorithm, is used. The extracted voiceprint feature vector is encrypted using the receiving end public key. After receiving the ciphertext information, the receiving end uses its own receiving end private key to decrypt it. For the privacy protection policy, the storage time of the voiceprint feature vector is set to 72 hours, and it is automatically deleted after this time. The anonymization of the voiceprint feature vector is achieved by hashing the feature vector, and the original voiceprint feature vector is restored at the receiving end through decryption and a specific hash inverse operation. User authorization is completed by pressing a specific key combination on the phone to ensure that voiceprint recognition can only be performed with user authorization.

[0148] In summary, the voiceprint recognition privacy protection method based on the telephone channel provided in this embodiment provides users with an efficient and secure voiceprint recognition privacy protection solution without affecting the normal operation of telephone communication.

[0149] In the description of the present invention, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0150] In the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", "connected", "fixed" and the like should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium; it can be the internal connection of two elements or the interaction relationship between two elements. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0151] In the present invention, unless otherwise clearly specified and limited, when a first feature is “on” or “below” a second feature, it may be that the first and second features are in direct contact, or that the first and second features are in indirect contact through an intermediate medium. Moreover, when a first feature is “above”, “above” or “above” a second feature, it may be that the first feature is directly above or obliquely above the second feature, or it may simply mean that the first feature is higher in level than the second feature. When a first feature is “below”, “below” or “below” a second feature, it may be that the first feature is directly below or obliquely below the second feature, or it may simply mean that the first feature is lower in level than the second feature.

[0152] In the description of this specification, the description of the terms "one embodiment", "some embodiments", "embodiment", "example", "specific example" or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, unless they are contradictory.

[0153] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may alter, modify, replace and modify the above embodiments within the scope of the present invention.

Claims

1. A privacy protection method for voiceprint recognition based on telephone channels, characterized in that: include: S11, the sending end inputs the received voice signal into a pre-trained voiceprint feature extraction model to obtain a voiceprint feature vector corresponding to the voice signal; The voiceprint feature extraction model is used to extract voiceprint features from the received speech signal; S12, the sending end encrypts the voiceprint feature vector according to a preset encryption algorithm to obtain corresponding ciphertext information, and sends the ciphertext information to the receiving end through the telephone channel, so that the receiving end decrypts the received ciphertext information using a decryption key, and then outputs an identity judgment result corresponding to the voice signal through the voiceprint recognition model; The decryption key is a key corresponding to the encryption algorithm.

2. The privacy protection method for voiceprint recognition based on telephone channel according to claim 1 is characterized in that: The S11 includes: The sending end pre-processes the received voice signal; The sending end inputs the preprocessed speech signal into the pre-trained voiceprint feature extraction model to obtain the voiceprint feature vector corresponding to the speech signal; The voiceprint feature extraction model comprises: a feature extraction model obtained by training a convolutional neural network using a first training set; The first training set includes: at least one speech signal and a voiceprint feature vector corresponding to each speech signal; The voiceprint feature vector is a comprehensive feature that integrates the time domain features and frequency domain features of the speech signal.

3. The privacy protection method for voiceprint recognition based on telephone channel according to claim 1 is characterized in that: The S11 includes: The sending end obtains the corresponding process feature vector according to the received voice signal and the preset voice processing algorithm; The process feature vector includes a speech formant vector and a filter bank feature vector; The sending end obtains the voiceprint feature vector corresponding to the speech signal according to the process feature vector and a pre-set residual convolution network.

4. The privacy protection method for voiceprint recognition based on telephone channel according to claim 1 is characterized in that: The encryption algorithm includes a symmetric encryption algorithm or an asymmetric encryption algorithm; Then the S12 includes: When a symmetric encryption algorithm is used to encrypt a voiceprint feature vector, the sender encrypts the extracted voiceprint feature vector using a pre-set shared key to obtain ciphertext information corresponding to the voiceprint feature vector; The sending end sends the ciphertext information to the receiving end through the telephone channel, so that the receiving end decrypts the ciphertext information according to the preset shared key, and then outputs the identity judgment result corresponding to the voice signal through the voiceprint recognition model; And, when an asymmetric encryption algorithm is used to encrypt the voiceprint feature vector, the sending end encrypts the extracted voiceprint feature vector by using the preset receiving end public key to obtain the ciphertext information corresponding to the voiceprint feature vector; The sending end sends the ciphertext information to the receiving end through the telephone channel, so that the receiving end decrypts the ciphertext information according to the preset receiving end private key to obtain the decrypted voiceprint feature vector; The shared key is a key pre-generated and shared by the sending end and the receiving end in a secure environment; The receiving end public key is a key that the receiving end sends to the sending end in advance in a secure environment; The decryption key includes a shared key or a receiving end private key; The receiving end private key is a key corresponding to the receiving end public key.

5. The privacy protection method for voiceprint recognition based on telephone channel according to claim 1 is characterized in that: The S11 further includes: The voiceprint feature vector is saved in a pre-set voiceprint feature information database, and the saving time of each voiceprint feature vector is recorded; The voiceprint feature information library includes at least one voiceprint feature vector; The saving time is the time when the voiceprint feature vector is saved in the voiceprint feature information database; Then, the method further comprises: The sending end determines in real time whether the storage time corresponding to each voiceprint feature vector in the voiceprint feature information database exceeds a preset time threshold; If it exceeds, the voiceprint feature vector and the storage time corresponding to the voiceprint feature vector will be deleted.

6. The privacy protection method for voiceprint recognition based on telephone channel according to claim 1 is characterized in that: The S12 includes: The sending end encrypts the voiceprint feature vector according to a preset encryption algorithm to obtain preliminary process information; The sending end performs anonymization processing on the preliminary process information according to a preset hash algorithm to obtain corresponding ciphertext information; The sending end sends the ciphertext information to the receiving end through the telephone channel, so that the receiving end decrypts the ciphertext information based on the received ciphertext information using the decryption key, and then outputs the identity judgment result corresponding to the voice signal through the voiceprint recognition model.

7. The privacy protection method for voiceprint recognition based on telephone channel according to claim 1 is characterized in that: The S12 further includes: When the sending end obtains the corresponding ciphertext information, the sending end sends a prompt message to prompt the user whether to authorize voiceprint recognition; The sending end receives the authorization result sent by the user in real time within a preset time, indicating whether to perform voiceprint recognition authorization; When the user authorizes voiceprint recognition, the sending end sends the ciphertext information to the receiving end through the telephone channel.

8. A voiceprint recognition privacy protection system based on telephone channels, characterized in that: Includes: a sending end and a receiving end; The transmitting end is used to input the received voice signal into a pre-trained voiceprint feature extraction model to obtain a voiceprint feature vector corresponding to the voice signal; And, encrypting the voiceprint feature vector according to a preset encryption algorithm to obtain corresponding ciphertext information, and sending the ciphertext information to a receiving end through a telephone channel; The receiving end is used to decrypt the corresponding ciphertext information according to a preset decryption key when receiving the corresponding ciphertext information through the telephone channel, so as to obtain a decrypted voiceprint feature vector; And, the decrypted voiceprint feature vector is input into a preset voiceprint recognition model to obtain an identity judgment result corresponding to the voice signal to verify whether the voice signal successfully matches the corresponding user identity.

9. The voiceprint recognition privacy protection system based on telephone channel according to claim 8, characterized in that: The voiceprint recognition model includes a GMM-UBM model obtained by training with the second training set; Then, the receiving end inputs the decrypted voiceprint feature vector into a preset voiceprint recognition model to obtain the identity judgment result corresponding to the voice signal, including: The receiving end inputs the decrypted voiceprint feature vector into the pre-set GMM-UBM model to obtain the identity judgment result corresponding to the voice signal; The second training set includes at least one voiceprint feature vector and an identity determination result corresponding to each voiceprint feature vector.

10. The voiceprint recognition privacy protection system based on telephone channel according to claim 8, characterized in that: The receiving end inputs the decrypted voiceprint feature vector into a preset voiceprint recognition model to obtain an identity judgment result corresponding to the voice signal, including: Preprocessing the decrypted voiceprint feature vector; the preprocessing includes reconstruction and noise reduction; The preprocessed voiceprint feature vector is used to distinguish people and segment the audio to obtain the corresponding voice feature vector; Performing voiceprint analysis on the human voice feature vector to obtain human voiceprint parameters corresponding to the human voice feature vector; Calculate the Euclidean distance between the human voiceprint parameters and the preset target voiceprint parameters; According to the Euclidean distance between the human voiceprint parameters and the preset target voiceprint parameters, the identity judgment result corresponding to the voice signal is obtained to verify whether the voice signal successfully matches the corresponding user identity.

Images