Voice communication privacy protection method and system based on voice confusion, and medium

Abstract

The invention relates to a voice communication privacy protection method and system based on voice confusion, and a medium. The method comprises the following steps: receiving first voice confusion data sent by a first user to a second user; acquiring a first random number corresponding to the first user and a first identity of the first user; searching target voices in the voice data set according to the first random number and the first identity, and splicing the target voices to generate voice confusion noise; according to the voice confusion noise, performing de-confusion on the first voice confusion data, and performing reduction to obtain first voice data; even if the data transmission is damaged, the voice can be restored based on the de-confusion, so that the receiver can successfully decrypt the voice data.

Description

Technical Field

[0001] The present application relates to the field of communication security technology, and in particular to a voice communication privacy protection method, system and medium based on voice obfuscation. Background Art

[0002] Voice communication technologies, including traditional telephones, VoIP (Voice over Internet Protocol), and voice messaging, have developed rapidly in recent years. As these technologies become more and more important in people's lives and appear more and more frequently, they have also raised a large number of privacy issues. During voice communication, the transmission of voice data on different links may be accessed and eavesdropped by unauthorized attackers. Since voice data itself contains a wealth of user personal privacy information, the leakage of voice data may lead to unauthorized surveillance of users, leakage of user privacy, and even identity theft.

[0003] In order to reduce the above-mentioned privacy leakage problem, related technologies have proposed some end-to-end encryption transmission solutions. For example, the communication service provider integrates the end-to-end encryption method in the network protocol layer of voice transmission. For another example, the user encrypts the voice data with a key at the sending end and then transmits it. However, this encrypted transmission method is prone to data loss during transmission, making it difficult for the receiver to decrypt.

[0004] Currently, no effective solution has been proposed for the problem in the related art that the receiver cannot decrypt the voice data. Summary of the invention

[0005] Based on this, it is necessary to provide a voice communication privacy protection method, system and medium based on voice obfuscation that can ensure that the recipient successfully decrypts the voice data in response to the above technical problems.

[0006] In a first aspect, the present application provides a voice communication privacy protection method based on voice confusion. The method comprises: Receiving first speech confusion data sent by a first user to a second user; Obtaining a first random number corresponding to the first user and a first identity of the first user; According to the first random number and the first identity, searching for a target voice in a voice data set, and concatenating the target voice to generate voice confusion noise; The first speech confusion data is de-obfuscated according to the speech confusion noise to restore the first speech data.

[0007] In one of the embodiments, before obtaining a first random number corresponding to the first user and a first identity of the first user, the method further includes: receiving a first certificate sent by the first user, and returning a second certificate of the second user; receiving first encrypted data sent by the first user, and decrypting the first encrypted data using the private key of the second user to obtain the first random number and the first identity; A second random number is generated, and the second random number and the second identity of the second user are encrypting by using the public key of the first user, and second encrypted data is generated and returned.

[0008] In one embodiment, de-obfuscating the first speech obfuscated data according to the speech obfuscation noise to restore the first speech data includes: Dividing and aligning the first speech confusion data and the speech confusion noise according to each signal sampling point to obtain a plurality of groups of corresponding first speech confusion data blocks and speech confusion noise blocks; Setting a parameter set, wherein the parameter set includes a plurality of one-dimensional optimization parameters related to the signal frequency; An objective function is set based on the one-dimensional optimization parameter, the parameter set is traversed, the minimum value of the objective function is obtained, and the one-dimensional optimization parameter corresponding to the minimum value is used as the target parameter; wherein the objective function is used to measure the degree of noise residual in the original signal after applying the one-dimensional optimization parameter; Based on the target parameter, calculating an improvement factor corresponding to each frequency sampling point; wherein the improvement factor is used to measure the noise suppression effect after optimization; The time-frequency spectra of each of the first speech confusion data blocks are weighted using the improvement factor as a weight to obtain the time-frequency spectra of the first speech data after deobfuscation.

[0009] In one embodiment, setting an objective function based on the one-dimensional optimization parameter, traversing the parameter set, and finding the minimum value of the objective function includes: For each frequency sampling point, scaling the speech confusion noise block based on the one-dimensional optimization parameter to obtain a first value; Compare the first value with the corresponding first speech confusion data block by taking absolute values ​​to obtain a ratio, and subtract the ratio from 1 to obtain a second value; The second value is used as a weight to weight each of the first speech confusion data blocks to obtain the value of the objective function.

[0010] In one embodiment, based on the target parameter, calculating the improvement factor corresponding to each frequency sampling point includes: For each frequency sampling point, scaling the speech confusion noise block based on the target parameter to obtain a first value; Compare the first value with the corresponding first speech confusion data block by taking absolute values ​​to obtain a ratio, and subtract the ratio from 1 to obtain a second value; The second value is compared with zero, and the larger value is taken as the value of the improvement factor.

[0011] In one of the embodiments, after calculating the improvement factor corresponding to each frequency sampling point based on the target parameter, the method further includes: Acquire a channel loss rate of a first link, and compare the improvement factor with the channel loss rate; the first link is a communication link from the first user to the second user; If the improvement factor is less than or equal to the channel loss rate, the improvement factor is set to zero.

[0012] In one embodiment, before de-obfuscating the first speech confusion data according to the speech confusion noise to restore the first speech data, the method further includes: A frequency response of the first link is obtained, and the speech confusion noise is compensated by using the frequency response.

[0013] In one embodiment, before de-obfuscating the first speech confusion data according to the speech confusion noise to restore the first speech data, the method further includes: Receiving second voice data sent by the first user based on the first link, and acquiring prior data of the second voice data; According to the second voice data and the priori data, the channel state of the first link is estimated to obtain channel state parameters; wherein the channel state parameters include frequency response and / or channel loss rate.

[0014] In a second aspect, the present application provides a voice communication privacy protection method based on voice confusion. The method comprises: Receiving a voice input from a first user to obtain first voice data; Determining a first identity in the voice data set that matches the voiceprint information of the first user; Determine a first random number of the current session, and generate voice confusion noise according to the first random number, the first identity, and the voice data set; The speech confusion noise is superimposed on the first speech data to obtain first speech confusion data.

[0015] In one embodiment, generating speech confusion noise according to the first random number, the first identity and the speech data set includes: Generate an index sequence according to the first random number; Searching for a target speech corresponding to the first identity in the speech data set according to the index sequence; The target speech is concatenated to obtain the speech confusion noise.

[0016] In one of the embodiments, before determining the first random number of the current session, the method further includes: sending a first certificate of the first user to the second user; receiving a second certificate returned by the second user; Generate the first random number, encrypt the first random number and the first identity using the public key of the second user, and send the generated first encrypted data to the second user; Receive second encrypted data returned by the second user, and decrypt the second encrypted data according to the private key of the first user; wherein the second encrypted data is generated by encrypting a second random number and a second identity of the second user with the public key of the first user.

[0017] In one embodiment, after superimposing the speech confusion noise to the first speech data to obtain the first speech confusion data, the method further includes: sending the first speech confusion data to the second user; or, The first speech confusion data is sent to a first communication terminal corresponding to the first user, and then sent to the second user via the first communication terminal.

[0018] In a third aspect, the present application provides a speech processing system, comprising: a processor, a speaker and a microphone, wherein the speaker and the microphone are connected to the processor respectively; The processor is used to execute the voice communication privacy protection method based on voice confusion described in the first aspect or the second aspect above.

[0019] In a fourth aspect, the present application provides a voice communication system, including: a voice processing system and a communication terminal; wherein: The communication terminal is used to receive the voice confusion data and send the voice confusion data to the voice processing system, and the voice processing system is used to execute the voice communication privacy protection method based on voice confusion described in the first aspect above; or, The speech processing system is used to execute the voice communication privacy protection method based on speech confusion described in the second aspect above, and send the generated speech confusion data to the communication terminal.

[0020] In a fifth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the method described in the first aspect or the second aspect are implemented.

[0021] The above-mentioned voice communication privacy protection method, system and medium based on voice obfuscation, by obfuscating the original voice data, generates voice obfuscated data which does not need to maintain a high degree of data integrity during transmission, and even if the data transmission is damaged, the voice can be restored based on deobfuscation. Therefore, through the method of this embodiment, it can be guaranteed that the receiving party successfully decrypts the voice data and can work normally even when the channel is damaged. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 is a structural block diagram of a speech processing system in one embodiment; Figure 2 A flowchart of a voice communication privacy protection method based on voice obfuscation of a receiver in one embodiment; Figure 3 is a flow chart of data exchange between users in one embodiment; Figure 4 A flowchart of a speech deconfusion method in one embodiment; Figure 5 A flowchart of a voice communication privacy protection method based on voice obfuscation of a sender in one embodiment; Figure 6 A schematic diagram of an application environment of a voice communication system in an embodiment; Figure 7 for Figure 6 The operation principle diagram of the voice communication system in the Figure 8 is a diagram showing experimental results of user voice privacy protection capability in one embodiment; Fig. 9 FIG. 4 is a diagram showing experimental results of voice communication quality in an embodiment. DETAILED DESCRIPTION

[0023] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0024] Unless otherwise defined, the technical terms or scientific terms involved in this application shall have the general meaning understood by people with general skills in the technical field to which this application belongs. The words "one", "a", "the", "these" and the like in this application do not indicate a quantitative limitation, and they may be singular or plural. The terms "include", "comprise", "have" and any variants thereof involved in this application are intended to cover non-exclusive inclusions; for example, a process, method and system, product or device comprising a series of steps or modules (units) is not limited to the listed steps or modules (units), but may include unlisted steps or modules (units), or may include other steps or modules (units) inherent to these processes, methods, products or devices. The words "connect", "connected", "coupled" and the like involved in this application are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. The "multiple" involved in this application refers to two or more. "And / or" describes the association relationship of associated objects, indicating that there may be three relationships, for example, "A and / or B" may mean: A exists alone, A and B exist at the same time, and B exists alone. Generally, the character " / " indicates that the objects associated with each other are in an "or" relationship. The terms "first", "second", "third", etc. in this application are only used to distinguish similar objects and do not represent a specific ordering of the objects.

[0025] In one embodiment, Figure 1 A structural block diagram of a speech processing system is provided, such as Figure 1 As shown, the speech processing system includes: a processor 101, a speaker 102 and a microphone 103, and the speaker 102 and the microphone 103 are connected to the processor 101 respectively. The processor 101 can be one or more ( Figure 1 Only one is shown in the figure), the processor 101 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA.

[0026] Optionally, the above-mentioned voice processing system further includes a memory 104 for storing data and a transmission device 105 for communication functions. A person skilled in the art can understand that Figure 1 The structure shown is only for illustration and does not limit the structure of the above-mentioned speech processing system. Figure 1 More or fewer components as shown, or with Figure 1 Different configurations shown.

[0027] The memory 104 can be used to store computer programs, for example, software programs and modules of application software, such as the computer program corresponding to the voice communication privacy protection method based on voice confusion in this embodiment. The processor 101 executes various functional applications and data processing by running the computer program stored in the memory 104, that is, to implement the above method. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include a memory remotely arranged relative to the processor 101, and these remote memories may be connected to the terminal via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0028] The transmission device 105 is used to receive or send data via a network. The above network includes a wireless network provided by a communication provider. In one example, the transmission device 105 includes a network adapter (Network Interface Controller, referred to as NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 105 can be a radio frequency (Radio Frequency, referred to as RF) module, which is used to communicate with the Internet wirelessly.

[0029] When the user corresponding to the speech processing system is the sender, the processor 101 executes Figure 2 When the user corresponding to the speech processing system is the recipient, the processor 101 executes Figure 5 Of course, the speech processing system can be used as both a sender and a receiver, so the processor 101 also has the function of executing Figure 2 and Figure 5 The following embodiments will introduce the voice communication privacy protection method based on voice confusion of the present application from the perspectives of the receiver and the sender respectively.

[0030] In one embodiment, Figure 2 A flowchart of a method for protecting the privacy of voice communication based on voice obfuscation of a receiver is provided. Figure 1 Taking the speech processing system as an example, the process includes the following steps: Step S21, receiving first speech confusion data sent by a first user to a second user.

[0031] The first voice confusion data is generated by superimposing the voice confusion noise of the first user onto the original voice data. In this embodiment, the first voice confusion data can be directly received by the voice processing system on the second user side, or the first voice confusion data can be first received by the communication terminal on the second user side, and then sent to the voice processing system by the communication terminal, which is not limited in this embodiment.

[0032] Optionally, before step S21, the method further includes a system registration step: each speech processing system instance is registered with the same trusted third-party certification authority before leaving the factory, and obtains a speech processing system certificate Cert dn The certificate of the trusted third-party certification authority ca , and obtain a public voice data set D, such as the ai_datatang data set. After the user obtains the corresponding voice processing system, the user registration stage is entered, and a key pair can be generated using the RSA-1024 algorithm (p k ,s k ), and record a user voice signal of about 20 to 60 seconds in length to accurately extract the user's voiceprint information. The user's voiceprint information is extracted from the voice signal, and based on the voiceprint information, the cosine similarity is used to match the speaker identity s closest to the voiceprint information from the voice dataset D, which is the user identity.

[0033] Step S22: Acquire a first random number corresponding to the first user and a first identity of the first user.

[0034] The first random number and the first identity can be obtained through data exchange between users. For example, the first user directly sends the first random number and the first identity to the second user; or, after the first user verifies the identity of the second user, the first random number and the first identity are sent to the second user.

[0035] Step S23, searching for the target speech in the speech data set according to the first random number and the first identity, and concatenating the target speech to generate speech confusion noise.

[0036] An index sequence may be generated according to a first random number; a target speech corresponding to the first identity may be searched in a speech data set according to the index sequence; and the target speech may be concatenated to obtain speech confusion noise.

[0037] Step S24, de-obfuscating the first speech obfuscated data according to the speech obfuscation noise to restore the first speech data.

[0038] In this step, a target parameter that can minimize the ratio of speech confusion noise to the original signal can be found from multiple frequency-related one-dimensional optimization parameters, and then an improvement factor can be calculated using the target parameter. The first speech confusion data is processed based on the improvement factor to restore clear first speech data.

[0039] From an end-to-end perspective, the channel is lossy and will destroy the encrypted data. Traditional voice communication privacy protection methods have high requirements for the integrity of encrypted data. If some bits are missing during encrypted data transmission, it will be difficult for the recipient to decrypt. In the above steps S21 to S24, by obfuscating the original voice data, the generated voice obfuscated data does not need to maintain a high degree of data integrity during transmission. Even if the data transmission is lossy, the voice can be restored based on deobfuscation. Therefore, the method of this embodiment can ensure that the recipient successfully decrypts the voice data and can work normally even when the channel is lossy.

[0040] In one embodiment, Figure 3 A flow chart of data exchange between users is provided, such as Figure 3 As shown, before obtaining the first random number corresponding to the first user and the first identity of the first user in step S22, the method further includes the following steps: Step 1: receiving a first certificate sent by a first user, and returning a second certificate of a second user.

[0041] The two communicating parties use the voice communication channel to exchange data. The data modulation method in the data exchange process is frequency modulation, and the carrier frequency is 800+i×400Hz; where i=0, 1, ..., 12, and the length of each data modulation signal is 0.075 seconds. In the 13-bit data, the first bit is the reference bit, which is always 0; the 10th to 13th bits are the check bits, and the check method is CRC-4.

[0042] During the data exchange process, the two communicating parties first generate a random number and send it to each other. The two parties compare the size of the random number to determine the master-slave relationship. Assuming that the two communicating parties are a (the second user) and b (the first user), the random number generated by the former is 0.5 and the random number generated by the latter is 0.6, then in this communication process, b is the master and a is the slave. After determining the master-slave relationship, the master first sends its certificate Cert to the slave. b After receiving the master certificate, the slave will first send an ACK signal to the master and then send its certificate Cert. a The response signal ACK is composed of two sweep frequency signals with a frequency range of 1Hz~8000Hz and a length of 0.24 seconds.

[0043] Step 2: Receive first encrypted data sent by the first user, and use the private key of the second user to decrypt the first encrypted data to obtain a first random number and a first identity.

[0044] After receiving the certificate from the slave, the host generates a first random number RN b Then, a response signal ACK is sent to the slave and a first random number RN encrypted with the slave public key is sent. b With the first identity b .

[0045] Step 3: Generate a second random number, encrypt the second random number and the second identity of the second user using the public key of the first user, and generate and return second encrypted data.

[0046] The slave receives the encrypted first random number RN sent by the master. b and first identity b , will first use its own private key to decrypt the data, and at the same time send a response signal ACK and a second random number RN encrypted with the host public key a With the second identity a After receiving the encrypted second random number RN sent by the slave, the host a With the second identity a , will use its own private key to decrypt the data.

[0047] In this embodiment, the communicating parties use the voice communication channel to exchange data and use the negotiated key to encrypt and decrypt the exchanged data, thereby ensuring the security of the data exchange.

[0048] In one embodiment, Figure 4 A flow chart of a speech deconfusion method is provided, such as Figure 4 As shown, the above step S24, according to the speech confusion noise, de-obfuscates the first speech confusion data to restore the first speech data, includes the following steps: Step S241, dividing and aligning the first speech confusion data and speech confusion noise according to each signal sampling point to obtain a plurality of groups of corresponding first speech confusion data blocks and speech confusion noise blocks.

[0049] Assume that the received confused speech sequence is y(t), and use the first random number RN b 、First identity bAnd the corresponding noise sequence n(t) is generated from the speech data set D. The first speech confusion data y(t) and the speech confusion noise n(t) are cut into data blocks of length 2048, and each first speech confusion data block and speech confusion noise block are aligned in the time domain based on cosine similarity to obtain multiple groups of corresponding first speech confusion data blocks and speech confusion noise blocks.

[0050] Step S242: setting a parameter set, where the parameter set includes a plurality of one-dimensional optimization parameters related to the signal frequency.

[0051] Set multiple one-dimensional optimization parameters related to signal frequency , a parameter set is formed based on these multiple one-dimensional optimization parameters.

[0052] Step S243, setting the objective function based on the one-dimensional optimization parameters, traversing the parameter set, finding the minimum value of the objective function, and using the one-dimensional optimization parameters corresponding to the minimum value as the target parameters; wherein the objective function is used to measure the degree of noise residue in the original signal after applying the one-dimensional optimization parameters.

[0053] The objective function calculation formula is as follows:

[0054] in, , i represents the frequency sampling point number, represents the signal frequency, t represents the time, represents multiplication. Step S243 specifically includes the following steps: Step 1: For each frequency sampling point, the speech confusion noise block is scaled based on the one-dimensional optimization parameter to obtain a first value ; Step 2: Set the first value Confused with the corresponding first speech data block Take the absolute value and compare it to get the ratio , subtract the ratio from 1 to get the second value ; Step 3: Use the second value is the weight, and for each first speech confusion data block Weighted, get the value of the objective function.

[0055] It should be noted that before calculating the objective function, the speech confusion noise block and the first speech confusion data block need to be converted from time domain data to frequency domain data. Therefore, the calculation of the objective function and improvement factor in this application is for frequency domain data.

[0056] Step S244, based on the target parameters, calculate the improvement factor corresponding to each frequency sampling point; wherein the improvement factor is used to measure the noise suppression effect after optimization.

[0057] The improvement factor formula is as follows:

[0058] in, represents the improvement factor, i represents the frequency sampling point number, represents the signal frequency, t represents the time, represents multiplication. Step S244 specifically includes the following steps: Step 1: For each frequency sampling point, the speech confusion noise block is scaled based on the target parameter to obtain a first value ; Step 2: Set the first value Confused with the corresponding first speech data block Take the absolute value and compare it to get the ratio , subtract the ratio from 1 to get the second value ; Step 3: compare the second value with zero, and take the larger value as the value of the improvement factor.

[0059] Step S245: weighting the time-frequency spectra of each first speech confusion data block with the improvement factor as the weight to obtain the time-frequency spectra of the first speech data after de-obfuscation.

[0060] In this embodiment, by traversing the parameter set to calculate the minimum value of the objective function, the target parameter α can be obtained, which is the optimal noise coefficient. The improvement factor is calculated based on the optimal noise coefficient, and the improvement factor is applied to the time-frequency spectrum of the confused speech to obtain the purest original speech possible.

[0061] In one embodiment, after calculating the improvement factor corresponding to each frequency sampling point based on the target parameter in the above step S244, the method further includes: The channel loss rate of the first link is obtained, and the improvement factor is compared with the channel loss rate; the first link is a communication link from the first user to the second user; if the improvement factor is less than or equal to the channel loss rate, the improvement factor is set to zero.

[0062] For any frequency , and for Each data in ,if Then Set to 0. The final time-frequency spectrum of the deobfuscated speech s(t) is ,in is the time spectrum of y(t). In this embodiment, for a certain frequency ,if , it is considered that the signal at this frequency has been severely damaged by the transmission link compression algorithm to the point where it cannot be restored, so it is set to zero to completely exclude the data at this frequency. This setting retains more original signal components in some places where noise is effectively suppressed, and reduces or even eliminates signal contributions in frequencies where noise is too severe, ultimately improving signal quality.

[0063] In one embodiment, in the above step S24, before de-obfuscating the first speech confusion data according to the speech confusion noise to restore the first speech data, the method further includes: Obtain the frequency response of the first link, and use the frequency response to compensate for the voice confusion noise. In this embodiment, the voice confusion noise n(t) can be compensated in the frequency domain based on the frequency response curve FR(f). The compensation method is N(f,t)=N0(f,t)×FR(f), where N0(f,t) is the time-frequency spectrum of n(t) (i.e., the short-time Fourier transform of the original voice confusion noise). Taking into account the loss of actual channel transmission, the voice will be damaged after transmission. Assuming that the user voice is j, the noise is k, and the confused voice is j+k, it may be attenuated to j'+k' during the channel transmission process. By compensating k', the frequency domain characteristics of the noise sequence can be adjusted to match the actual channel conditions, which is beneficial to improving the quality of the first voice data subsequently recovered.

[0064] In one embodiment, in the above step S24, before de-obfuscating the first speech confusion data according to the speech confusion noise and restoring the first speech data, the method further comprises the following steps: Based on the first link, second voice data sent by the first user is received, and prior data of the second voice data is obtained; according to the second voice data and the prior data, a channel state of the first link is estimated to obtain a channel state parameter; wherein the channel state parameter includes a frequency response (Frequency Response) and / or a channel loss ratio (Loss Ratio).

[0065] In this embodiment, the second user estimates the channel state and obtains relevant parameters of the channel state by receiving the voice data of the first user.

[0066] For example, the first user sends a swept frequency signal s(t) with a length of 5 seconds and a frequency range of 5Hz~8000Hz to the second user. After lossy transmission through the channel, assuming that the signal received by the second user is s'(t), calculate the Fourier transforms S(f) and S'(f) of the two signals and estimate the channel frequency response:

[0067] Wherein, S(f) is the Fourier transform of s(t), and S'(f) is the Fourier transform of s'(t).

[0068] For example, the first user sends a normal speech signal s(t) of 5 seconds to the second user. After lossy transmission through the channel, it is assumed that the signal received by the second user is s'(t). First, the second user performs energy normalization processing on s'(t): , ensuring that the energy of the transmitted and received voice signals is equal. Next, the time spectrum S(t, f) of s(t) is calculated with a window length of 2048 and a step length of 1024, and the time spectrum S'(t, f) of s'(t) is calculated. Based on the time spectrum of these two signals, the loss percentage of the transmitted / received signal at different frequencies is calculated. Finally, the channel loss rate is calculated as:

[0069] This embodiment estimates the channel state of the first link to obtain the frequency response and / or the channel loss rate, so that when the second user de-confuses the first voice confusion data, the second user can perform frequency domain compensation for the voice confusion noise based on the frequency response and / or process the improvement factor based on the channel loss rate, thereby ultimately improving the quality of the restored first voice data.

[0070] Optionally, the channel state parameter estimation is a bidirectional estimation, that is, the first user can also estimate the channel state parameters of the second link by receiving the signal of the second user, wherein the second link is a communication link from the second user to the first user, and is configured in this way to facilitate de-obfuscation of the second user's speech confusion data when it is subsequently received.

[0071] In one embodiment, Figure 5 A flowchart of a method for protecting the privacy of voice communication based on voice obfuscation is provided. Figure 1 Taking the speech processing system as an example, the process includes the following steps: Step S11, receiving a voice input from a first user to obtain first voice data.

[0072] The voice communication system collects the voice of the first user to obtain first voice data.

[0073] Optionally, before step S11, the method further includes a system registration step: each speech processing system instance is registered with the same trusted third-party certification authority before leaving the factory, and obtains a speech processing system certificate Cert dn The certificate of the trusted third-party certification authority ca , and obtain a public speech dataset D, such as the ai_datatang dataset.

[0074] Step S12: determining a first identity matching the voiceprint information of the first user in the voice data set.

[0075] After the user obtains the corresponding voice processing system, the user registration phase can be started. A key pair can be generated using the RSA-1024 algorithm (p k ,s k ), and record a user voice signal of about 20 to 60 seconds in length to accurately extract the user's voiceprint information. The user's voiceprint information is extracted from the voice signal, and based on the voiceprint information, the cosine similarity is used to match the speaker identity s closest to the voiceprint information from the voice dataset D, which is the user identity.

[0076] Step S13, determining a first random number of the current session, and generating voice confusion noise according to the first random number, the first identity, and the voice data set.

[0077] An index sequence may be generated according to a first random number; a target voice corresponding to the first identity may be searched in the voice data set according to the index sequence; and the target voice may be concatenated to obtain voice confusion noise. Optionally, the target voice may be vowel data, because vowel data is less likely to be cracked than consonant data. This configuration may improve the security of the first voice confusion data.

[0078] Step S14: superimpose the speech confusion noise onto the first speech data to obtain the first speech confusion data.

[0079] The voice confusion noise is superimposed on the first voice data to confuse the voice of the first user. Optionally, the signal-to-noise ratio of the voice confusion noise during the confusion is -9.

[0080] From an end-to-end perspective, the channel is lossy and will destroy the encrypted data. Traditional voice communication privacy protection methods have high requirements for the integrity of encrypted data. If some bits are missing during encrypted data transmission, it will be difficult for the recipient to decrypt. In the above steps S11 to S14, by obfuscating the original voice data, the generated voice obfuscated data does not need to maintain a high degree of data integrity during transmission. Even if the data transmission is lossy, the voice can be restored based on deobfuscation. Therefore, through the method of this embodiment, it can be guaranteed that the recipient can successfully decrypt the voice data and can work normally even when the channel is lossy.

[0081] In one embodiment, Figure 3 A flow chart of data exchange between users is provided, such as Figure 3 As shown, before determining the first random number of the current session in step S13, the method further includes the following steps: Step 1: Send a first certificate of a first user to a second user.

[0082] The two communicating parties use the voice communication channel to exchange data. The data modulation method in the data exchange process is frequency modulation, and the carrier frequency is 800+i×400Hz; where i=0, 1, ..., 12, and the length of each data modulation signal is 0.075 seconds. In the 13-bit data, the first bit is the reference bit, which is always 0; the 10th to 13th bits are the check bits, and the check method is CRC-4.

[0083] During the data exchange process, the two communicating parties first generate a random number and send it to each other. The two parties compare the size of the random number to determine the master-slave relationship. Assuming that the two communicating parties are a (the second user) and b (the first user), the random number generated by the former is 0.5 and the random number generated by the latter is 0.6, then in this communication process, b is the master and a is the slave. After determining the master-slave relationship, the master first sends its certificate Cert to the slave. b .

[0084] Step 2: Receive a second certificate returned by the second user.

[0085] After receiving the host certificate, the slave will first send an ACK signal to the host and then send its certificate Cert a The response signal ACK is composed of two sweep frequency signals with a frequency range of 1Hz~8000Hz and a length of 0.24 seconds.

[0086] Step 3: Generate a first random number, encrypt the first random number and the first identity using the public key of the second user, and send the generated first encrypted data to the second user.

[0087] After receiving the certificate from the slave, the host generates a first random number RN b Then, a response signal ACK is sent to the slave and a first random number RN encrypted with the slave public key is sent. b With the first identity b .

[0088] Step 4, receiving the second encrypted data returned by the second user, and decrypting the second encrypted data according to the private key of the first user; wherein the second encrypted data is generated by encrypting the second random number and the second identity of the second user with the public key of the first user.

[0089] The slave receives the encrypted first random number RN sent by the master. b and first identity b , will first use its own private key to decrypt the data, and at the same time send a response signal ACK and a second random number RN encrypted with the host public key a With the second identity a After receiving the encrypted second random number RN sent by the slave, the hosta With the second identity a , will use its own private key to decrypt the data.

[0090] In this embodiment, the communicating parties use the voice communication channel to exchange data and use the negotiated key to encrypt and decrypt the exchanged data, thereby ensuring the security of the data exchange.

[0091] In one embodiment, in step S14, after the speech confusion noise is superimposed on the first speech data to obtain the first speech confusion data, the method further includes: The voice processing system directly sends the first voice confusion data to the second user, that is, the voice processing system can communicate with the external device using the service of the communication service provider.

[0092] Alternatively, the voice communication system first sends the first voice obfuscation data to the first communication terminal corresponding to the first user, and then sends it to the second user via the first communication terminal. This setting is based on the consideration that if the first communication terminal is attacked by a third party, it will become a dishonest device. Therefore, by using a voice processing system to collect and process user voices, and then sending the processed user voices to the communication terminal, potential third-party attacks on the communication terminal can be isolated.

[0093] In one embodiment, before the above step S14, sending the first speech confusion data to the second user, the method further includes the following steps: Receive a voice input from a first user to obtain second voice data; send the second voice data to a second user based on the first link, and provide prior data of the second voice data to the second user.

[0094] In this embodiment, the first user sends and provides voice data to the second user, so that the second user can estimate the channel state and obtain relevant parameters of the channel state.

[0095] For example, the first user sends a swept frequency signal s(t) of 5 Hz to 8000 Hz to the second user. Assuming that the signal received by the second user is s'(t), the frequency response is calculated as:

[0096] Wherein, S(f) is the Fourier transform of s(t), and S'(f) is the Fourier transform of s'(t).

[0097] For example, the first user sends a normal speech signal s(t) to the second user. Assuming that the signal received by the second user is s'(t), the second user first performs energy normalization processing: s(t)=s(t), ; Calculate the time spectrum S(t,f) of s(t), and calculate the time spectrum S'(t,f) of s'(t); The final calculated channel loss rate is:

[0098] This embodiment estimates the channel state of the first link to obtain the frequency response and / or the channel loss rate, so that when the second user de-confuses the first voice confusion data, the second user can perform frequency domain compensation for the voice confusion noise based on the frequency response and / or process the improvement factor based on the channel loss rate, thereby ultimately improving the quality of the restored first voice data.

[0099] Optionally, the channel state parameter estimation is a bidirectional estimation, that is, the first user can also estimate the channel state parameters of the second link by receiving the signal of the second user, wherein the second link is a communication link from the second user to the first user, and is configured in this way to facilitate de-obfuscation of the second user's speech confusion data when it is subsequently received.

[0100] In one embodiment, a voice communication system is provided. The voice communication system includes: a voice processing system and a communication terminal; the voice processing system and the communication terminal are connected. When the voice communication system acts as a sender, the voice processing system is used to execute the voice communication privacy protection method based on voice obfuscation corresponding to the sender, and send the generated voice obfuscation data to the other end through the communication terminal. When the voice communication system acts as a receiver, the voice obfuscation data is received through the communication terminal, and the voice obfuscation data is sent to the voice processing system, and the voice processing system is used to execute the voice communication privacy protection method based on voice obfuscation of the receiver, and de-obfuscate the voice obfuscation data.

[0101] The above-mentioned voice communication system can be applied to Figure 6 In the application environment shown. Figure 6 As shown, the first user side has a first voice processing system and a first communication terminal, and the second user side has a second voice processing system and a second communication terminal. The voice processing systems of both parties are connected to their respective communication terminals, and the communication terminals of both parties are connected to each other in communication. For example, the voice processing system can be a portable wearable device such as a smart watch, a smart bracelet, a head-mounted device, etc. The wearable device has a microphone and a speaker, and the processor is built into the wearable device. The communication terminal can be, but is not limited to, various personal computers, laptops, smart phones, tablet computers, Internet of Things devices, etc.

[0102] Figure 7 for Figure 6 The operation principle diagram of the voice communication system in Figure 7 As shown, the operation process of the voice communication system includes the following steps: Step S1, voice processing system registration. Each voice processing system instance is registered with a trusted third-party certification authority before leaving the factory to obtain a personal certificate and a third-party certification authority certificate. The voice processing system contains a public voice data set.

[0103] Step S2: User registration and key generation: After the user obtains the voice processing system, a key pair is generated, and the system is used to record a personal voice signal, and the identity of the voice signal in the voice data set is matched based on the voiceprint.

[0104] Step S3: channel parameter estimation: After the user starts voice communication, the systems of both communicating parties estimate the channel state and obtain relevant parameters of the channel state.

[0105] Step S4, data key exchange. After the channel state parameter estimation is completed, the communicating parties automatically exchange keys through the voice transmission channel. After the key exchange is completed, the communicating parties use the key to encrypt and transmit the current session random number; Step S5, sender noise generation and voice confusion. During the user voice communication process, the user's voice input is received, voice confusion noise is generated using random numbers and the user's phoneme data set, the confusion noise is added to the user's voice to confuse the voice data, and the confused voice data is sent to the communication terminal; Step S6, receiving side noise generation and voice deobfuscation. During the user voice communication process, the obfuscated voice is received from the communication terminal, the obfuscated noise is generated using a random number, the obfuscated voice is deobfuscated based on the obfuscated noise, and the deobfuscated voice is played to the user.

[0106] Compared with the encryption methods of existing communication service providers, the voice processing system or voice communication system of this embodiment does not rely on a trusted service provider and is more universal. Compared with other existing encryption-based methods, the voice processing system or voice communication system of this embodiment can work normally in the case of lossy channels, is more practical, and can be used in actual situations. The voice processing system or voice communication system of this embodiment can adapt to different types of voice communication channels without requiring additional operations by the user.

[0107] Figure 8This is an experimental result diagram for the user voice privacy protection capability, which is used to verify the voice processing system's ability to protect the user's voice privacy during actual voice message communication. This experiment shows the effect of user voice privacy protection under different voice communication channels and different attacker conditions, where the horizontal axis represents different voice message communication channels (T: Telegram, V: Viber, M: Messenger, S: Skype, L: Line, D: DingTalk, WA: WhatsAPP, WE: WeChat). The experimental results show that in the absence of any protection, if the attacker can obtain the voice signal during the transmission process, the attacker can obtain the user's private semantic information from it and achieve a word error rate of less than 5%. In the case of voice processing system protection, for any attacker, even if he can obtain the obfuscated voice signal during the transmission process, the word error rate of the semantic information obtained is greater than 75%.

[0108] Fig. 9 This is an experimental result chart for voice communication quality, which is used to verify that the voice processing system will not have a significant impact on the user's voice communication quality. This experiment shows the impact on the quality of voice communication under different voice communication channels, where the horizontal axis represents different voice message communication channels (T: Telegram, V: Viber, M: Messenger, S: Skype, L: Line, D: DingTalk, WA: WhatsAPP, WE: WeChat). In the absence of protection, the error rate of semantic information obtained by users during voice communication is about 5% on average. In the case of voice processing system protection, the accuracy of semantic information that users can obtain in the first four communication channels is almost unaffected. The error rate in the last four channels is slightly higher, but still less than 40%.

[0109] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0110] In addition, in combination with the voice communication privacy protection method based on voice confusion provided in the above embodiment, the present application also provides a storage medium, on which a computer program is stored; when the computer program is executed by a processor, the following steps are implemented: Receiving first speech confusion data sent by a first user to a second user; Obtaining a first random number corresponding to a first user and a first identity of the first user; According to the first random number and the first identity, searching for a target voice in a voice data set, and concatenating the target voice to generate voice confusion noise; According to the speech confusion noise, the first speech confusion data is de-obfuscated to restore the first speech data.

[0111] In one embodiment, the computer program implements the following steps when executed by a processor: before obtaining a first random number corresponding to the first user and a first identity of the first user, the method further comprises; Receive a first certificate sent by a first user, and return a second certificate of a second user; Receiving first encrypted data sent by the first user, and decrypting the first encrypted data using the private key of the second user to obtain a first random number and a first identity; A second random number is generated, the second random number and the second identity of the second user are encrypting by using the public key of the first user, and second encrypted data is generated and returned.

[0112] In one embodiment, when the computer program is executed by a processor, the following steps are implemented: according to the speech confusion noise, the first speech confusion data is deobfuscated to restore the first speech data, including: Dividing and aligning the first speech confusion data and speech confusion noise according to each signal sampling point to obtain a plurality of corresponding groups of first speech confusion data blocks and speech confusion noise blocks; Setting a parameter set, the parameter set includes a plurality of one-dimensional optimization parameters related to the signal frequency; The objective function is set based on the one-dimensional optimization parameter, the parameter set is traversed, the minimum value of the objective function is obtained, and the one-dimensional optimization parameter corresponding to the minimum value is used as the target parameter; wherein the objective function is used to measure the degree of noise residual in the original signal after applying the one-dimensional optimization parameter; Based on the target parameters, an improvement factor corresponding to each frequency sampling point is calculated; wherein the improvement factor is used to measure the noise suppression effect after optimization; The time-frequency spectra of each first speech confusion data block are weighted with the improvement factor as the weight to obtain the time-frequency spectra of the first speech data after de-obfuscation.

[0113] In one embodiment, when the computer program is executed by a processor, the following steps are implemented: setting an objective function based on a one-dimensional optimization parameter, traversing the parameter set, and finding the minimum value of the objective function, including: For each frequency sampling point, scaling the speech confusion noise block based on the one-dimensional optimization parameter to obtain a first value; Compare the first value with the corresponding first speech confusion data block after taking the absolute value to obtain a ratio, and subtract the ratio from 1 to obtain a second value; The second value is used as the weight to weight each first speech confusion data block to obtain the value of the objective function.

[0114] In one embodiment, when the computer program is executed by a processor, the following steps are implemented: based on the target parameter, calculating the improvement factor corresponding to each frequency sampling point, including: For each frequency sampling point, scaling the speech confusion noise block based on the target parameter to obtain a first value; Compare the first value with the corresponding first speech confusion data block after taking the absolute value to obtain a ratio, and subtract the ratio from 1 to obtain a second value; The second value is compared with zero, and the larger value is taken as the value of the improvement factor.

[0115] In one embodiment, when the computer program is executed by a processor, the following steps are implemented: after calculating the improvement factor corresponding to each frequency sampling point based on the target parameter, the method further includes: Obtaining a channel loss rate of a first link, and comparing the improvement factor with the channel loss rate; the first link is a communication link from a first user to a second user; If the improvement factor is less than or equal to the channel loss rate, the improvement factor is set to zero.

[0116] In one embodiment, when the computer program is executed by a processor, the following steps are implemented: before de-obfuscating the first speech confusion data according to the speech confusion noise to restore the first speech data, the method further includes: A frequency response of the first link is obtained, and the frequency response is used to compensate for the speech confusion noise.

[0117] In one embodiment, when the computer program is executed by a processor, the following steps are implemented: before de-obfuscating the first speech confusion data according to the speech confusion noise to restore the first speech data, the method further includes: Receiving second voice data sent by the first user based on the first link, and acquiring prior data of the second voice data; According to the second speech data and the priori data, the channel state of the first link is estimated to obtain channel state parameters; wherein the channel state parameters include frequency response and / or channel loss rate.

[0118] In one embodiment, the computer program, when executed by a processor, implements the following steps: Receiving a voice input from a first user to obtain first voice data; Determining a first identity in the voice data set that matches the voiceprint information of the first user; Determine a first random number of the current session, and generate voice confusion noise according to the first random number, the first identity, and the voice data set; The speech confusion noise is superimposed on the first speech data to obtain the first speech confusion data.

[0119] In one embodiment, when the computer program is executed by a processor, the following steps are implemented: generating speech confusion noise according to the first random number, the first identity and the speech data set, including: Generate an index sequence according to the first random number; Searching for a target speech corresponding to a first identity in the speech dataset according to the index sequence; The target speech is concatenated to obtain speech confusion noise.

[0120] In one embodiment, when the computer program is executed by a processor, the following steps are implemented: before determining the first random number of the current session, the method further includes: sending a first certificate of the first user to a second user; receiving a second certificate returned by the second user; Generate a first random number, encrypt the first random number and the first identity using the public key of the second user, and send the generated first encrypted data to the second user; The second encrypted data returned by the second user is received, and the second encrypted data is decrypted according to the private key of the first user; wherein the second encrypted data is generated by encrypting the second random number and the second identity of the second user with the public key of the first user.

[0121] In one embodiment, when the computer program is executed by a processor, the following steps are implemented: after adding the speech confusion noise to the first speech data to obtain the first speech confusion data, the method further includes: Sending the first voice confusion data to the second user; or, The first speech confusion data is sent to a first communication terminal corresponding to the first user, and then sent to the second user via the first communication terminal.

[0122] In one embodiment, when the computer program is executed by a processor, the following steps are implemented: before sending the first speech confusion data to the second user, the method further includes: Receiving a voice input from a first user to obtain second voice data; The second voice data is sent to the second user based on the first link, and the priori data of the second voice data is provided to the second user.

[0123] In addition, in combination with the voice communication privacy protection method based on voice obfuscation provided in the above embodiment, the present application also provides a computer program product, including a computer program, which implements the following steps when executed by a processor: Receiving first speech confusion data sent by a first user to a second user; Obtaining a first random number corresponding to a first user and a first identity of the first user; According to the first random number and the first identity, searching for a target voice in a voice data set, and concatenating the target voice to generate voice confusion noise; According to the speech confusion noise, the first speech confusion data is de-obfuscated to restore the first speech data.

[0124] In one embodiment, the computer program implements the following steps when executed by a processor: before obtaining a first random number corresponding to the first user and a first identity of the first user, the method further comprises; Receive a first certificate sent by a first user, and return a second certificate of a second user; Receiving first encrypted data sent by the first user, and decrypting the first encrypted data using the private key of the second user to obtain a first random number and a first identity; A second random number is generated, the second random number and the second identity of the second user are encrypting by using the public key of the first user, and second encrypted data is generated and returned.

[0125] In one embodiment, when the computer program is executed by a processor, the following steps are implemented: according to the speech confusion noise, the first speech confusion data is deobfuscated to restore the first speech data, including: Dividing and aligning the first speech confusion data and speech confusion noise according to each signal sampling point to obtain a plurality of corresponding groups of first speech confusion data blocks and speech confusion noise blocks; Setting a parameter set, the parameter set includes a plurality of one-dimensional optimization parameters related to the signal frequency; The objective function is set based on the one-dimensional optimization parameter, the parameter set is traversed, the minimum value of the objective function is obtained, and the one-dimensional optimization parameter corresponding to the minimum value is used as the target parameter; wherein the objective function is used to measure the degree of noise residual in the original signal after applying the one-dimensional optimization parameter; Based on the target parameters, an improvement factor corresponding to each frequency sampling point is calculated; wherein the improvement factor is used to measure the noise suppression effect after optimization; The time-frequency spectra of each first speech confusion data block are weighted with the improvement factor as the weight to obtain the time-frequency spectra of the first speech data after de-obfuscation.

[0126] In one embodiment, when the computer program is executed by a processor, the following steps are implemented: setting an objective function based on a one-dimensional optimization parameter, traversing the parameter set, and finding the minimum value of the objective function, including: For each frequency sampling point, scaling the speech confusion noise block based on the one-dimensional optimization parameter to obtain a first value; Compare the first value with the corresponding first speech confusion data block after taking the absolute value to obtain a ratio, and subtract the ratio from 1 to obtain a second value; The second value is used as the weight to weight each first speech confusion data block to obtain the value of the objective function.

[0127] In one embodiment, when the computer program is executed by a processor, the following steps are implemented: based on the target parameter, calculating the improvement factor corresponding to each frequency sampling point, including: For each frequency sampling point, scaling the speech confusion noise block based on the target parameter to obtain a first value; Compare the first value with the corresponding first speech confusion data block after taking the absolute value to obtain a ratio, and subtract the ratio from 1 to obtain a second value; The second value is compared with zero, and the larger value is taken as the value of the improvement factor.

[0128] In one embodiment, when the computer program is executed by a processor, the following steps are implemented: after calculating the improvement factor corresponding to each frequency sampling point based on the target parameter, the method further includes: Obtaining a channel loss rate of a first link, and comparing the improvement factor with the channel loss rate; the first link is a communication link from a first user to a second user; If the improvement factor is less than or equal to the channel loss rate, the improvement factor is set to zero.

[0129] In one embodiment, when the computer program is executed by a processor, the following steps are implemented: before de-obfuscating the first speech confusion data according to the speech confusion noise to restore the first speech data, the method further includes: A frequency response of the first link is obtained, and the frequency response is used to compensate for the speech confusion noise.

[0130] In one embodiment, when the computer program is executed by a processor, the following steps are implemented: before de-obfuscating the first speech confusion data according to the speech confusion noise to restore the first speech data, the method further includes: Receiving second voice data sent by the first user based on the first link, and acquiring prior data of the second voice data; According to the second speech data and the priori data, the channel state of the first link is estimated to obtain channel state parameters; wherein the channel state parameters include frequency response and / or channel loss rate.

[0131] In one embodiment, the computer program, when executed by a processor, implements the following steps: Receiving a voice input from a first user to obtain first voice data; Determining a first identity in the voice data set that matches the voiceprint information of the first user; Determine a first random number of the current session, and generate voice confusion noise according to the first random number, the first identity, and the voice data set; The speech confusion noise is superimposed on the first speech data to obtain the first speech confusion data.

[0132] In one embodiment, when the computer program is executed by a processor, the following steps are implemented: generating speech confusion noise according to the first random number, the first identity and the speech data set, including: Generate an index sequence according to the first random number; Searching for a target speech corresponding to a first identity in the speech dataset according to the index sequence; The target speech is concatenated to obtain speech confusion noise.

[0133] In one embodiment, when the computer program is executed by a processor, the following steps are implemented: before determining the first random number of the current session, the method further includes: sending a first certificate of the first user to a second user; receiving a second certificate returned by the second user; Generate a first random number, encrypt the first random number and the first identity using the public key of the second user, and send the generated first encrypted data to the second user; The second encrypted data returned by the second user is received, and the second encrypted data is decrypted according to the private key of the first user; wherein the second encrypted data is generated by encrypting the second random number and the second identity of the second user with the public key of the first user.

[0134] In one embodiment, when the computer program is executed by a processor, the following steps are implemented: after adding the speech confusion noise to the first speech data to obtain the first speech confusion data, the method further includes: Sending the first voice confusion data to the second user; or, The first speech confusion data is sent to a first communication terminal corresponding to the first user, and then sent to the second user via the first communication terminal.

[0135] In one embodiment, when the computer program is executed by a processor, the following steps are implemented: before sending the first speech confusion data to the second user, the method further includes: Receiving a voice input from a first user to obtain second voice data; The second voice data is sent to the second user based on the first link, and the priori data of the second voice data is provided to the second user.

[0136] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0137] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.

[0138] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0139] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A voice communication privacy protection method based on voice obfuscation, characterized in that: include: Receiving first speech confusion data sent by a first user to a second user; Obtaining a first random number corresponding to the first user and a first identity of the first user; According to the first random number and the first identity, searching for a target voice in a voice data set, and concatenating the target voice to generate voice confusion noise; The first speech confusion data is de-obfuscated according to the speech confusion noise to restore the first speech data.

2. The voice communication privacy protection method based on voice obfuscation according to claim 1, characterized in that: De-obfuscating the first speech obfuscated data according to the speech obfuscation noise to restore the first speech data includes: Dividing and aligning the first speech confusion data and the speech confusion noise according to each signal sampling point to obtain a plurality of groups of corresponding first speech confusion data blocks and speech confusion noise blocks; Setting a parameter set, wherein the parameter set includes a plurality of one-dimensional optimization parameters related to the signal frequency; An objective function is set based on the one-dimensional optimization parameter, the parameter set is traversed, the minimum value of the objective function is obtained, and the one-dimensional optimization parameter corresponding to the minimum value is used as the target parameter; wherein the objective function is used to measure the degree of noise residual in the original signal after applying the one-dimensional optimization parameter; Based on the target parameter, calculating an improvement factor corresponding to each frequency sampling point; wherein the improvement factor is used to measure the noise suppression effect after optimization; The time-frequency spectra of each of the first speech confusion data blocks are weighted using the improvement factor as a weight to obtain the time-frequency spectra of the first speech data after deobfuscation.

3. The voice communication privacy protection method based on voice obfuscation according to claim 2 is characterized in that: Setting an objective function based on the one-dimensional optimization parameter, traversing the parameter set, and finding the minimum value of the objective function includes: For each frequency sampling point, scaling the speech confusion noise block based on the one-dimensional optimization parameter to obtain a first value; Compare the first value with the corresponding first speech confusion data block by taking absolute values ​​to obtain a ratio, and subtract the ratio from 1 to obtain a second value; The second value is used as a weight to weight each of the first speech confusion data blocks to obtain the value of the objective function.

4. The voice communication privacy protection method based on voice obfuscation according to claim 2, characterized in that: Based on the target parameter, calculating an improvement factor corresponding to each frequency sampling point includes: For each frequency sampling point, scaling the speech confusion noise block based on the target parameter to obtain a first value; Compare the first value with the corresponding first speech confusion data block by taking absolute values ​​to obtain a ratio, and subtract the ratio from 1 to obtain a second value; The second value is compared with zero, and the larger value is taken as the value of the improvement factor.

5. The voice communication privacy protection method based on voice obfuscation according to claim 2, characterized in that: After calculating the improvement factor corresponding to each frequency sampling point based on the target parameter, the method further includes: Acquire a channel loss rate of a first link, and compare the improvement factor with the channel loss rate; the first link is a communication link from the first user to the second user; If the improvement factor is less than or equal to the channel loss rate, the improvement factor is set to zero.

6. The voice communication privacy protection method based on voice obfuscation according to claim 1, characterized in that: Before de-obfuscating the first speech confusion data according to the speech confusion noise to restore the first speech data, the method further includes: A frequency response of the first link is obtained, and the speech confusion noise is compensated by using the frequency response.

7. The voice communication privacy protection method based on voice obfuscation according to any one of claims 1 to 6, characterized in that: Before de-obfuscating the first speech confusion data according to the speech confusion noise to restore the first speech data, the method further includes: receiving second voice data sent by the first user based on the first link, and acquiring prior data of the second voice data; According to the second voice data and the priori data, the channel state of the first link is estimated to obtain channel state parameters; wherein the channel state parameters include frequency response and / or channel loss rate.

8. A voice communication privacy protection method based on voice obfuscation, characterized in that: The method comprises: Receiving a voice input from a first user to obtain first voice data; Determining a first identity in the voice data set that matches the voiceprint information of the first user; Determine a first random number of the current session, and generate voice confusion noise according to the first random number, the first identity, and the voice data set; The speech confusion noise is superimposed on the first speech data to obtain first speech confusion data.

9. The voice communication privacy protection method based on voice obfuscation according to claim 8, characterized in that: Generating speech confusion noise according to the first random number, the first identity, and the speech data set includes: Generate an index sequence according to the first random number; Searching for a target speech corresponding to the first identity in the speech data set according to the index sequence; The target speech is concatenated to obtain the speech confusion noise.

10. A speech processing system, characterized in that: include: A processor, a speaker and a microphone, wherein the speaker and the microphone are connected to the processor respectively; wherein, The processor is used to execute the voice communication privacy protection method based on voice confusion described in any one of claims 1 to 7; or, the processor is used to execute the voice communication privacy protection method based on voice confusion described in claim 8 or claim 9.

11. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 9 are implemented.

Images