Anti-eavesdropping sound recording interference method oriented to Chinese speech and based on ultrasonic waves

Abstract

The invention discloses a Chinese speech anti-eavesdropping-oriented ultrasonic-based recording interference method, which designs a user-non-inductive speech eavesdropping interference technology, fully considers the ability of an eavesdropper and the risk of acoustic privacy leakage caused by the ability under real conditions, enhances the effectiveness and security of a system, and improves the reliability of the system. User corpora are expanded by using a speech generation technology, and the corpora are cut into phonemes without semantics according to Chinese speech characteristics, so that the speech registration time cost of a user is reduced, the privacy of user registration speech is protected, the protection capability for Chinese speech is enhanced, noise coupled with the user sound is generated according to the user sound, and the user experience is improved. The difficulty of noise separation by an eavesdropper is increased, the voice of a user is more completely protected from being eavesdropped, the ultrasonic signal is corrected for the audible sound leakage problem of the ultrasonic transmitting equipment, the leakage degree of the audible sound is weakened, the use experience of the user is enhanced, and the risk that the ultrasonic transmitting equipment is found by the eavesdropper is reduced.

Description

Technical Field

[0001] The invention relates to the technical field of eavesdropping defense, and in particular to an ultrasonic-based recording interference method for preventing Chinese voice eavesdropping. Background Art

[0002] With the growing demand for voice privacy protection, various anti-eavesdropping technologies have emerged. Among them, the ultrasonic recording jamming system has attracted widespread attention from researchers and the consumer market due to its unique advantages, such as being imperceptible to the human ear, widely applicable to a wide range of devices, large coverage, and low deployment cost. The ultrasonic jamming system uses the nonlinear characteristics of the microphone and uses ultrasonic signals that are inaudible to the human ear to generate low-frequency noise in the electronic recording device, thereby effectively resisting eavesdropping and protecting voice privacy. Due to the nonlinear characteristics of the microphones in electronic recording devices, high-frequency ultrasonic signals will be distorted when they exceed their frequency response range, and then distorted into low-frequency noise in the audible frequency band. By carefully designing and modulating ultrasonic signals, electronic recording devices can be efficiently interfered with to prevent the occurrence and harm of eavesdropping. However, in actual scenarios, eavesdroppers will use various technical means to reduce the noise interference in the recording, thereby restoring the original voice information and achieving the purpose of stealing voice privacy.

[0003] In order to deal with the existing security risks in the field of eavesdropping interference, the present invention proposes an ultrasonic recording interference method designed specifically for Chinese speech. Taking into account that the eavesdropper may use a variety of denoising methods, the present invention designs a coupling noise generation algorithm based on the characteristics of Chinese speech. There is no need to collect the user's voice content in real time. The ultrasonic interference noise generated by the algorithm is highly correlated with the voice signal to be protected and difficult to separate. In addition, in order to improve the practicality and user experience of the system, the present invention designs an ultrasonic carrier modulation strategy, which effectively avoids the generation and leakage of audible frequency band noise caused by hardware imperfections such as ringing effects, ensures that the user's hearing will not be affected in any way during use, and takes into account the efficient noise injection of different types of eavesdropping devices. Summary of the invention

[0004] In view of the problems raised in the existing methods, the present invention provides an ultrasonic recording interference method for preventing Chinese voice eavesdropping, which solves the security risks that can be avoided by algorithms in the existing methods. The present invention is implemented through the following schemes:

[0005] The invention discloses an ultrasonic recording interference method for preventing Chinese voice eavesdropping, comprising the following steps:

[0006] Obtain a short voice segment of the user to be protected;

[0007] Use speech generation technology to generate a large amount of user corpus data based on short segments of user speech;

[0008] Generate the user's phoneme-level corpus by using the phoneme processing algorithm to process the user's corpus data;

[0009] Through the coupled noise generation algorithm, any number of phonemes are randomly selected from the phoneme-level corpus each time noise is generated to generate coupled noise;

[0010] The coupled noise is modulated by an ultrasonic carrier wave, and the corrected sound wave is emitted through an ultrasonic speaker using an ultrasonic correction algorithm.

[0011] As a further improvement, the phoneme processing algorithm described in the present invention is specifically:

[0012] 1) Use the maximum inter-class variance method to calculate the strength threshold of the corpus data and cut the user's corpus data into phonemes;

[0013] 2) clustering the cut phonemes according to vowels, diphthongs and consonants, and treating different tones of the same phoneme as different categories during clustering;

[0014] 3) The clustered phoneme signals of each category are aligned and averaged using a dynamic time warping algorithm to obtain the average signal of each phoneme of the user as the phoneme library.

[0015] As a further improvement, the coupling noise generation algorithm described in the present invention is specifically:

[0016] 1) Randomly extract phonemes from the user's phoneme-level corpus, that is, randomly extract any number of phonemes at short intervals to generate the user's semantic-free phoneme sequence s(t);

[0017] 2) Generate a noise signal n with similar characteristics to the user's voice based on the user's phoneme sequence s(t) 1 (t):

[0018]

[0019] Among them, M 1×2 (t) is a 1x2 non-singular random matrix; n r1 (t) is a random noise; L[·] is a non-integrable nonlinear aliasing function:

[0020]

[0021] 3) Based on n 1 (t), a random noise is introduced in the time domain and frequency domain through convolution operation to generate a time-frequency noise signal n 2 (t):

[0022]

[0023] Among them, * is the convolution operation; n ri (t)(i=1,2,3) are random noises that are independent of each other and have different distributions;

[0024] 4) Based on n 2 (t), using frequency hopping and spectrum spreading technology to broaden its spectrum bandwidth, and finally form coupled noise n 3 (t):

[0025]

[0026] in, is the frequency hopping signal; n is the nth frequency point in the frequency hopping table; is the phase modulation function, where d is the signal code after error coding, t is the time, and Δω is the frequency difference component before and after frequency hopping.

[0027] As a further improvement, the ultrasonic noise correction algorithm described in the present invention is specifically:

[0028] 1) Use the frequency domain filter to smooth the ultrasonic signal. The expression of the frequency domain filter is:

[0029]

[0030] Where ω is the angular frequency;

[0031] 2) Decompose the smoothed filtered ultrasonic signal into multiple narrow bandwidth signals, and make the frequency of each signal segment lower than 50 Hz. Each of the decomposed signals is transmitted using a different ultrasonic transmitter.

[0032] The beneficial effects of the present invention are as follows:

[0033] The invention discloses a recording interference method based on ultrasonic waves.

[0034] Aiming at the problem of preventing Chinese voice eavesdropping, the present invention designs a voice eavesdropping interference technology that is imperceptible to users, and fully considers the capabilities of eavesdroppers under realistic conditions and the risks of acoustic privacy leakage caused by them, thereby enhancing the effectiveness and security of the system.

[0035] The present invention uses speech generation technology to expand user corpus, and cuts the corpus into semantically non-semantic phonemes based on the characteristics of Chinese speech, thereby reducing the user's voice registration time cost, protecting the privacy of the user's registered voice, and enhancing the protection capability for Chinese speech.

[0036] The present invention generates noise coupled with the user's voice according to the user's voice, thereby increasing the difficulty for an eavesdropper to separate the noise and more completely protecting the user's voice from being eavesdropped.

[0037] The present invention corrects the ultrasonic signal for the problem of audible sound leakage of the ultrasonic transmitting device, thereby reducing the degree of audible sound leakage, enhancing the user experience, and reducing the risk of the ultrasonic transmitting device being discovered by an eavesdropper. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 Flow chart of the system of the present invention. DETAILED DESCRIPTION

[0039] The technical solution of the present invention is further described below by means of specific embodiments in conjunction with the accompanying drawings:

[0040] The invention discloses an ultrasonic recording interference method for preventing Chinese speech from eavesdropping. Figure 1 This is a flow chart of the system of the present invention. Aiming at the problem of anti-eavesdropping of Chinese voice, a user-imperceptible voice eavesdropping interference technology is proposed, and unique coupling noise is generated according to the user's voice characteristics to enhance the effectiveness and security of the system.

[0041] The present invention comprises the following steps:

[0042] Step 1: Obtain a short voice segment of the user to be protected;

[0043] Step 2: Use speech generation technology to generate a large amount of user corpus data based on the user's short speech segments;

[0044] Step 3: Generate the user's phoneme-level corpus by using the phoneme processing algorithm on the user's corpus data;

[0045] Step 4: Using the coupled noise generation algorithm, any number of phonemes are randomly selected from the phoneme-level corpus each time noise is generated to generate coupled noise;

[0046] Step 5: The coupled noise is modulated by an ultrasonic carrier wave, and the corrected sound wave is emitted through an ultrasonic speaker using an ultrasonic correction algorithm.

[0047] Specifically, the steps of the present invention are detailed as follows.

[0048] Step 1: Get a short voice clip of the user to be protected:

[0049] Use recording equipment to collect the user's spoken voice for a few seconds. This is used for subsequent user corpus generation.

[0050] Step 2: Use speech generation technology to generate a large amount of user corpus data based on the user's short speech segments:

[0051] The collected short-duration user voice data is used to generate more user corpus through the SV2TTS speech generation network.

[0052] Step 3: Generate the user's phoneme-level corpus by using the phoneme processing algorithm:

[0053] In order to further improve the correlation between the coupling noise based on the generated voice signal and the content to be protected, the present invention uses the vowel-consonant system and intonation characteristics of Chinese to generate more targeted user corpus. The present invention divides the generated corpus content according to phonemes, and classifies it according to vowels, consonants and tones, and obtains the average corpus of each phoneme of the user to be protected. It is specifically divided into the following steps:

[0054] 1) Use the maximum inter-class variance method (also known as the Big Law method, OTSU) to calculate the speech signal strength threshold and use it as a basis for cutting phonemes.

[0055] 2) Cluster each phoneme according to vowels, diphthongs and consonants. Considering that the tones in Chinese speech have a great influence on the speech spectrum characteristics, different tones of the same phoneme are also regarded as different categories. The number of categories of the final clustering = (number of vowel categories + number of diphthong categories + number of consonant categories) × number of tones. The number of categories here can be adjusted according to the language type. For example, some dialects have diconsonant phonemes, and Wu dialects have tones such as "yin shang" and "yang shang".

[0056] 3) The corpus signals within the same pitch phonemes are aligned based on the Dynamic Time Warping (DTW) algorithm and then averaged to obtain the average signal of each phoneme of the user.

[0057] Step 4: Through the coupled noise generation algorithm, any number of phonemes are randomly selected from the phoneme-level corpus each time noise is generated to generate coupled noise:

[0058] In order to further improve the security of coupled noise and prevent eavesdroppers from using noise elimination algorithms to circumvent anti-eavesdropping interference, the present invention randomly selects phonemes from the user's phoneme-level corpus to form semantic-free speech, and uses this to generate coupled noise to enhance the noise's anti-noise denoising ability. Specifically, it is divided into the following steps:

[0059] 1) Randomly extract phonemes from the user's phoneme-level corpus, that is, randomly extract any number of phonemes at short intervals to generate the user's semantic-free phoneme sequence s(t).

[0060] 2) The phoneme sequence s(t) is nonlinearly mixed with random noise in the time domain to generate a noise signal n with similar characteristics to the user's voice. 1 (t):

[0061]

[0062] Among them, M1×2 (t) is a 1x2 non-singular random matrix; n r1 (t) is a random noise; L[·] is a non-integrable nonlinear aliasing function:

[0063]

[0064] Since nonlinear aliasing has a many-to-many mapping relationship, its inverse transform is not unique. This feature makes the generated coupled noise highly complex and robust, increasing the difficulty of separating noise from speech signals.

[0065] 3) Based on the noise signal n that is similar to the user's voice characteristics 1 (t), and enhance its interference capability in the time domain and frequency domain respectively. Specifically, a random noise is introduced in the time domain and frequency domain respectively through convolution operation. The resulting coupling noise is:

[0066]

[0067] Among them, * is the convolution operation; n ri (t)(i=1,2,3) are random noises that are independent of each other and have different distributions. Through convolution operations in the time domain and frequency domain, the resulting noise is tightly coupled with the speech signal, which enhances the ability of the noise to resist the denoising measures of eavesdroppers.

[0068] 4) Based on n 2 (t), in order to resist the denoising methods such as ultrasonic sniffing that use time-frequency domain characteristics, the frequency hopping spread spectrum technology is used to broaden the spectrum bandwidth of the coupled noise. The resulting coupled noise is:

[0069]

[0070] in, is the frequency hopping signal; n is the nth frequency point in the frequency hopping table; is the phase modulation function, where d is the signal after error coding, t is the time, and Δω is the frequency difference component before and after frequency hopping. 3 The bandwidth of (t) is greatly expanded. In this case, it is difficult for the eavesdropper to obtain all the characteristics of the interference noise, so it is impossible to recover the speech signal.

[0071] Step 5: Modulate the coupled noise through the ultrasonic carrier, and transmit the corrected sound wave through the ultrasonic speaker through the ultrasonic correction algorithm:

[0072] Due to the imperfect characteristics of the ultrasonic transmitter, the ultrasonic wave will generate audible noise during the transmission process, affecting the user's daily use and more likely to alert the eavesdropper, so that he will take measures to circumvent anti-eavesdropping. In order to solve this problem, the present invention designs an ultrasonic correction algorithm from two aspects: reducing the ringing effect and splitting the broadband spectrum.

[0073] 1) Reduce the ringing effect: The ringing effect will cause audible noise to be generated at the ultrasonic transmitter. When the frequency of the sound signal changes dramatically, a pulse signal will be generated in the hardware device. This pulse wave has higher energy in the audible sound frequency band, and the oscillation on its sound waveform is similar to a ringing tone, so it is called the ringing effect. In the sound generation process of the ultrasonic speaker, the bandpass filter is used to cut off the signal frequency to ensure that the ultrasonic signal of a specific frequency is generated without deviation. However, due to the hardware imperfections of the bandpass filter, a sudden change in frequency may occur in this process, thereby triggering the appearance of the ringing effect. In principle, the mutated sound signal can be regarded as being truncated by a rectangular window function. The spectrum distribution characteristics of the rectangular window function are as follows: when the input frequency ω is less than the cutoff frequency ω 0 When X(ω)=1, the signal in the frequency band is retained; on the contrary, when the input frequency ω is greater than or equal to the cutoff frequency ω 0 When , X(ω) = 0, that is, the signal in this frequency band is filtered out. The inverse Fourier transform can be obtained that the time domain expression of this bandpass filter is the Single function:

[0074]

[0075] The response of this bandpass filter in the time domain is a main pulse with a high amplitude, followed by numerous periodic pulses that gradually decay. In this phenomenon, the main pulse with the highest amplitude is called a ringing pulse wave signal. Due to defects in actual hardware, this significant ringing pulse wave signal may leak into the low-frequency audible sound band.

[0076] In order to reduce the influence of ringing pulses, the present invention uses a frequency domain filter to smooth the ultrasonic signal. The present invention applies the Singer function in the frequency domain to the bandpass filter to replace the rectangular window function. The expression of the Singer function is:

[0077]

[0078] According to the reversibility of Fourier transform, the expression of the Singer function bandpass filter in the time domain is similar to that of the rectangular window function. This filter design can effectively smooth the sudden change of signal frequency, thereby eliminating the ringing pulse phenomenon. Although this processing will produce some out-of-band frequency components, the energy of these frequency components outside the passband is low and still within the ultrasonic frequency band. Therefore, this modulation method can avoid the generation of audible noise while not significantly reducing the ultrasonic energy in the passband.

[0079] 2) Splitting the broadband spectrum: The internal amplifier of the ultrasonic speaker will produce nonlinear effects, causing the ultrasonic signal to be demodulated into low-frequency sound. Assume that the center frequency of the ultrasonic signal to be transmitted is f c , bandwidth is 2B. Due to the nonlinear characteristics of the ultrasonic speaker, it will generate a low-frequency sound signal with a center frequency of 0 and a bandwidth of B. This nonlinear effect will turn the ultrasonic signal into audible noise when it propagates in the air.

[0080] In order to solve this problem, the present invention decomposes the ultrasonic signal into multiple narrow bandwidth signals and makes the frequency lower than 50Hz, that is, converts it into infrasound waves that cannot be perceived by the human ear, thereby avoiding the appearance of audible noise. Taking an ultrasonic signal with a bandwidth of 2B as an example, it is divided into k equal parts in the frequency domain and decomposed into k bandwidths of The nonlinear effect will produce a signal with a center frequency of 0 and a bandwidth of By using a larger k, That is, the low-frequency signal generated by the self-demodulation cannot be detected by the human ear. The k-divided ultrasonic signal is transmitted using k different ultrasonic transmitters, and these k signals will be demodulated separately on the eavesdropper's device and pieced together into the original noise signal.

[0081] The above contents are only for explaining the technical idea of ​​the present invention and cannot be used to limit the protection scope of the present invention. Any changes made on the basis of the technical solution in accordance with the technical idea proposed by the present invention shall fall within the protection scope of the claims of the present invention.

Claims

1. An ultrasonic recording jamming method for preventing Chinese speech eavesdropping, characterized in that: The following steps are involved: Obtain a short voice segment of the user to be protected; Use speech generation technology to generate a large amount of user corpus data based on short segments of user speech; Generate the user's phoneme-level corpus by using the phoneme processing algorithm to process the user's corpus data; Through the coupled noise generation algorithm, any number of phonemes are randomly selected from the phoneme-level corpus each time noise is generated to generate coupled noise; The coupled noise is modulated by an ultrasonic carrier wave, and the corrected sound wave is emitted through an ultrasonic speaker using an ultrasonic correction algorithm.

2. The ultrasonic-based recording interference method for preventing Chinese speech eavesdropping according to claim 1 is characterized in that: The phoneme processing algorithm is specifically: 1) Use the maximum inter-class variance method to calculate the strength threshold of the corpus data and cut the user's corpus data into phonemes; 2) clustering the cut phonemes according to vowels, diphthongs and consonants, and treating different tones of the same phoneme as different categories during clustering; 3) The clustered phoneme signals of each category are aligned and averaged using a dynamic time warping algorithm to obtain the average signal of each phoneme of the user as the phoneme library.

3. The ultrasonic-based recording interference method for preventing Chinese speech eavesdropping according to claim 1 is characterized in that: The coupling noise generation algorithm is specifically: 1) Randomly extract phonemes from the user's phoneme-level corpus, that is, randomly extract any number of phonemes at short intervals to generate the user's semantic-free phoneme sequence s(t); 2) Generate a noise signal n1(t) with similar characteristics to the user's voice based on the user's phoneme sequence s(t): Among them, M 1×2 (t) is a 1x2 non-singular random matrix; n r1 (t) is a random noise; L[·] is a non-integrable nonlinear aliasing function: 3) Based on n1(t), a random noise is introduced in the time domain and frequency domain through convolution operation to generate the time-frequency noise signal n2(t): Among them, * is the convolution operation; n ri t)(i=1,2,3) are random noises that are independent of each other and have different distributions; 4) Based on n2(t), the frequency hopping spread spectrum technology is used to broaden its spectrum bandwidth, and finally the coupled noise n3(t) is formed: in, is the frequency hopping signal; n is the nth frequency point in the frequency hopping table; is the phase modulation function, where d is the signal code after error coding, t is the time, and Δω is the frequency difference component before and after frequency hopping.

4. The ultrasonic-based recording interference method for preventing Chinese speech eavesdropping according to claim 1 is characterized in that: The ultrasonic noise correction algorithm is specifically: 1) Use the frequency domain filter to smooth the ultrasonic signal. The expression of the frequency domain filter is: Where ω is the angular frequency; 2) Decompose the smoothed filtered ultrasonic signal into multiple narrow bandwidth signals, and make the frequency of each signal segment lower than 50 Hz. Each of the decomposed signals is transmitted using a different ultrasonic transmitter.

Images