A silent and precise anti-eavesdropping method

Abstract

The present invention discloses a silent and precise anti-eavesdropping method. The method includes: step one, preprocessing stage: first, process the target audio into two audio files with opposite phases, and add flags to the two audio files; send the two audio files in dual channels in sequence through a sending device, and collect audio signals through a receiving device; obtain two preprocessed audio signals; the two preprocessed audio signals are mono audios of different channels respectively; step two, anti-eavesdropping audio construction stage, align the phases of the two preprocessed audio signals, and save them as dual-channel data of the same audio after adaptive adjustment; and exchange the audio data of the two channels; step three, the cancellation stage, send the exchanged dual-channel audio data through a sending device. The present invention uses the principle of sound cancellation to ensure that the cancellation signal strength is the same as the audible sound strength, and compared with the prior art, the overhead is smaller.

Description

Technical Field

[0001] The invention belongs to the technical field of electrical communications, and in particular relates to a silent interference anti-eavesdropping method based on the nonlinear characteristics of a microphone. Background Art

[0002] With the development of science and technology, eavesdropping technology is becoming more and more advanced. Eavesdropping has become a very serious problem, and there are hidden dangers to privacy and security. Once the eavesdropped voice information is leaked, it will not only infringe people's privacy, damage the interests of the company, but also threaten national security. Thus, anti-eavesdropping technology was born. Nowadays, anti-eavesdropping technology has become an indispensable part of life. Anti-eavesdropping technology has a wide range of applications, such as: installing anti-eavesdropping devices on mobile phones, PN junction detectors, and various eavesdropping alarms, etc.

[0003] Anti-eavesdropping technologies can be divided into three categories: electromagnetic interference, audible sound cancellation, and ultrasonic interference. Since the range of sounds that humans can hear is limited, ultrasonic interference can interfere with eavesdropping devices without humans noticing, so ultrasonic interference is more widely used.

[0004] Existing ultrasonic interference technology mainly relies on the nonlinear effect of the amplifier inside the microphone. Ultrasonic waves can leak energy into the sound spectrum and broadcast high-frequency, inaudible strong noise. Due to the nonlinear effect of the microphone, the high-frequency noise is demodulated to the low-frequency human vocal band. The noise power is greater than the sound intensity of human speech, thereby covering the sound.

[0005] The existing ultrasonic interference technology has certain defects, for example: the interference method is too expensive, more costly, and lacks specificity. It is impossible to interfere only with unauthorized devices, and it is easy to prevent authorized devices from recording while interfering. Summary of the invention

[0006] The purpose of the present invention is to overcome the shortcomings and deficiencies of the prior art, such as the common interference method of using Gaussian white noise as an interference signal, weak targeting, and high cost. A silent and precise cancellation anti-eavesdropping interference technology using the nonlinear characteristics of a microphone and the principle of sound cancellation is proposed. The technology uses the sound of two waves with opposite vibration characteristics to weaken and eliminate the interaction of the sound waves to perform precise interference.

[0007] In order to achieve the above purpose, the technical solution adopted by the present invention is: a silent and precise anti-eavesdropping method, comprising:

[0008] Step 1: Preprocessing

[0009] First, the target audio is processed into two audio files with opposite phases, and flags are added to the two audio files; the two audio files are sent in dual channels in sequence by a sending device, the sending device includes a low-frequency speaker and an ultrasonic transducer; the low-frequency speaker and the ultrasonic transducer are respectively connected to different channels; the audio signal is collected by a receiving device to obtain two pre-processed audio signals; the two pre-processed audio signals are respectively mono audios of different channels;

[0010] Step 2: Anti-eavesdropping audio construction stage

[0011] After the two pre-processed audio signals are phase-aligned, they are adaptively adjusted and saved as dual-channel data of the same audio; and the audio data of the two channels are exchanged;

[0012] Step 3: Offset Phase

[0013] The exchanged dual-channel audio data is then sent through the sending device.

[0014] Furthermore, the flag bit is a single audio signal, and there is a blank interval between the flag bit and the target audio data.

[0015] Furthermore, in the preprocessing stage, the flags of the two audio files are sent on the same channel.

[0016] Furthermore, in the preprocessing stage, the data of another audio file is modulated using AM.

[0017] Furthermore, the distance between the sending device and the receiving device is 40cm-100cm.

[0018] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0019] (1) As a microphone nonlinear anti-eavesdropping technology, the present invention proposes a precise anti-eavesdropping method, which is different from the common microphone anti-eavesdropping device playing strong noise interference masking method.

[0020] (2) Lower overhead. The present invention uses the principle of sound cancellation to ensure that the canceling signal strength is the same as the audible sound strength. Compared with the common microphone anti-eavesdropping equipment that uses strong noise interference to cover up the audible sound, the overhead is lower. Under the same power as cancellation, using noise cannot achieve effective interference.

[0021] (3) Strong security. The offset audio was tested and recognized, and the word error rate (WER) reached over 99%. Two audio recovery methods were used to recover the audio after interference and extract the speech content. After the audio was recovered, human recognition was performed and the speech content could not be distinguished.

[0022] (4) Targetedness. The present invention can effectively implement single interference against unauthorized eavesdropping, and will not affect the recording of voice content by authorized recording equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 It is a flowchart of a silent and precise cancellation anti-eavesdropping method in an embodiment of the present invention;

[0024] Figure 2 Schematic diagram of the principle of phase alignment in an embodiment of the present invention;

[0025] Figure 3 The frequency response difference between two different loudspeakers in the embodiment of the present invention;

[0026] Figure 4 The frequency response after the two channels are swapped in the embodiment of the present invention;

[0027] Figure 5 The audio quality evaluation results after accurate cancellation by the method of the present invention are: (a) PESQ of different languages; (a) STOI of different languages; (c) different distances;

[0028] Figure 6 This is an analysis of the cancellation and noise interference effects under equal power. DETAILED DESCRIPTION

[0029] The technical solutions in the embodiments of the present invention will be clearly and completely described below in combination with the embodiments of the present invention and the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. The following description of at least one exemplary embodiment is actually only illustrative and is by no means intended to limit the present invention and its application or use. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0030] The basic idea and technical principle involved in the present invention are mainly based on the nonlinear characteristics of microphones and the principle of sound cancellation.

[0031] 1. Microphone nonlinearity

[0032] Microphone nonlinear characteristics But in practice, microphone amplifiers usually exhibit nonlinearity, and the frequency f of the input signal S s Higher than the maximum receiving frequency f of the microphone c When , the input signal will be affected by nonlinearity, and the output signal is in a nonlinear state. The relationship is expressed as:

[0033]

[0034] Although the nonlinear output is an infinite power series in theory, since the energy of harmonics above the third order is very weak, we only consider the second order terms. At this point, we can express the microphone nonlinearity as:

[0035] S out =A 1 S+A 2 S 2

[0036] Therefore, when a frequency f is played 1 and f 2 When the ultrasonic wave is composed, the signal can be expressed as:

[0037] S=cos(2πf 1 t)+cos(2πf 2 t)

[0038] When the ultrasonic microphone diaphragm amplifier outputs S out We can express it as:

[0039] S out =A 1 (f 1 +f 2 )+A 2 (f 1 +f 2 ) 2

[0040] =A 1 {cos(2πf 1 t)+cos(2πf 2 t)}+A 2 {cos 2 (2πf 1 t)+cos 2 (2πf 2 t)+2cos(2πf 2 t)cos(2πf 2 t)}

[0041] Now, the frequency of the first order term is 2πf 1 and 2πf 2 , which are outside the cutoff point of the microphone. The second-order terms can be expressed as:

[0042]

[0043] Generally, the microphone of a mobile phone will pass through a low-pass filter before being sampled by the ADC. The cut-off frequency of the low-pass filter is generally 24kHz. 1 and f 2All frequencies above 24kHz are filtered out by low-pass filters, except for Cos(2π(f 1 -f 2 )t). When ((f 1 -f 2 )<24kHz, a non-ultrasonic frequency will be obtained, which can be retained when passing through a low-pass filter, that is, the high-frequency inaudible sound frequency emitted by ultrasound will be received by a microphone as audio audible to humans.

[0044] 2. Sound cancellation

[0045] Sound wave cancellation theory and active noise control technology Sound wave cancellation theory is mostly used for active noise control. Its principle is that when two sound waves propagate to the same point in the medium at the same time, if the vibration characteristics generated by the two sound waves at this point are in opposite phases, the vibration here will be weakened or even eliminated due to the interaction of the waves, and the energy here will be attenuated. This phenomenon is called destructive interference of sound waves.

[0046] Based on the above theory, we simultaneously send two sound waves with the same frequency and fixed phase difference, and the sound pressure P generated at a certain position in the sound field is a , P b It is expressed as:

[0047]

[0048]

[0049] in, and is the phase of the two sound waves; ω is the frequency of the sound waves; P 1 and P 2 is the sound pressure amplitude of the sound wave.

[0050] At this time, the superimposed sound pressure of the two sound waves at a certain point is

[0051]

[0052] In the sound field, the average sound energy density is usually To express the energy of sound, the sound energy density of the sound field where the two sound waves overlap is:

[0053]

[0054] Where c is the speed of sound propagation, ρ 0 The density of the medium in which sound waves propagate in the sound field. is the phase difference between the two sound waves

[0055] From the formula (), it can be seen that the synthetic sound energy density at this point in the sound field is mainly related to the phase difference between the two sound waves. When the sound pressure amplitudes of the two sound waves are the same, P 1 =P 2 , phase difference hour, The synthetic sound energy density at this point in the sound field The two sound wave signals have opposite phases, achieving destructive interference of the sound waves.

[0056] Embodiment 1 utilizes the nonlinear characteristics of microphones and the principle of sound cancellation to provide a silent and precise anti-eavesdropping interference cancellation method. Figure 1 and Figure 2 As shown, the method specifically comprises the following steps:

[0057] Step 1, preprocessing stage:

[0058] First, the target audio is divided into two identical audio files, and one of the audio files is inverted. A 1-second single-frequency signal is added to both audio files as a flag, and there is a one-second blank interval between the flag and the target audio data. The flags of the two audio files are sent by the ultrasonic transducer on the right channel (or by the low-frequency speaker on the left channel) to determine the delay difference of the signals sent by the two different devices to ensure phase alignment.

[0059] In this embodiment, two mutually inverted audio files with added flags are sequentially transmitted through a transmitting device, wherein the transmitting device includes a low-frequency speaker and an ultrasonic transducer; the low-frequency speaker is connected to the left channel, and the ultrasonic transducer is connected to the right channel. A dual-channel audio file is transmitted through the transmitting device. The flags can be transmitted through a low-frequency speaker or an ultrasonic transducer, but need to be transmitted on the same channel, for example, both are transmitted on the left channel, or both are transmitted on the right channel.

[0060] As in this embodiment, two mutually inverted audio files with added flags are named Audio1 and Audio2 respectively. Audio1 is played in two channels through a low-frequency speaker, and received by a receiving device to collect the audio signal Audio3 of the left channel (L). Audio2 is first modulated into an ultrasonic signal through AM, and then played in two channels through an ultrasonic transducer, and received by a receiving device to collect the audio signal Audio4 of the right channel (R). Audio3 is a mono audio mixed from the two channels in Audio1, and Audio4 is a mono audio mixed from the two channels in Audio2.

[0061] The receiving device is a device of the same model as the eavesdropping device, such as a microphone, wherein the distance between the receiving device and the transmitting device is 40cm-100cm.

[0062] In this embodiment, the AM modulation method is used to play Audio 2, which can reduce the distortion of the ultrasonic signal collected by the receiving device (microphone).

[0063] Step 2: Anti-eavesdropping audio construction phase:

[0064] The two collected mono audio signals Audio3 and Audio4 are aligned through phase alignment technology, that is, the preset flags are aligned. After the two sound signals are phase-aligned through the flags, they can be saved as a dual-channel audio file: Audio3 (L) and Audio4 (R).

[0065] The above audio files are adaptively adjusted. For example, in order to make the target signal and the cancellation signal have the same volume when they reach the microphone, and just completely cancel each other out, the volume of the two channels is adjusted and low-pass filtered. Then the channels of Audio3 and Audio4 are swapped, that is, Audio3(R) and Audio4(L).

[0066] The principle of phase alignment is as follows: Figure 2 As shown, the flag bit is sent through an ultrasonic transducer as an example: since the flag bit audio and the target audio of Audio2 are both sent by the right channel (R) of the ultrasonic transducer, the blank interval time t does not change, that is, the ultrasonic transducer sends the target audio with a delay of time t; the blank interval between the flag bit audio and the target audio of Audio4 collected by the receiving device is also t.

[0067] The audio data of Audio1 is sent through the right channel (R) of the ultrasonic transducer due to the flag bit, and the target audio is sent by the low-frequency speaker. Since they are not the same sending device, the distances between different sending devices and receiving devices are often different, causing the blank interval time of t between the received Audio3 flag bit and the target audio data to change, which is recorded as t1, t=t1±Δt, Δt is the change, which is expressed as a delay increment or an advance increment.

[0068] After Audio3 and Audio4 are aligned through the flag bit, they are saved as a dual-channel file and the channels are exchanged. When sending, Audio3 is sent through the ultrasonic transducer connected to the right channel, and Audio4 is sent through the low-frequency speaker connected to the left channel. At this time, two channels of audio data are played at the same time, and the two channels of audio data reach the eavesdropping microphone at the same time, and the phases are opposite, thereby achieving a cancellation effect.

[0069] Step 3, interference cancellation stage:

[0070] The dual-channel audio files Audio3 and Audio4 after the channels are swapped are sent simultaneously through the sending device. Specifically, the audio of Audio3 is played through the ultrasonic transducer, and the audio of Audio4 is played through the low-frequency speaker. Since Audio3 and Audio4 are in anti-phase with each other and reach the eavesdropping device at the same time, precise cancellation and interference effects can be achieved.

[0071] The transmitting device of the whole system is composed of a low-frequency speaker and an ultrasonic transducer, which separates the left and right channels. The left channel is connected to the low-frequency speaker to play audible sound, and the right channel is connected to the RIGOL DG5072 signal transmitter to modulate the signal into ultrasonic waves, and then connected to the ultrasonic transducer to send non-sound wave signals. The equipment layout in the interference cancellation stage must be absolutely consistent with the position and distance of the low-frequency speaker, ultrasonic transducer, and microphone receiving device in the pre-processing stage to achieve effective and precise interference.

[0072] Example 2 The present invention uses a chirp signal of 50-5000 Hz as the test audio to test the frequency response difference of two different playback devices and confirms the feasibility of the method of the present invention.

[0073] Frequency response difference: The sound playback and reception are composed of speakers and microphones. Due to the characteristics and structure of the hardware materials themselves, different devices have different response capabilities to the same frequency. Therefore, there are differences in the audio collected by using different devices to play the same audio. For example, Figure 3 Shown is the difference in frequency response between an ultrasonic sensor and a low-frequency speaker.

[0074] The present invention found that when the sound passes through two speakers with different frequency responses in sequence, the obtained audio waveform is consistent, regardless of the order. Therefore, the present invention adopts the method of exchanging channels to reduce the frequency response difference. After aligning the two channels of audio collected in the preprocessing stage, the two channels are exchanged and played again after the exchange. The audible audio is played through the ultrasonic sensor, and the ultrasonic audio is played through the low-frequency speaker to solve the frequency response difference. That is, the waveforms of "audio-low-frequency speaker-ultrasonic sensor-collection" and "audio-ultrasonic sensor-low-frequency speaker-collection" are equal, such as Figure 4 shown.

[0075] Example 3 In order to evaluate the interference effect of the method of the present invention, the present invention conducts experimental tests from the aspects of the influence of different languages, the influence of distance, the influence of equipment, and the audio anti-restoration effect.

[0076] Experimental environment: All experiments were conducted in an office with an ambient noise level of approximately 35dB, and the loudness of the audible speech audio obtained by the receiving device was guaranteed to be no less than 75db.

[0077] The present invention evaluates the audio quality after precise cancellation, i.e., intelligibility, using the following indicators: Perceptual Evaluation of Sound Quality (PESQ), Short-Time Objective Intelligibility (STOI), and Word Error Rate (WER).

[0078] PESQ is standardized in P.862 and is widely used for objective speech quality assessment. Technically, the input consists of a clear speech signal as a reference and a signal to be measured. The output is a mean opinion score (MOS)

[28] ranging from -0.5 to 4.5. A high PESQ score means that the corresponding speech has a high listening quality, and vice versa. Generally, PESQ values ​​from 1.00 to 1.99 mean "there is no feasible way to understand it".

[0079] WER is the percentage of the number of wrong words divided by the total number of words in the standard word sequence. The present invention adopts human recognition and machine recognition for evaluation. Invite 10 volunteers to listen to the undisturbed recorded audio and the disturbed audio, and count the words heard.

[0080] Short-term objective intelligibility (STOI) is an important indicator for measuring speech intelligibility. For a word in a speech signal, there are only two situations: it can be understood or it cannot be understood. In this invention, the STOI value of 0.4 is taken as the baseline. When the STOI value is below 0.4, it is considered unintelligible.

[0081] (1) Influence of language

[0082] Four languages, namely Chinese, English, Italian and Japanese, were used for the test, with male and female voices in each language. Figure 5 As shown in (a)h and (b), the PESQ and STOI of different languages ​​and genders are not much different, indicating that different languages ​​and genders have little impact on the interference effect.

[0083] (2) Effect of distance

[0084] The recording devices were placed at 40cm, 60cm, 80cm, and 100cm away from the playback device to collect sound signals.

[0085] The results are as follows Figure 5 As shown in (c), within the distance range of 40cm-100cm, PESQ and STOI remain consistent, indicating that the interference effect is stable within this distance range.

[0086] (3) Impact of different devices

[0087] The present invention uses four smartphones, namely iPhone10 X, iPhone13pro, iPhone13promax, and Redmi k30ultra, as eavesdropping devices for testing, and uses two recovery methods: 1) DPCRN convolutional recursive network voice enhancement; 2) using the SOX tool to extract noise features for noise elimination, simulating the situation where an opponent attempts to use these methods to recover information from unauthorized recordings, to prove the security of the method of the present invention.

[0088] Table 1 Effect of different equipment on the offset effect and anti-recovery effect of the method of the present invention

[0089]

[0090] It can be concluded from the data in Table 1 that the method of the present invention has an excellent interference effect, and the PESQ is stable between 1.02 and 1.5, which means that the audio after cancellation has enough speech distortion and there is no feasible way to understand it. WER also accepts this conclusion and reaches 100% in machine recognition and human recognition, which means that the opponent cannot extract anything from the recording.

[0091] Example 4 Analysis of the cancellation and noise interference effects of the present invention at the same power

[0092] Since it is impossible to control the intensity of two sounds to be absolutely the same in actual applications, the present invention adds 30% of the noise of the cancellation signal power to the original cancellation signal to further ensure the cancellation effect and safety of the present invention. At the same time, the interference effects of cancellation and pure noise at the same power, and the interference effects of cancellation + noise and noise of the same power are compared.

[0093] Depend on Figure 6 It can be seen that cancellation has lower PESQ and STOI. After adding 20% ​​power noise, the PESQ and STOI values ​​of cancellation + noise have a certain degree of reduction, which proves that the interference effect can be effectively improved. The audio after pure noise interference of two different powers has higher PESQ and STOI, indicating that the use of pure noise interference at this power is sufficiently intelligible and is not enough to cover up the voice content to achieve an effective interference effect.

Claims

1. A silent and precise anti-eavesdropping method, characterized in that: include: Step 1: Preprocessing First, the target audio is processed into two audio files with opposite phases, and flags are added to the two audio files; the two audio files are sent in dual channels in sequence by a sending device, the sending device includes a low-frequency speaker and an ultrasonic transducer; the low-frequency speaker and the ultrasonic transducer are respectively connected to different channels; the audio signal is collected by a receiving device to obtain two pre-processed audio signals; the two pre-processed audio signals are respectively mono audios of different channels; Step 2: Anti-eavesdropping audio construction stage After the two pre-processed audio signals are phase-aligned, they are adaptively adjusted and saved as dual-channel data of the same audio; And exchange the audio data of the two channels; Step 3: Offset Phase The exchanged dual-channel audio data is then sent through the sending device.

2. The anti-eavesdropping method of silent and precise cancellation according to claim 1 is characterized in that: The flag bit is a single audio signal, and there is a blank interval between the flag bit and the target audio data.

3. The anti-eavesdropping method of silent and precise cancellation according to claim 1 is characterized in that: In the preprocessing stage, the flags of the two audio files are sent on the same channel.

4. The anti-eavesdropping method of silent and precise cancellation according to claim 1 is characterized in that: In the pre-processing stage, the data of another audio file is modulated using AM.

5. The anti-eavesdropping method of silent and precise cancellation according to claim 1 is characterized in that: The distance between the sending device and the receiving device is 40cm-100cm.

Images