Recording interference prevention method and device

By monitoring and identifying recording devices in real time, and using acoustic parametric arrays and beamforming technology for directional transmission, targeted interference strategies are generated. This solves the problems of insufficient identification and inaccurate interference in existing ultrasonic anti-recording technologies, and achieves efficient and adaptive anti-recording protection.

CN121711059APending Publication Date: 2026-03-20GUANGZHOU WO YUN ELECTRONIC PROD CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202511817904.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-12-04
Publication Date
2026-03-20

AI Technical Summary

Technical Problem

Existing ultrasonic anti-recording technologies lack the ability to identify recording devices and the specificity of interference strategies, resulting in unstable protection effects when facing diverse recording devices. Furthermore, traditional interference methods are inefficient and may cause discomfort to people in the room, making it difficult to ensure interference effectiveness while avoiding disruption to normal conversations.

Method used

By monitoring sound field changes in real time through an auxiliary sensor array, identifying the type and model of recording equipment, generating targeted interference strategies, using acoustic parametric arrays and beamforming technology for directional transmission, and combining multiple dynamic sequences and composite baseband signals to achieve precise and adaptive anti-recording protection.

Benefits of technology

It achieves accurate identification and targeted interference of recording devices, improves the reliability and applicability of anti-recording, reduces the impact on normal voice communication, and enhances the unpredictability and anti-filtering ability of interference signals.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN121711059A_ABST
    Figure CN121711059A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of voice information security, and discloses a recording interference prevention method and device, and the method comprises an environment perception step, an equipment recognition and positioning step, an interference strategy decision-making step, a dynamic sequence generation step, a signal fusion construction step, a parametric array modulation step, a directional emission step, and an evaluation adjustment step. The device corresponds to the method. According to the application, different recording devices are accurately distinguished through an acoustic fingerprint identification technology, and a targeted interference strategy can be generated according to specific acoustic features of the devices; by combining an acoustic parametric array and a beam forming technology, accurate directional transmission of interference sound waves is realized, interference energy is enabled to act on target equipment in a concentrated manner, and the interference efficiency is improved; through fusion of multiple dynamic sequences and construction of composite baseband signals, the unpredictability and anti-filtering capability of interference signals are enhanced. Therefore, the high-efficiency interference performance is maintained, the influence on normal voice communication is reduced, and accurate and self-adaptive anti-recording protection is realized.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of voice information security technology, specifically a method and device for preventing recording interference. Background Technology

[0002] Voice information security is directly related to the protection of state secrets and trade secrets. Current mainstream anti-recording technologies mainly include sound interference, soundproofing, electromagnetic shielding, frequency diffusion, and encrypted communication. Among these, ultrasonic-based sound interference technology has attracted attention due to its relatively simple implementation.

[0003] However, existing ultrasonic anti-recording technologies have significant limitations. Traditional ultrasonic jamming methods mostly employ fixed jamming frequencies and patterns, and this "one-size-fits-all" approach lacks specificity. Because different recording devices vary significantly in hardware characteristics, acoustic response, and noise reduction algorithms, fixed jamming patterns are ineffective against intelligent recording devices employing advanced noise reduction algorithms. Furthermore, the broad range of jamming signal transmissions is not only inefficient but may also cause discomfort to people indoors.

[0004] More importantly, existing technologies lack the ability to identify recording devices and the specificity of their interference strategies. The system cannot distinguish between different types of recording devices, nor can it adjust its interference strategy based on the device's acoustic characteristics, resulting in inconsistent protection against a wide variety of recording devices. Furthermore, traditional ultrasonic jamming methods are insufficient in terms of precision and concealment, making it difficult to ensure effective interference while avoiding disruption to normal conversation.

[0005] Therefore, there is an urgent need for a more reliable technology to provide effective and accurate continuous protection for recording equipment. Summary of the Invention

[0006] The purpose of this invention is to provide a method and apparatus for preventing recording interference, so as to solve the technical problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention discloses the following technical solutions: In a first aspect, the present invention discloses a method for preventing recording interference, the method comprising: Environmental perception steps: Real-time monitoring of sound field changes in the target protection area using an auxiliary sensor array, and collection of ambient background noise; Device identification and positioning steps: When an abnormal sound signal is detected, the acoustic fingerprint feature of the sound signal is extracted and matched with a preset recording device voiceprint feature database to identify the type and model of the recording device and determine its spatial location. Interference strategy decision-making steps: Based on the identified recording device model, retrieve a targeted combination of interference parameters from a pre-set interference strategy library; the interference strategy library records the vulnerability characteristics of different recording devices against various interference signals; Dynamic sequence generation step: Generate at least two dynamic sequences, including a first random sequence and a second environment adaptive sequence, wherein the second environment adaptive sequence is reconstructed based on the noise characteristics of the background noise of the environment collected in the environment perception step; Signal fusion construction steps: The first random sequence and the second environmental adaptive sequence are fused according to the interference parameter combination to construct a composite baseband signal; Parametric array modulation step: The composite baseband signal is modulated onto a pair of ultrasonic carriers with a fixed frequency difference to generate an ultrasonic transmission signal with two original ultrasonic signals. Directional transmission step: Based on the spatial orientation of the recording device, the ultrasonic transmission signal is directionally focused and transmitted to the spatial area where the recording device is located using an ultrasonic transmitter array and beamforming technology; wherein, the two original ultrasonic signals propagate in the air and converge in the spatial area, and after nonlinear interaction based on the principle of acoustic parametric array, they self-demodulate to generate a highly directional audible frequency band interference sound field, which acts on the recording device; Evaluation and adjustment steps: Evaluate the interference effect of the audible frequency band interference sound field on the recording device in real time, and dynamically adjust any one of the interference parameter combination, the modulation mode of the composite baseband signal, or the transmission power of the ultrasonic transmission signal based on the evaluation results.

[0008] In one optional implementation, the extraction of the acoustic fingerprint features in the device identification and positioning step includes: A linear frequency modulated detection signal with a frequency of 18kHz-22kHz is transmitted toward the target protection area; The microphone of the recording device located within the target protection area receives the linear frequency modulation (LFM) probe signal and the intermodulation distortion products generated due to its nonlinear characteristics. The amplitude ratio of the second harmonic to the third harmonic of the LFM probe signal, and the frequency of the LFM probe signal are extracted. The energy of the intermodulation distortion components, of which , and These are two different frequency components within the linear frequency modulated detection signal; The extracted amplitude ratio and the energy composition feature vector of the intermodulation distortion component are input into a convolutional neural network model pre-trained by transfer learning to complete the classification and identification of the device model.

[0009] In one optional implementation, the method for constructing and invoking the interference strategy library in the interference strategy decision-making step includes: A policy function dynamically updated based on a reinforcement learning model is used as the combination of interference parameters for each type of recording device in the interference policy library; wherein, the state space of the policy function includes the environmental signal-to-noise ratio, device azimuth angle and recognition confidence, and the action space is the joint adjustment of the complexity parameter of the first random sequence and the mixed weight of the second environmental adaptive sequence; By training in online and simulated environments, the reinforcement learning model converges to the optimal strategy that maximizes the evaluation function of the interference effect.

[0010] In one optional implementation, the method for constructing and invoking the interference strategy library further includes: When constructing the policy function, an adversarial neural network is used to simulate the noise reduction algorithm of the recording device; By alternating training of the reinforcement learning model and the adversarial network, the policy function is optimized to generate interference signal parameters that can deceive or penetrate the noise reduction algorithm.

[0011] In one optional embodiment, the method for generating the second environment-adaptive sequence in the dynamic sequence generation step includes: Empirical mode decomposition is performed on the collected environmental background noise to obtain a set of intrinsic mode functions; Calculate the coherence coefficient between each intrinsic mode function and the currently generated first random sequence in the time-frequency domain; The K intrinsic mode function components with the lowest coherence coefficients are selected and recombined to generate the second environment adaptive sequence that is statistically orthogonal to the first random sequence.

[0012] In one alternative embodiment, the original ultrasonic signal is generated by simultaneously modulating the composite baseband signal onto a pair of ultrasonic carriers with frequencies of 40 kHz and 40.5 kHz.

[0013] In one alternative implementation, the two original ultrasonic signals converge in the spatial region by using an FPGA-based programmable delay chip to independently control the transmission delay and phase of each element in the ultrasonic transmitter array for the two ultrasonic signals.

[0014] In one alternative implementation, the method further includes: Acoustic stealth post-processing steps: After the directional emission step, a canceling sound wave is emitted through an independent auxiliary loudspeaker. This sound wave has the same amplitude and opposite phase to the target interference sound wave constituting the audible frequency band interference sound field. The canceling sound wave is adjusted by an adaptive filter so that it cancels out the target interference sound wave in a protected area other than the spatial orientation of the recording device.

[0015] In one alternative implementation, the method further includes: Signal fingerprint encryption step: Before the parametric array modulation step, the acoustic channel impulse response from the ultrasonic transmitter array to the recording device is estimated, and the characteristics of the acoustic channel response are used as a dynamic key to encrypt the seed value of the first random sequence or the modulation method of the composite baseband signal.

[0016] Secondly, this application discloses an apparatus for applying the anti-recording interference method described above, the apparatus comprising: The environmental perception module is configured to: monitor the sound field changes of the target protection area in real time through an auxiliary sensor array, and collect the ambient background noise; The device identification and positioning module is configured to: when an abnormal sound signal is detected, extract the acoustic fingerprint features of the sound signal, match it with a preset recording device voiceprint feature database, in order to identify the type and model of the recording device and determine its spatial location; The interference strategy decision module is configured to: based on the identified recording device model, call a targeted combination of interference parameters from a preset interference strategy library; the interference strategy library records the vulnerability characteristics of different recording devices against various interference signals; The dynamic sequence generation module is configured to generate at least two dynamic sequences, including a first random sequence and a second environment adaptive sequence, wherein the second environment adaptive sequence is reconstructed based on the noise characteristics of the background noise of the environment collected in the environment perception step; The signal fusion construction module is configured to fuse the first random sequence and the second environmental adaptive sequence according to the interference parameter combination to construct a composite baseband signal; The parametric array modulation module is configured to modulate the composite baseband signal onto a pair of ultrasonic carriers with a fixed frequency difference to generate an ultrasonic transmission signal with two original ultrasonic signals. The directional transmission module is configured to: based on the spatial orientation of the recording device, use an ultrasonic transmitter array and beamforming technology to directionally focus and transmit the ultrasonic transmission signal to the spatial area where the recording device is located; wherein, the two original ultrasonic signals propagate in the air and converge in the spatial area, and after nonlinear interaction based on the principle of acoustic parametric array, they self-demodulate to generate a highly directional audible frequency band interference sound field, which acts on the recording device; The evaluation and adjustment module is configured to: evaluate the interference effect of the audible frequency band interference sound field on the recording device in real time, and dynamically adjust any one of the interference parameter combination, the modulation mode of the composite baseband signal, or the transmission power of the ultrasonic transmission signal based on the evaluation results.

[0017] Beneficial Effects: The anti-recording interference method and device of the present invention accurately distinguishes different recording devices through acoustic fingerprint recognition technology, and can generate targeted interference strategies based on the specific acoustic characteristics of the devices; it adopts a method combining acoustic parametric array and beamforming technology to achieve precise directional transmission of interference sound waves, so that the interference energy is concentrated on the target device, improving interference efficiency; through the fusion of multiple dynamic sequences and the construction of composite baseband signals, it enhances the unpredictability and anti-filtering ability of the interference signal. Thus, while maintaining high-efficiency interference performance, it reduces the impact on normal voice communication, achieving precise and adaptive anti-recording protection. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 A flowchart illustrating the method for preventing audio recording interference provided in an embodiment of this application; Figure 2 This is a structural block diagram of the anti-recording interference device provided in the embodiments of this application. Detailed Implementation

[0020] To facilitate understanding of the technical solutions provided in the embodiments of this application, the background technology involved in the embodiments of this application will be described below.

[0021] Voice information security, as a crucial aspect of confidentiality, is directly related to national security and commercial interests. Preventing unauthorized recording of voice messages in sensitive scenarios such as important meetings and business negotiations has become an urgent need. Currently, the main anti-recording technologies include sound interference, soundproofing, electromagnetic shielding, frequency diffusion, and encrypted communication. Among these, ultrasonic-based sound interference technology has attracted much attention due to its flexible deployment and lack of disruption to normal conversation; however, existing ultrasonic anti-recording solutions still have significant shortcomings in practical applications.

[0022] Specifically, the existing technology has the following core flaws: The interference strategy is singular: Traditional ultrasonic interference often uses interference signals with fixed frequencies or preset patterns, lacking specificity for the acoustic characteristics of different recording devices. Since different brands and models of recording devices vary significantly in microphone sensitivity, frequency response, and noise reduction algorithms, a uniform interference pattern is insufficient to effectively address the diverse range of recording equipment. Insufficient interference accuracy: Existing solutions generally use wide beam transmission, resulting in an excessively large interference signal coverage area. This not only reduces the interference energy density and leads to limited interference effect, but may also cause unnecessary noise pollution to non-target areas. Weak anti-filtering capability: With the continuous upgrading of noise reduction algorithms in intelligent recording devices, fixed interference patterns are easily identified and filtered. In particular, when facing advanced noise reduction algorithms based on deep learning, the difficulty of cracking traditional interference signals has increased significantly.

[0023] To address the aforementioned issues, this application provides a method and apparatus for preventing audio recording interference. By constructing an intelligent system for preventing audio recording interference, it achieves accurate identification and targeted interference of recording devices, effectively improving the reliability and applicability of the anti-recording measures.

[0024] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present application. Secondly, in this document, the term "comprising" is intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements, but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0025] This embodiment provides a method for preventing recording interference in the first aspect, such as Figure 1 As shown, the method sequentially includes an environmental perception step, a device identification and localization step, an interference strategy decision step, a dynamic sequence generation step, a signal fusion construction step, a parametric array modulation step, a directional transmission step, and an evaluation and adjustment step, as detailed below.

[0026] The environmental perception step specifically includes: real-time monitoring of sound field changes in the target protected area using an auxiliary sensor array, and collecting ambient background noise. This step utilizes a high-sensitivity sensor array to capture sound field changes and background noise characteristics in real time, avoiding blind spots inherent to single sensors. On one hand, it can quickly identify abnormal sound signals, reducing misjudgments of inherent environmental noise (such as air conditioner noise and footsteps); on the other hand, the collected background noise characteristics provide a basis for the subsequent second environmental adaptive sequence, preventing interference signals from overlapping with environmental noise and causing interference failure, thus improving the environmental adaptability of interference signals.

[0027] In one specific implementation example, the auxiliary sensor array consists of four omnidirectional acoustic sensors, model Sennheiser MKH 416, evenly distributed at the four corners of a 10m × 10m target protection area (such as a conference room). Secondly, the ambient background noise is collected for a default duration of 10 seconds, recording characteristics such as the noise power spectrum distribution and time-domain peak values. For example, the typical ambient noise level in a conference room is 35dB(A).

[0028] The device identification and positioning steps specifically include: when an abnormal acoustic signal is detected, extracting the acoustic fingerprint features of the signal and matching it with a pre-set database of recording device acoustic fingerprint features to identify the type and model of the recording device and determine its spatial location. This step utilizes the unique acoustic fingerprint of the recording device (such as microphone nonlinear distortion characteristics and circuit noise characteristics during operation) for matching with the database, combined with array time delay difference positioning. This accurately distinguishes recording devices from non-recording devices (such as ordinary speakers and mobile phone call modes), avoiding indiscriminate interference; and improves spatial orientation, providing a precise target location for subsequent directional transmission, reducing energy waste of interference signals, and improving interference efficiency.

[0029] In a specific implementation example, the determination of abnormal acoustic signals can be achieved by comparing the collected sound field changes in the target protected area with a preset abnormality threshold (e.g., 5 dB). If the threshold is exceeded, an abnormal acoustic signal is determined to have occurred. The extracted acoustic fingerprint features can include signal spectral envelope, time-domain impulse response, etc. The recording device voiceprint feature database contains voiceprint samples of common recording device types, such as mobile phones, voice recorders, and mini recorders. Each sample contains voiceprint data under three different operating modes. The spatial location of the recording device can be determined using a time-delay difference positioning method with an auxiliary sensor array. The positioning accuracy error is ≤0.3m, and the three-dimensional coordinates of the recording device within the target protected area are output.

[0030] The interference strategy decision-making steps specifically include: Based on the identified recording device model, retrieving a targeted combination of interference parameters from a pre-set interference strategy library; this library records the vulnerability characteristics of different recording devices against various interference signals. This step, by retrieving targeted parameters based on the vulnerability characteristics of the recording device, avoids the inefficiency of general interference parameters. Parameters are matched to the sensitive frequency bands and anti-interference capabilities of different devices (e.g., strengthening interference in the 1kHz-3kHz sensitive frequency band of a recorder, and increasing sequence complexity for the wideband anti-interference characteristics of a mobile phone), ensuring that the interference signal accurately hits the device's weaknesses, improving interference efficiency while reducing unnecessary power consumption.

[0031] In a specific implementation example, the interference strategy library stores interference parameter combinations categorized by device model. For example, the parameter combination for the iPhone 15 is "first random sequence complexity parameter: Lyapunov exponent (a numerical characteristic representing the average exponential divergence rate of adjacent trajectories in phase space) 0.8, second environmental adaptive sequence mixing weight 0.6". It also records the vulnerability characteristics of each device to frequency sweep interference and random noise interference. For example, the Sony ICD-UX570 is more sensitive to interference in the 1kHz-3kHz frequency band.

[0032] The dynamic sequence generation step specifically includes: generating at least two dynamic sequences, including a first random sequence and a second environment-adaptive sequence. The second environment-adaptive sequence is reconstructed based on the noise characteristics of the ambient background noise collected in the environment perception step. This step employs a dual-sequence design of a first random sequence (strong randomness) and a second environment-adaptive sequence (adapting to the environment), and utilizes the background noise characteristics to reconstruct the second environment-adaptive sequence. The statistical characteristics of the two sequences are complementary, enabling coverage of a wider time-frequency domain and avoiding frequency gaps in interference signals (such as a lack of coverage in a certain frequency band allowing recording equipment to capture human voices). Simultaneously, the second environment-adaptive sequence incorporates ambient noise, making interference signals more difficult to filter in the environment and improving resistance to environmental interference.

[0033] In a specific implementation example, the first random sequence uses a 32-bit M-sequence as the base sequence, with the seed value initially set to 0x12345678. The sequence length is 2048 bits, exhibiting good randomness. The noise characteristics of the ambient background noise can include, for example, the proportion of low-frequency components and the periodic fluctuations in the time domain.

[0034] The signal fusion construction steps specifically include: fusing a first random sequence and a second environmentally adaptive sequence according to interference parameter combinations to construct a composite baseband signal. This step, by fusing sequences according to interference parameter combinations, ensures that the fused composite baseband signal carries targeted interference characteristics. Furthermore, the composite baseband signal simultaneously possesses randomness (anti-noise reduction) and environmental adaptability (anti-environmental filtering), accurately matching the vulnerability characteristics of devices. For example, for devices sensitive to mixed signals, the fused signal can trigger recording distortion more quickly.

[0035] In one specific implementation example, the first random sequence and the second environment-adaptive sequence are linearly superimposed and fused, with a mixing weight of 0.6:0.4 during fusion. The bandwidth of the composite baseband signal is adapted to audible frequency band interference requirements.

[0036] The parametric array modulation step specifically includes: modulating the composite baseband signal onto a pair of ultrasonic carriers with a fixed frequency difference to generate an ultrasonic transmission signal with two original ultrasonic signals. In this step, the baseband signal is modulated onto the ultrasonic carriers (frequency higher than 20kHz, inaudible to the human ear), utilizing the fixed frequency difference to prepare for subsequent self-demodulation. The modulated ultrasonic transmission signal does not affect human hearing during propagation (avoiding premature exposure of interference behavior), while the fixed frequency difference ensures that subsequent self-demodulation can generate interference signals in the target's audible frequency band, laying the foundation for directional interference.

[0037] In one specific implementation example, the modulation of the composite baseband signal is achieved by amplitude modulation (AM), and the frequency difference between a pair of ultrasonic carriers is preset to 0.5kHz-2kHz to adapt to the coverage of the audible frequency band.

[0038] The directional transmission step specifically includes: based on the spatial orientation of the recording equipment, using an ultrasonic transmitter array and beamforming technology, the ultrasonic transmission signal is directionally focused and transmitted to the spatial area where the recording equipment is located. Specifically, the two original ultrasonic signals propagate in the air and converge in the spatial area. Based on the nonlinear interaction principle of the acoustic parametric array (i.e., the main lobes of the two ultrasonic signals form constructive interference when they converge, generating a local high sound pressure zone, where the air will produce nonlinear acoustic effects (such as harmonic distortion and difference frequency generation)), a highly directional audible frequency interference sound field is generated through self-demodulation and acts on the recording equipment. This step uses beamforming technology to focus the signal, combined with the nonlinear effect of the acoustic parametric array (the ultrasonic carrier generates audible sound through difference frequency generation in the high sound pressure zone). It generates an audible frequency interference sound field with a high directivity angle (e.g., ≤15°), which only acts on the area where the recording equipment is located, preventing the interference from spreading to other locations in the protected area; and the strong directivity can make the interference sound pressure level ≥100dB in the target area, which can effectively suppress human voice signals (usually human voice sound pressure level 60dB-70dB), causing the signal collected by the recording equipment to be mainly interference, and unable to reproduce human voice.

[0039] In one specific implementation example, the ultrasonic transmitter array consists of eight ultrasonic transducers. The beamforming technique employed is delay-sum beamforming.

[0040] The evaluation and adjustment steps specifically include: real-time evaluation of the interference effect of the audible frequency band interference field on the recording equipment, and dynamic adjustment of any one of the following based on the evaluation results: the combination of interference parameters, the modulation method of the composite baseband signal, or the transmission power of the ultrasonic transmission signal. This step, through real-time evaluation and dynamic adjustment capabilities, can adaptively optimize the interference strategy based on interference effect feedback and environmental changes, thereby improving interference effectiveness and reliability.

[0041] In one alternative implementation, the method for real-time evaluation of the interference effect includes at least one of the following: Method 1: Using a reference microphone array (consisting of a linear array of 6 microphones) deployed in the target area, acquire the interfered sound field signal and calculate its normalized mean square error compared to the expected interference signal. The calculation formula is as follows: ,in The expected time-domain waveform of the interference signal. For the actual sound field signal time-domain waveform collected, when If the value is 0.2, the interference is considered valid; otherwise, it is invalid. The value is used as an indicator of interference effectiveness. Method 2: Transmit a low-frequency detection signal to the recording device, and collect the nonlinear acoustic response characteristics returned by the recording device (such as the amplitude changes of the second and third harmonics in the response signal) through an auxiliary sensor array. Use the change in characteristics as an indirect evaluation index of the interference effect.

[0042] Furthermore, methods for dynamically adjusting parameters based on evaluation results include: Establish based on interference effectiveness index ( A control model with the optimization objective of minimizing the emission power and sequence complexity parameters (variable range Lyapunov exponent 0.5-1.2) is constructed using a particle swarm optimization algorithm to build a nonlinear model with an iteration step size of 0.001. When the interference effectiveness index is not lower than the preset threshold ( When the gradient descent algorithm or reinforcement learning strategy is initiated, the transmission power and / or sequence complexity parameters are optimized in the control model. The default number of iterations is 100, and the convergence condition is that the parameter change between two adjacent iterations is less than 0.01. The optimized parameters are then fed back to the interference strategy decision step and the directional transmission step.

[0043] Therefore, the two evaluation methods (direct measurement of sound field error and indirect measurement of equipment response) are combined with the control model optimization to form a closed-loop adjustment, thereby monitoring the interference effect in real time. When environmental changes (such as the background noise rising to 45dB(A)) or equipment movement cause interference, the system can adjust the system accordingly. At the same time, it can quickly optimize parameters through algorithms (such as increasing the transmit power from 10W to 15W and adjusting the sequence Lyapunov exponent from 0.8 to 1.0) to ensure that the interference effectiveness is always maintained within the effective range, avoid interference effect decay, and achieve continuous and stable interference.

[0044] In one optional implementation, the extraction of acoustic fingerprint features specifically includes: A linear frequency modulated detection signal with a frequency of 18kHz-22kHz is transmitted to the target protection area (with 20kHz as the center frequency, a frequency modulation bandwidth of 2kHz, a duration of 100ms, and a transmission sound pressure level of 70dB, using one sensor from the auxiliary sensor array as the transmitter). The microphone of the recording equipment located within the target protection area receives the linear frequency modulated (LFM) probe signal and generates intermodulation distortion products due to its nonlinear characteristics (the microphone's nonlinearity causes the input signal to generate harmonics and intermodulation components; for example, a 20kHz signal will generate 40kHz (2nd harmonic) and 60kHz (3rd harmonic) components, as well as intermodulation signals of different frequency components). The amplitude of the second harmonic of the LFM probe signal (such as the peak amplitude of the 40kHz component) is extracted. ) and the third harmonic amplitude (such as the peak amplitude of the 60kHz component) The amplitude ratio (calculated as follows) ), and frequency of The energy of the intermodulation distortion components, of which , and For example, taking two different frequency components within a linear frequency modulated probe signal. , The energy is obtained by calculating the sum of the signal power within ±50Hz of that frequency point; The extracted amplitude ratio and the energy of the intermodulation distortion components are combined to form a feature vector (2×1 dimension, vector form [R, energy]), which is then input into a convolutional neural network model (CNN model, the base model is ResNet50, the pre-training dataset is ImageNet, and after transfer learning, it is fine-tuned with 5000 sets of harmonic and intermodulation feature data from different recording devices) to complete the classification and recognition of device models.

[0045] This approach leverages the nonlinear characteristics of microphones in recording devices (all electronic microphones generate harmonic and intermodulation distortion products of the input signal due to the nonlinearity of their internal components (such as diaphragms and amplifiers), and different microphone models exhibit unique differences in their degree of nonlinearity and distortion characteristics). By emitting a specific linear frequency modulated (LFM) probe signal to trigger distortion, unique harmonic amplitude ratios and intermodulation energy characteristics are extracted, and then a pre-trained CNN model is used for accurate classification. Compared to traditional spectral envelope features, microphone nonlinear distortion features have stronger device uniqueness (different microphone models have different diaphragm materials and circuit designs, resulting in significant differences in distortion products). Combined with a pre-trained CNN model, it can effectively distinguish recording devices with similar appearances but different models (such as the microphones of iPhone 13 and iPhone 16), avoiding subsequent interference parameter mismatches due to model misjudgment and improving interference accuracy. Secondly, a linear frequency modulated (LFM) probe signal of 18kHz-22kHz is selected (close to the upper limit of human hearing, and the microphone of recording equipment usually covers this frequency band). This avoids the probe signal being detected by the human ear (reducing exposure to interference) while ensuring that the microphone of the recording equipment can receive it stably and produce distortion products, thus preventing the inability to identify the signal due to the probe signal frequency exceeding the equipment's receiving range. In addition, the duration of the LFM probe signal can be set to 100ms, the time for feature extraction (calculating amplitude ratio and intermodulation energy) can be controlled within 50ms, and the time for CNN model inference can be controlled within 100ms. The overall recognition time is relatively short, enabling rapid device recognition and buying time for subsequent interference actions, preventing the recording equipment from acquiring a valid human voice signal during the recognition process.

[0046] In one optional implementation, the method for constructing and invoking the interference strategy library specifically includes: A dynamically updated policy function based on a reinforcement learning model (using a Deep Q-Network (DQN), with a network structure of "input layer (3 neurons, corresponding to state space parameters) - hidden layer (2 layers, 64 neurons per layer, activation function ReLU) - output layer (2 neurons, corresponding to action space parameters)") is used as the interference parameter combination for each type of recording device in the interference policy library; wherein, the state space of the policy function includes the environmental signal-to-noise ratio (calculated as...). , For human voice signal power, The parameters are: ambient noise power (range 0-30dB), device azimuth angle (angle between the line connecting the recording device and the ultrasonic transmitter array and the array normal, range 0-360°, accuracy 1°), and recognition confidence (device model recognition probability output by the CNN model, range 0-1, e.g., 0.92 represents 92% confidence). The action space is a joint adjustment of the complexity parameter of the first random sequence (using the Lyapunov exponent of chaotic sequences, range 0.5-1.2, the larger the exponent, the stronger the randomness) and the mixed weight of the second environmental adaptive sequence (range 0-1, e.g., 0.6 represents the first sequence accounting for 60%). Through online and simulated environment training, the reinforcement learning model converges to the evaluation function that maximizes the interference effect (the evaluation function is defined as follows). ,in For normalized mean square error, The signal reproduction rate of the recording equipment. The larger the value, the better the interference effect; the target makes... The optimal strategy can be set to 10,000 training iterations, with the convergence condition being 1,000 consecutive iterations. Fluctuations do not exceed 0.01.

[0047] This approach transforms the parameter combinations of the interference policy library into a reinforcement learning policy function. Using environmental signal-to-noise ratio (SNR), device azimuth angle, and recognition confidence as state inputs, and sequence complexity and mixing weights as action outputs, the policy function is trained to output optimal actions based on real-time states, maximizing the interference effect evaluation function and achieving dynamic policy updates. The policy function can adjust actions in real-time based on state space parameters (e.g., when the SNR decreases, the Lyapunov exponent is adjusted from 0.8 to 1.1 to increase sequence randomness, and the mixing weights are adjusted from 0.6 to 0.7 to enhance environmental adaptability), ensuring that the interference parameters always adapt to the current state, thus maximizing the interference effect evaluation function. Maintaining an effective range of no less than 0.8, the perturbation success rate is improved compared to a fixed strategy. Secondly, through joint online and simulated training, the simulated environment can generate a massive number of states (such as 1000 different signal-to-noise ratio and azimuth angle combinations), accelerating model convergence (convergence is achieved in 10,000 iterations), avoiding the problems of insufficient data and slow convergence caused by relying solely on online training; simultaneously, the evaluation function is fused. (Direct interference effect) and (Recording equipment fidelity) can comprehensively measure the interference effect and avoid misjudgment optimization caused by a single indicator (such as looking only at...). (Ignoring the actual restoration capability of the device). In addition, the policy function can be dynamically updated (continuously optimized during online training). When a new recording device model is added, only the vulnerability feature data of the device needs to be input for minor fine-tuning (iteration 1000 times) to generate an adaptation policy, without the need to rebuild the library, thus reducing maintenance costs.

[0048] Furthermore, the methods for constructing and invoking the interference strategy library also include: When constructing the policy function, an adversarial neural network (using a generative adversarial network (GAN) structure, including a generator G and a discriminator D; the generator G simulates the noise reduction algorithm of the recording equipment (such as noise reduction based on wavelet threshold, adaptive filtering noise reduction), with the input being the interference signal and the output being the noise-reduced signal; the discriminator D is a binary classifier, with the input being the noise-reduced signal and the clean human voice signal, and the output being the signal class probability) is used to simulate the noise reduction algorithm of the recording equipment. By alternating training of a reinforcement learning model and an adversarial network, the policy function is optimized to generate interference signal parameters that can deceive or penetrate the denoising algorithm. The training process is as follows: 1- The reinforcement learning model generates interference parameters and generates interference signals; 2- The generator G denoises the interference signals; 3- The discriminator D determines whether the denoised signal is a clean human voice; 4- If the discriminator D determines it is an interference signal (probability < 0.5), the reinforcement learning model is fed back to adjust the parameters until the discriminator D determines it is a clean human voice (probability ≥ 0.8). At this point, the parameters are the parameters that penetrate the denoising algorithm.

[0049] By leveraging the adversarial training logic of adversarial neural networks, a generator simulates the noise reduction algorithm of a recording device, a discriminator judges the noise reduction effect, and a reinforcement learning model continuously optimizes the interference parameters in adversarial training against the adversarial neural network. This allows the generated interference signal to deceive the noise reduction algorithm (i.e., it is still identified as human voice after noise reduction, or it cannot be filtered by the noise reduction algorithm), thus achieving effective interference against recording devices with noise reduction functions. Since most current recording devices have noise reduction functions (such as AI noise reduction in mobile phones and environmental noise reduction in voice recorders), traditional interference signals are easily filtered by noise reduction algorithms, leading to interference failure. However, this method, through adversarial training, enables the signal generated by the interference parameters to penetrate the noise reduction algorithm (e.g., simulating the spectral characteristics of human voice, causing the noise reduction algorithm to misjudge it as human voice that should be preserved and not filter it). Secondly, the adversarial training-trained interference signal can adapt to the features preserved by the noise reduction algorithm (such as the energy distribution of human voice in the 1kHz-3kHz frequency band), making the interference signal appear as a human voice in the recording device, thus avoiding being directly identified as interference noise and eliminated by the noise reduction algorithm. At the same time, this human voice-like interference can more effectively suppress the real human voice (the two have overlapping spectra), causing the recording device to be unable to distinguish and restore the real human voice, reducing the recording restoration rate. In addition, adversarial neural networks can simulate different types of noise reduction algorithms (such as wavelet noise reduction and adaptive filtering noise reduction). Through alternating training, the policy function generated by the reinforcement learning model can be adapted to multiple noise reduction algorithms simultaneously, without the need to design parameters separately for each noise reduction algorithm.

[0050] In one alternative implementation, the method for generating the second environment-adaptive sequence includes: The collected environmental background noise was subjected to empirical mode decomposition (EMD decomposition, with 4 decomposition layers. The number of intrinsic mode functions (IMFs) was determined by the screening criteria (standard deviation less than 0.3), and finally 4 IMF components were obtained, which were denoted as IMF1 (high frequency component, 10kHz-20kHz), IMF2 (mid-high frequency component, 5kHz-10kHz), IMF3 (mid-low frequency component, 1kHz-5kHz), and IMF4 (low frequency component, 20Hz-1kHz) respectively, thus obtaining a set of intrinsic mode functions. Calculate the coherence coefficient of each intrinsic mode function (EMF) with the currently generated first random sequence in the time-frequency domain (using Short-Time Fourier Transform (STFT) to convert the EMF and the currently generated first random sequence into a time-frequency graph with a window length of 256 points and an overlap rate of 50%). The formula for calculating the coherence coefficient is as follows: ,in This is the IMF's time-frequency signal. The time-frequency signal of the first random sequence. for The conjugate of complex numbers, For mathematical expectation operator operations, Range 0-1, The smaller the value, the lower the correlation between the two. The K intrinsic mode function components with the lowest coherence coefficients (K value is set according to the interference bandwidth requirement, here K=2, for example, IMF3 has C=0.2 and IMF4 has C=0.15, which are the two lowest) are selected for recombination (the recombination method is linear superposition, and the superposition weight is allocated according to their respective energy proportions, IMF3 has a weight of 0.6 if it has an energy proportion of 60%, and IMF4 has a weight of 0.4 if it has an energy proportion of 40%), to generate a second environmental adaptive sequence that is statistically orthogonal to the first random sequence, so as to ensure that the fusion of the two sequences can cover a wider range of correlation function zeros. By means of EMD decomposition, the ambient background noise is broken down into IMF components of different frequency bands. The coherence coefficient of each IMF with the first random sequence is calculated, and the component with the lowest correlation is selected for recombination. Utilizing the statistical orthogonality property (low coherence coefficient), the two sequences are made complementary in the time-frequency domain, avoiding interference frequency gaps caused by overlap. The K IMFs with the lowest coherence coefficients are selected for recombination, which reduces the correlation between the second environmental adaptive sequence and the first random sequence, reduces the overlap area between the two in the time-frequency domain, and avoids the frequency coverage repetition caused by the fusion of similar sequences (such as all concentrated in 2kHz-5kHz). The fused composite signal can cover a wider frequency band, increase the number of zeros in the correlation function, fill interference frequency gaps, and prevent the recording equipment from collecting human voices through frequency band filtering. The second environmental adaptive sequence is generated by recombination of the IMF components of the ambient background noise. Its statistical characteristics (such as power spectrum distribution and time-domain fluctuations) are highly consistent with the ambient noise, allowing the interference signal to be hidden in the noise background and avoiding being identified as external interference and rejected by the ambient noise filtering function of the recording equipment.

[0051] In one optional implementation, the original ultrasonic signal is generated by simultaneously modulating a composite baseband signal onto a pair of ultrasonic carriers with frequencies of 40kHz and 40.5kHz. The modulation method is double-sideband amplitude modulation (DSB-AM), the carrier signal is a sine wave with an amplitude of 5V, and the modulation depth of the composite baseband signal is set to 0.8 to ensure that the modulated signal is not distorted. The frequency difference between 40kHz and 40.5kHz is 0.5kHz, which falls within the audible frequency range and corresponds to the key high-frequency band of human voice, effectively suppressing human voice in this frequency band.

[0052] Therefore, based on the difference frequency effect of the acoustic parametric array: when two ultrasonic carriers with similar frequencies propagate in the air and form a high sound pressure zone, the nonlinearity of the air causes the two carriers to interact, generating an audible sound signal with a frequency difference between the two carriers; at the same time, the composite baseband signal is modulated onto the carrier, so that the audible sound generated by the difference frequency carries the interference characteristics of the composite baseband, becoming the target interference signal, and preventing overmodulation distortion caused by excessive modulation of the composite baseband signal, ensuring that the original ultrasonic signal after modulation can accurately carry the interference characteristics; at the same time, the 5V carrier amplitude is matched with the composite baseband signal, which can generate an ultrasonic signal of sufficient strength, providing the energy basis for the subsequent formation of a high sound pressure zone and triggering the nonlinear difference frequency effect, thus improving the generation efficiency of the difference frequency signal.

[0053] Furthermore, the two original ultrasonic signals converge in the spatial region by using an FPGA-based programmable delay chip (using a Xilinx Spartan-7 series FPGA chip, model XC7S50, with a delay control accuracy of 1ns, a phase control step size of 0.1°, supporting parallel control of 8 signals, and adapting to the 8 elements of the ultrasonic transmitter array). This allows for independent control of the transmission delay and phase of each element in the ultrasonic transmitter array for the two ultrasonic signals. For example, if the recording device is located at (3m, 4m, 1.5m), the distance difference between each element and the target area is calculated. The FPGA chip controls element 1 to have a transmission delay of 10ns and a phase of 0° for the 40kHz signal, and a transmission delay of 12ns and a phase of 0.2° for the 40.5kHz signal; element 2 to have a transmission delay of 8ns and a phase of 0° for the 40kHz signal, and a transmission delay of 10ns and a phase of 0.2° for the 40.5kHz signal, ensuring that the two signals arrive and converge simultaneously in the target area.

[0054] By employing the precise delay and phase control of the FPGA programmable delay chip, the transmission timing of each element in the ultrasonic transmitter array is adjusted: based on the distance of each element to the target area, the required delay of the two ultrasonic signals is calculated, and the elements are controlled to transmit signals according to this delay, ensuring that the two signals arrive at the target area simultaneously, forming constructive interference (high sound pressure zone) and triggering nonlinear difference frequency effects; simultaneously, the phase is independently controlled to ensure that the phase difference between the two signals is stable when they intersect, avoiding phase disturbances that could cause distortion of the difference frequency signal. Furthermore, independent phase control stabilizes the phase difference between the two signals when they intersect, preventing random phase fluctuations that could cause disturbances in the amplitude and frequency of the difference frequency signal; a stable difference frequency signal can continuously suppress human voice in the target frequency band, avoiding interference gaps caused by signal fluctuations (such as when the interference signal weakens at a certain moment, allowing the recording equipment to pick up human voice), thus improving interference continuity.

[0055] In one alternative implementation, the method further includes: Acoustic stealth post-processing steps: After the directional emission step, a cancellation sound wave is emitted through an independent auxiliary loudspeaker (a full-range loudspeaker, such as the BOSE FreeSpace DS 16S, with a frequency response of 20Hz-20kHz, positioned at the center of the target protected area). This sound wave has the same amplitude and opposite phase to the target interference sound wave that constitutes the audible frequency interference sound field (for example, if the sound pressure level of the target interference sound wave at the edge of the protected area is 60dB and the phase is 0°, then the sound pressure level of the cancellation sound wave is set to 60dB and the phase is 180°). This cancellation sound wave is adjusted by an adaptive filter (the filter uses the minimum mean square error (LMS) algorithm, the filter order is set to 32, the iteration step size is 0.001, and the reference signal is a copy of the target interference sound wave), so that it cancels out the target interference sound wave within the protected area excluding the spatial orientation of the recording equipment. This achieves the goal of reducing the auditory impact on legitimate listeners in the area without reducing the interference effect on the target equipment.

[0056] Based on the principle of sound wave superposition and cancellation, when two sound waves with the same amplitude but opposite phases are superimposed in the same space, they will cancel each other out due to their opposite vibration directions (the sound pressure level is reduced to the level of ambient noise). An adaptive filter adjusts the amplitude and phase of the canceling sound wave in real time, ensuring precise cancellation with the interfering sound wave in the non-target area (the area where legitimate listeners are located). In the target area (the area where the recording equipment is located), the intensity of the interfering sound wave is much greater than the canceling sound wave, so the cancellation effect is negligible and does not affect the interference. In particular, traditional anti-recording interference methods are difficult to use in occupied settings (such as conference rooms or negotiation rooms) because the interfering sound affects legitimate listeners. This step, however, uses acoustic stealth to achieve precise interference with equipment without affecting people, allowing the method to be used normally in occupied settings and improving its applicability (such as business negotiations or confidential meetings).

[0057] In one alternative implementation, the method further includes: Signal fingerprinting encryption step: Before the parametric array modulation step, the acoustic channel impulse response between the ultrasonic transmitter array and the recording device is estimated using the impulse response method. A 5kHz pulse signal (sound pressure level 70dB) with a duration of 1ms is transmitted to the recording device. The reflected signal is collected through an auxiliary sensor array, and the channel impulse response is obtained by fitting using the least squares algorithm. (The length is set to 1024 points), and the characteristics of the acoustic channel response are extracted. Three key characteristics: peak amplitude (e.g., 0.5V), peak delay (e.g., 2.8ms), energy-concentrated frequency band (e.g., 1kHz-3kHz), press " The format is encoded as a 128-bit dynamic key, for example, "01101001-10011000-...-00110101". This dynamic key is used to encrypt the seed value of the first random sequence (initial seed value 0x12345678) or the modulation method of the composite baseband signal (such as ASK, FSK, PSK). The encryption algorithm is AES-128. The dynamic key is used as the AES key to XOR the seed value, and the encrypted seed value becomes 0x78563412. Alternatively, the modulation method is determined based on the last 4 bits of the key. For example, if the last 4 bits of the key are "0001", then ASK is selected, and if they are "0010", then FSK is selected.

[0058] This approach leverages the uniqueness of acoustic channels: the acoustic channels between recording devices and ultrasonic transmitter arrays at different locations (affected by distance, obstacles, and environmental reflections) possess unique impulse responses, and their characteristics (peak value, delay, frequency band) can serve as dynamic keys. By encrypting key parameters (seed value, modulation scheme) of the interference signal using this key, the interference signal carries a unique encrypted identifier, preventing unauthorized cracking or forgery. The dynamic key is generated based on real-time channel characteristics; each recording device and each location has a different key, avoiding the vulnerability of traditional fixed keys. The AES-128 encryption algorithm is extremely difficult to crack (brute-force cracking requires 2^128 attempts). Even if the interference signal is intercepted, unauthorized individuals cannot obtain the seed value of the first random sequence (unable to reproduce the sequence) or the modulation scheme (unable to parse the signal) by cracking the key, thus improving the security of the interference signal. Secondly, the encrypted interference signal carries a unique channel fingerprint. If an unauthorized individual forges the interference signal (without dynamic key encryption), its sequence seed value or modulation scheme will be inconsistent with the legitimate interference signal. At the recording device end, the interference effect will fail due to signal characteristic mismatch (such as inconsistent sequence randomness or incorrect modulation scheme). Meanwhile, the dynamic key updates as the channel changes, preventing unauthorized individuals from pre-preparing keys and further eliminating the possibility of forgery. Furthermore, the estimation of the acoustic channel impulse response is achieved using existing auxiliary sensor arrays and pulse signals, requiring no additional hardware; AES-128 encryption can be implemented using embedded algorithms on an FPGA chip (already used for delay control), eliminating the need for additional encryption chips, thus improving security while reducing cost increases, balancing security and economy.

[0059] In summary, the anti-recording interference method and apparatus of this embodiment accurately distinguishes different recording devices through acoustic fingerprint recognition technology, and can generate targeted interference strategies based on the specific acoustic characteristics of the devices. It employs a combination of acoustic parametric array and beamforming technology to achieve precise directional transmission of interference sound waves, concentrating interference energy on the target device and improving interference efficiency. Furthermore, it enhances the unpredictability and anti-filtering capability of the interference signal through the fusion of multiple dynamic sequences and the construction of composite baseband signals. Thus, it achieves precise and adaptive anti-recording protection while maintaining high-efficiency interference performance and reducing the impact on normal voice communication.

[0060] This embodiment provides a recording interference prevention device in a second aspect, which applies the recording interference prevention method described above, such as... Figure 2 As shown, the device includes: The environmental perception module is configured to: monitor the sound field changes of the target protection area in real time through an auxiliary sensor array, and collect the ambient background noise; The device identification and positioning module is configured to: when an abnormal sound signal is detected, extract the acoustic fingerprint features of the sound signal, match it with a preset recording device voiceprint feature database, in order to identify the type and model of the recording device and determine its spatial location; The interference strategy decision module is configured to: based on the identified recording device model, call a targeted combination of interference parameters from a preset interference strategy library; the interference strategy library records the vulnerability characteristics of different recording devices against various interference signals; The dynamic sequence generation module is configured to generate at least two dynamic sequences, including a first random sequence and a second environment adaptive sequence, wherein the second environment adaptive sequence is generated by reconstructing the noise features of the background noise of the environment collected in the environment perception step; The signal fusion construction module is configured to fuse the first random sequence and the second environmental adaptive sequence according to the interference parameter combination to construct a composite baseband signal; The parametric array modulation module is configured to modulate the composite baseband signal onto a pair of ultrasonic carriers with a fixed frequency difference to generate an ultrasonic transmission signal with two original ultrasonic signals. The directional transmission module is configured to: based on the spatial orientation of the recording device, use an ultrasonic transmitter array and beamforming technology to directionally focus and transmit ultrasonic transmission signals to the spatial area where the recording device is located; wherein, the two original ultrasonic signals propagate in the air and converge in the spatial area, and after nonlinear interaction based on the principle of acoustic parametric array, they self-demodulate to generate a highly directional audible frequency band interference sound field, which acts on the recording device. The evaluation and adjustment module is configured to: evaluate the interference effect of the audible frequency band interference sound field on the recording equipment in real time, and dynamically adjust any one of the following based on the evaluation results: the combination of interference parameters, the modulation method of the composite baseband signal, or the transmission power of the ultrasonic transmission signal.

[0061] It should be noted that the anti-recording interference device of this embodiment corresponds to the aforementioned anti-recording interference method. Therefore, the parts of the anti-recording interference device of this embodiment that are not specifically described (including but not limited to specific technical means and technical effects) can be referred to the relevant descriptions in the aforementioned anti-recording interference method, and will not be repeated here.

[0062] In the embodiments provided by this invention, it should be understood that the embodiments described herein can be implemented in hardware, software, firmware, middleware, code, or any suitable combination thereof. For hardware implementation, the processor can be implemented in one or more of the following: application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), processors, controllers, microcontrollers, microprocessors, other electronic units designed to implement the functions described herein, or combinations thereof. For software implementation, some or all of the processes of the embodiments can be performed by a computer program instructing the associated hardware. During implementation, the program can be stored in a computer-readable storage medium or transmitted as one or more instructions or code on a computer-readable storage medium. Computer-readable storage media include computer storage media and communication media, wherein communication media include any medium that facilitates the transmission of a computer program from one place to another. Storage media can be any available medium accessible to a computer. Computer-readable storage media can include, but are not limited to, RAM, ROM, EEPROM, CD-ROM or other optical disk storage, magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code having the form of instructions or data structures and accessible to a computer.

[0063] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for preventing recording interference, characterized in that, The method includes: Environmental perception steps: Real-time monitoring of sound field changes in the target protection area using an auxiliary sensor array, and collection of ambient background noise; Device identification and positioning steps: When an abnormal sound signal is detected, the acoustic fingerprint feature of the sound signal is extracted and matched with a preset recording device voiceprint feature database to identify the type and model of the recording device and determine its spatial location. Interference strategy decision-making steps: Based on the identified recording device model, retrieve a targeted combination of interference parameters from a pre-set interference strategy library; the interference strategy library records the vulnerability characteristics of different recording devices against various interference signals; Dynamic sequence generation step: Generate at least two dynamic sequences, including a first random sequence and a second environment adaptive sequence, wherein the second environment adaptive sequence is reconstructed based on the noise characteristics of the background noise of the environment collected in the environment perception step; Signal fusion construction steps: The first random sequence and the second environmental adaptive sequence are fused according to the interference parameter combination to construct a composite baseband signal; Parametric array modulation step: The composite baseband signal is modulated onto a pair of ultrasonic carriers with a fixed frequency difference to generate an ultrasonic transmission signal with two original ultrasonic signals. Directional transmission step: Based on the spatial orientation of the recording device, the ultrasonic transmission signal is directionally focused and transmitted to the spatial area where the recording device is located using an ultrasonic transmitter array and beamforming technology; wherein, the two original ultrasonic signals propagate in the air and converge in the spatial area, and after nonlinear interaction based on the principle of acoustic parametric array, they self-demodulate to generate a highly directional audible frequency band interference sound field, which acts on the recording device; Evaluation and adjustment steps: Evaluate the interference effect of the audible frequency band interference sound field on the recording device in real time, and dynamically adjust any one of the interference parameter combination, the modulation mode of the composite baseband signal, or the transmission power of the ultrasonic transmission signal based on the evaluation results.

2. The method for preventing recording interference according to claim 1, characterized in that, In the device identification and localization step, the extraction of the acoustic fingerprint features includes: A linear frequency modulated detection signal with a frequency of 18kHz-22kHz is transmitted toward the target protection area; The microphone of the recording device located within the target protection area receives the linear frequency modulation (LFM) probe signal and, due to its nonlinear characteristics, generates intermodulation distortion products. The amplitude ratio of the second harmonic to the third harmonic of the LFM probe signal, and the frequency of the LFM probe signal are extracted. The energy of the intermodulation distortion components, of which , and These are two different frequency components within the linear frequency modulated detection signal; The extracted amplitude ratio and the energy composition feature vector of the intermodulation distortion component are input into a convolutional neural network model pre-trained by transfer learning to complete the classification and identification of the device model.

3. The method for preventing recording interference according to claim 1, characterized in that, In the interference strategy decision-making step, the method for constructing and invoking the interference strategy library includes: A policy function dynamically updated based on a reinforcement learning model is used as the combination of interference parameters for each type of recording device in the interference policy library; wherein, the state space of the policy function includes the environmental signal-to-noise ratio, device azimuth angle and recognition confidence, and the action space is the joint adjustment of the complexity parameter of the first random sequence and the mixed weight of the second environmental adaptive sequence; By training in online and simulated environments, the reinforcement learning model converges to the optimal policy that maximizes the evaluation function of the interference effect.

4. The method for preventing recording interference according to claim 3, characterized in that, The method for constructing and calling the interference strategy library also includes: When constructing the policy function, an adversarial neural network is used to simulate the noise reduction algorithm of the recording device; By alternating training of the reinforcement learning model and the adversarial network, the policy function is optimized to generate interference signal parameters that can deceive or penetrate the noise reduction algorithm.

5. The method for preventing recording interference according to claim 1, characterized in that, In the dynamic sequence generation step, the method for generating the second environment-adaptive sequence includes: Empirical mode decomposition is performed on the collected environmental background noise to obtain a set of intrinsic mode functions; Calculate the coherence coefficient between each intrinsic mode function and the currently generated first random sequence in the time-frequency domain; The K intrinsic mode function components with the lowest coherence coefficients are selected and recombined to generate the second environment adaptive sequence that is statistically orthogonal to the first random sequence.

6. The method for preventing recording interference according to claim 1, characterized in that, The original ultrasonic signal is generated by simultaneously modulating the composite baseband signal onto a pair of ultrasonic carriers with frequencies of 40 kHz and 40.5 kHz.

7. The method for preventing recording interference according to claim 6, characterized in that, The two original ultrasonic signals converge in the spatial region by using an FPGA-based programmable delay chip to independently control the transmission delay and phase of each array element in the ultrasonic transmitter array for the two ultrasonic signals.

8. The method for preventing recording interference according to claim 1, characterized in that, The method also includes: Acoustic stealth post-processing steps: After the directional emission step, a canceling sound wave is emitted through an independent auxiliary loudspeaker. This sound wave has the same amplitude and opposite phase to the target interference sound wave constituting the audible frequency band interference sound field. The canceling sound wave is adjusted by an adaptive filter so that it cancels out the target interference sound wave in a protected area other than the spatial orientation of the recording device.

9. The method for preventing recording interference according to claim 1, characterized in that, The method also includes: Signal fingerprint encryption step: Before the parametric array modulation step, the acoustic channel impulse response from the ultrasonic transmitter array to the recording device is estimated, and the characteristics of the acoustic channel response are used as a dynamic key to encrypt the seed value of the first random sequence or the modulation method of the composite baseband signal.

10. An apparatus for applying the anti-recording interference method as described in any one of claims 1-9, characterized in that, The device includes: The environmental perception module is configured to: monitor the sound field changes of the target protection area in real time through an auxiliary sensor array, and collect the ambient background noise; The device identification and positioning module is configured to: when an abnormal sound signal is detected, extract the acoustic fingerprint features of the sound signal, match it with a preset recording device voiceprint feature database, in order to identify the type and model of the recording device and determine its spatial location; The interference strategy decision module is configured to: based on the identified recording device model, call a targeted combination of interference parameters from a preset interference strategy library; the interference strategy library records the vulnerability characteristics of different recording devices against various interference signals; The dynamic sequence generation module is configured to generate at least two dynamic sequences, including a first random sequence and a second environment adaptive sequence, wherein the second environment adaptive sequence is reconstructed based on the noise characteristics of the background noise of the environment collected in the environment perception step; The signal fusion construction module is configured to fuse the first random sequence and the second environmental adaptive sequence according to the interference parameter combination to construct a composite baseband signal; The parametric array modulation module is configured to modulate the composite baseband signal onto a pair of ultrasonic carriers with a fixed frequency difference to generate an ultrasonic transmission signal with two original ultrasonic signals. The directional transmission module is configured to: based on the spatial orientation of the recording device, use an ultrasonic transmitter array and beamforming technology to directionally focus and transmit the ultrasonic transmission signal to the spatial area where the recording device is located; wherein, the two original ultrasonic signals propagate in the air and converge in the spatial area, and after nonlinear interaction based on the principle of acoustic parametric array, they self-demodulate to generate a highly directional audible frequency band interference sound field, which acts on the recording device; The evaluation and adjustment module is configured to: evaluate the interference effect of the audible frequency band interference sound field on the recording device in real time, and dynamically adjust any one of the interference parameter combination, the modulation mode of the composite baseband signal, or the transmission power of the ultrasonic transmission signal based on the evaluation results.