Sound deception signal confrontation method and device based on time domain energy distribution feature recognition

Abstract

The invention discloses an acoustic deception signal confrontation method and device based on time domain energy distribution feature recognition. The method comprises the following steps that: underwater detection equipment acquires a noise signal when a target passes through an underwater acoustic sensor; calculating a power spectrum based on the noise signal, and extracting frequency band energy corresponding to the noise signal and a feature Ei corresponding to the bandwidth of the noise signal; selecting E0 as a reference energy value, performing ratio processing on other features Ep except E0 and the reference energy value to obtain a ratio feature ap, and inputting the ratio feature ap into the trained neural network model for calculation to obtain an identification result; the recognition result represents whether the noise signal contains an acoustic deception signal or not. The ship target identification method based on the time domain energy distribution characteristics is established, the balance difference of the frequency spectrum structure is extracted, effective identification can be achieved, the essential difference between the sound deception signal and the real ship noise can be excavated, and a decision can be made in real time.

Description

Technical Field

[0001] The present invention relates to the technical field of signal processing, and particularly relates to a method and device for countering acoustic deception signals based on time-domain energy distribution characteristics recognition. Background Art

[0002] The recognition system is an important part of underwater detection equipment, and has developed rapidly due to its wide range of application scenarios. The detection equipment detects and identifies ship targets. In order to protect itself from being detected and identified, ship targets usually configure acoustic deception devices on accompanying unmanned remotely controlled ships. The acoustic deception devices simulate the noise of the ship targets, and interfere with the detection and identification of the targets by the detection equipment through noise, thus seriously affecting the accuracy of the detection equipment in target recognition.

[0003] The detection equipment receives information such as signals and noises of ship targets through a passive acoustic system, and tracks and locates the targets, and finally determines the target category. This requires the detection equipment to adapt to ships with widely different noises, and also to face various acoustic deception signals that simulate the noises of target ships, without being confused and misidentified by them. Summary of the Invention

[0004] In view of this, the present invention provides a method and device for countering acoustic deception signals based on time-domain energy distribution characteristics recognition, which can solve the technical problem that the acoustic deception signals emitted by acoustic deception devices interfere with the detection of underwater detection equipment.

[0005] In order to solve the above technical problem, the present invention is implemented as follows.

[0006] A method for countering acoustic deception signals based on time-domain energy distribution characteristics recognition, the method comprising:

[0007] Step S1: An underwater detection device obtains a noise signal when a target passes through by means of an underwater acoustic sensor;

[0008] Step S2: Calculate a power spectrum based on the noise signal, extract the band energy corresponding to the noise signal, and a feature E corresponding to the bandwidth of the noise signal. i , where i is the number of frequency bands, determined according to the analysis frequency range, and the value range of i is 0 ≤ i ≤ M, and M is a natural number;

[0009] Step S3: Select E0 as a reference energy value, and perform a ratio process on other features E except E0 p with the reference energy value, 1 ≤ p ≤ M, to obtain a ratio feature a p , and the ratio feature a pInput the trained neural network model for calculation to obtain the recognition result; the recognition result indicates whether the noise signal contains a voice spoofing signal; the reference energy value is the band energy with an energy change degree less than a preset threshold.

[0010] Preferably, perform a Fourier transform on the noise signal to obtain a power spectrum, and extract the band energy corresponding to the noise signal and the feature E corresponding to the bandwidth of the noise signal i , and the extraction formula is:

[0011]

[0012] where N is the number of frequency points within the bandwidth, and P j is the power spectrum intensity corresponding to the frequency point.

[0013] Preferably, the neural network recognition model includes an input layer, a hidden layer, and an output layer connected in sequence; the entire neural network forms a classification decision maker, and the classification decision maker receives the ratio feature a p , and obtains the recognition result through calculation by the activation functions of the hidden layer and the output layer.

[0014] Preferably, E0 is used as the reference energy value, and the reference energy value is selected based on the band with an energy change degree less than a preset threshold.

[0015] Preferably, each neuron of the neural network model is set in the form of an S-type activation function.

[0016] An anti-voice spoofing signal device based on time-domain energy distribution feature recognition provided by the present invention, the device includes:

[0017] Noise signal acquisition module: configured to enable an underwater detection device to acquire a noise signal when a target passes through via an underwater acoustic sensor;

[0018] Power spectrum determination module: configured to calculate a power spectrum based on the noise signal, and extract the band energy corresponding to the noise signal and the feature E corresponding to the bandwidth of the noise signal i , i is the number of bands, determined according to the analysis frequency range, and the value range of i is 0 ≤ i ≤ M, where M is a natural number;

[0019] Recognition module: configured to select E0 as the reference energy value, perform a ratio process on other features E except E0 p with the reference energy value, 1 ≤ p ≤ M, to obtain a ratio feature a p , and input the ratio feature a p into the trained neural network model for calculation to obtain the recognition result; the recognition result indicates whether the noise signal contains a voice spoofing signal; the reference energy value is the band energy with an energy change degree less than a preset threshold.

[0020] A computer-readable storage medium provided by the present invention stores multiple instructions therein; the multiple instructions are used to be loaded and executed by a processor to perform the method as described above.

[0021] An electronic device provided by the present invention is characterized in that the electronic device includes:

[0022] A processor for executing multiple instructions;

[0023] A memory for storing multiple instructions;

[0024] Wherein, the multiple instructions are used to be stored by the memory and loaded and executed by the processor to perform the method as described above.

[0025] Beneficial technical effects brought by the present invention:

[0026] (1) In view of the characteristics of modern acoustic simulation devices with a wide frequency band, controllable spectral characteristics and dynamic range, the present invention establishes a ship target recognition method based on the time-domain energy distribution characteristics, extracts the difference in the balance of the spectral structure, and can effectively cope with all types of underwater recognition systems. It can dig out the essential difference between the acoustic deception signal and the real ship noise and make a decision in real time.

[0027] (2) The present invention describes the noise broadband spectral structure with a ratio sequence. This method can not only accurately extract the spectral structure characteristics of the noise, but also compress the dynamic range to ensure stable recognition at a long distance. By digging out the difference in the balance of the spectral structure, the relationship between the system and the total broadband sound level of the target is not significant, that is, it has little relationship with the target tonnage, distance, etc. The balance difference does not focus on the energy amplitude or change of a certain or a few frequency bands, but pays attention to the spectral balance covering all frequency bands. And this "spectral balance covering all frequency bands" is exactly the difficulty and the ultimate goal of the development of acoustic simulation devices.

[0028] (3) The present invention uses an adaptive neural network to construct an identification and classification decision-making device, breaking through the limitations of the previous method of artificially determining thresholds on the number of samples, the dimension of feature quantities, correlation, etc., and organically combining the underlying signal analysis technology and high-level knowledge representation. The knowledge that there is a balance difference in the spectral structure between the simulated ship noise signal and the real ship noise signal is transplanted into the underwater target recognition system, so that the system that is unattended for a long period of time after deployment has the ability to distinguish true and false targets in the face of a large number of ships and acoustic simulation devices with open and extremely different noise spectral structure characteristics. Description of the Drawings

[0029] Figure 1 It is a schematic flow chart of the method for combating acoustic deception signals based on time-domain energy distribution characteristics recognition of the present invention;

[0030] Figure 2 Schematic diagram of the principle of the method for combating acoustic spoofing signals based on time-domain energy distribution characteristics of the present invention;

[0031] Figure 3 Schematic diagram of the structure of the device for combating and identifying acoustic spoofing signals based on time-domain energy distribution characteristics of the present invention. Specific embodiments

[0032] The present invention will be described in detail below with reference to the accompanying drawings and embodiments.

[0033] As Figure 1 shown, the present invention proposes a method for combating and identifying acoustic spoofing signals based on time-domain energy distribution characteristics, and the method includes:

[0034] Step S1: An underwater detection device obtains a noise signal when a target passes through by means of an underwater acoustic sensor;

[0035] Step S2: Calculate the power spectrum based on the noise signal, and extract the band energy corresponding to the noise signal and the characteristic E corresponding to the bandwidth of the noise signal i , where i is the number of frequency bands, determined according to the analysis frequency range, and the value range of i is 0 ≤ i ≤ M, and M is a natural number;

[0036] Step S3: Select E0 as the reference energy value, and perform a ratio process on other characteristics E except E0 p with the reference energy value, 1 ≤ p ≤ M, to obtain a ratio characteristic a p , input the ratio characteristic a p into the trained neural network model for calculation to obtain an identification result. Each neuron of the neural network is set in the form of an S-shaped activation function; the identification result indicates whether the noise signal contains an acoustic spoofing signal; the reference energy value is the band energy with an energy change degree less than a preset threshold.

[0037] Modern acoustic simulation devices are characterized by a wide frequency band, controllable spectral characteristics and dynamic range, and can effectively cope with all types of underwater identification systems. Therefore, it is necessary to dig out the essential differences between this simulated signal and the real ship noise and make decisions in real time.

[0038] Further, perform FFT (fast Fourier transform) on the noise signal to obtain a power spectrum, and extract the band energy corresponding to the noise signal and the characteristic E corresponding to the bandwidth of the noise signal i , and the extraction formula is:

[0039]

[0040] where N is the number of frequency points within the bandwidth, Pj is the power spectral intensity corresponding to the frequency point.

[0041] Further, a ratio process is performed based on the band energy and characteristics. Here, E0 is used as a reference energy value, and the reference energy value is the band energy with an energy change degree less than a preset threshold. The ratio extraction formula is:

[0042]

[0043] The neural network recognition model includes an input layer, a hidden layer, and an output layer connected in sequence. The hidden layer neurons select the S-shaped activation function, and the entire neural network forms a classification decision maker. The classification decision maker receives the ratio feature a p , and obtains the recognition result through the calculation and processing of the activation functions of the hidden layer and the output layer.

[0044] Ship noise is caused by the vibration and movement of the entire ship driven by the main and auxiliary engines and other machines during operation. The frequency can range from very low infrasonic frequencies to very high ultrasonic frequencies. When the acoustic simulation device emits sound by combining single or multiple sound generators and simulates the ship noise signal, time-domain control and frequency-domain control are required, and broadband spectrum equalization control is its weak point and technical implementation difficulty.

[0045] The underwater recognition system cannot simply extract the time-domain and frequency-domain features in a broadband or several frequency bands, but should perform a refined spectrum analysis covering the entire frequency band. In the broadband range, there is an equalization difference in the spectral structure between the ship noise signal emitted by the acoustic simulation device and the real ship noise signal. In the shape of the spectrum diagram covering the entire frequency band, there are inevitably phenomena such as large mutations in the spectral values of adjacent frequency bands and overly large differences in the spectral values of some frequency bands, which exceed the difference range of general ship noise. This is the essential difference between the two.

[0046] At the same time, in the face of a large number of samples, feature quantities with many dimensions and strong correlations, an adaptive neural network is used to construct a recognition classification decision maker to complete the mapping from the ratio sequence of feature quantities to the target type. The processable feature quantities can be greatly improved, and the work that was difficult or impossible to complete by manual statistics in the past is handed over to the computer to be automatically completed according to the principle of minimum error. The neural network learning samples include measured data and virtual samples, enabling the system to have a wider adaptability and higher intelligence in the face of a large number of ships and mine sweepers with open and extremely different noise spectral structure characteristics.

[0047] The present invention provides an estimation device for the direction of an underwater target, and the device includes:

[0048] Noise signal acquisition module: configured to obtain the noise signal when the target passes through through an underwater detection device via an underwater acoustic sensor;

[0049] Power spectrum determination module: Configured to calculate the power spectrum based on the noise signal, extract the band energy corresponding to the noise signal, and the feature E corresponding to the bandwidth of the noise signal i , where i is the number of frequency bands, determined according to the analysis frequency range, and the value range of i is 0 ≤ i ≤ M, and M is a natural number;

[0050] Recognition module: Configured to select E0 as the reference energy value, and perform a ratio process on the other features E except E0 p with the reference energy value, 1 ≤ p ≤ M, to obtain the ratio feature a p , and input the ratio feature a p into the trained neural network model for calculation to obtain the recognition result; the recognition result represents whether the noise signal contains a voice spoofing signal; the reference energy value is the band energy with an energy change degree less than a preset threshold.

[0051] A computer-readable storage medium provided by the present invention, wherein multiple instructions are stored in the storage medium; the multiple instructions are used to be loaded and executed by a processor to perform the method as described above.

[0052] An electronic device provided by the present invention, characterized in that the electronic device includes:

[0053] A processor for executing multiple instructions;

[0054] A memory for storing multiple instructions;

[0055] Wherein, the multiple instructions are used to be stored by the memory and loaded and executed by the processor to perform the method as described above.

[0056] The above specific embodiments only describe the design principle of the present invention. The shapes and names of the components in this description can be different and are not limited. Therefore, those skilled in the art of the present invention can modify or equivalently replace the technical solutions recorded in the foregoing embodiments; and these modifications and replacements do not depart from the spirit and technical solutions of the present invention, and should all belong to the protection scope of the present invention.

Claims

1. An anti-acoustic spoofing signal method based on the recognition of time-domain energy distribution characteristics, characterized in that, The method includes: Step S1: The underwater detection device acquires the noise signal when the target passes through by means of an underwater acoustic sensor; Step S2: Calculate the power spectrum based on the noise signal, extract the band energy corresponding to the noise signal, and the feature E corresponding to the bandwidth of the noise signal i , where i is the number of frequency bands, determined according to the analysis frequency range, and the value range of i is 0 ≤ i ≤ M, where M is a natural number; Step S3: Select E0 as the reference energy value, and perform a ratio process on other features E except E0, where 1 ≤ p ≤ M, to obtain the ratio feature a p The ratio is calculated with the reference energy value, and the ratio feature a is obtained p , and input the ratio feature a p into the trained neural network model for calculation to obtain the recognition result; the recognition result indicates whether the noise signal contains a voice spoofing signal; the reference energy value is the band energy with an energy change degree less than the preset threshold.

2. The method according to claim 1, characterized in that, Perform a Fourier transform on the noise signal to obtain a power spectrum, and extract the band energy corresponding to the noise signal and the feature E corresponding to the bandwidth of the noise signal i , and the extraction formula is: where N is the number of frequency points within the bandwidth, and P j is the power spectral intensity corresponding to the frequency point.

3. The method according to claim 2, characterized in that, The neural network recognition model includes an input layer, a hidden layer, and an output layer connected in sequence; the entire neural network forms a classification decision maker, and the classification decision maker receives the ratio feature a p , and obtains the recognition result through the calculation and processing of the excitation functions of the hidden layer and the output layer.

4. The method according to claim 3, wherein Where E0 is used as a reference energy value, and the reference energy value is selected based on the frequency band with an energy change degree less than a preset threshold.

5. The method according to any one of claims 1-4, characterized in that, Each neuron of the neural network model is set in the form of an S-shaped activation function.

6. An acoustic spoofing signal countermeasure device based on the recognition of time-domain energy distribution characteristics, characterized in that, The device includes: Noise signal acquisition module: configured such that the underwater detection device acquires the noise signal when the target passes through by means of an underwater acoustic sensor; Power spectrum determination module: configured to calculate a power spectrum based on the noise signal, extract the band energy corresponding to the noise signal, and a feature E corresponding to the bandwidth of the noise signal i , where i is the number of frequency bands, determined according to the analysis frequency range, and the value range of i is 0 ≤ i ≤ M, where M is a natural number; Recognition module: Configured to select E0 as the reference energy value, and perform a ratio processing on other features E except E0, where 1 ≤ p ≤ M, to obtain the ratio feature a p and perform a ratio processing with the reference energy value to obtain the ratio feature a p , and input the ratio feature a p into the trained neural network model for calculation to obtain the recognition result; the recognition result indicates whether the noise signal contains a voice deception signal; the reference energy value is the band energy with an energy change degree less than the preset threshold.

7. A computer-readable storage medium, in which multiple instructions are stored; the multiple instructions are used to be loaded and executed by a processor to perform the method according to any one of claims 1-5.

8. An electronic device, characterized in that, The electronic device includes: A processor for executing multiple instructions; A memory for storing multiple instructions; Wherein, the multiple instructions are used to be stored by the memory and loaded and executed by the processor to perform the method according to any one of claims 1-5.

Images