CN115047448A - Indoor target rapid detection method and system based on acoustic-electromagnetic intermodulation - Google Patents

Abstract

The invention provides a method and system for fast detection of indoor targets based on acousto-electromagnetic intermodulation. The acousto-electromagnetic echoes of each of the multiple indoor targets are extracted, and the Doppler characteristics and amplitude characteristics of each of the acousto-electromagnetic echoes located in different frequency bands are extracted; Perform feature fusion to obtain the multi-band Doppler fusion feature and the multi-band amplitude fusion feature of each target; The machine learning model obtains the corresponding classification results and then performs decision fusion to determine the type of the target corresponding to each of the acousto-electromagnetic echoes. The invention improves the efficiency and accuracy of target detection, and solves the problem of single-means detection.

Description

A method and system for fast detection of indoor targets based on acousto-electromagnetic intermodulation

technical field

The invention relates to the technical field of target detection, in particular to an indoor target rapid detection method and system based on acousto-electromagnetic intermodulation.

Background technique

In disaster rescue and other operations, rescuers often need to go deep inside unfamiliar buildings, and the lack of information inside buildings will pose a greater threat to the smooth development of operations and the safety of personnel. Therefore, it has important practical significance and research value to study the penetration detection inside the building.

In recent years, penetration detection schemes have improved over the past due to advances in implementation algorithms and receiver performance. In the case of not destroying the scene, detection technologies such as acoustic waves, infrared waves, and electromagnetic waves can achieve penetration detection to varying degrees. Among them, electromagnetic signals have long been used to detect, characterize and identify targets in a variety of detection applications, including interferometry, navigation, and penetration detection; however, in penetration detection, electromagnetic signals are used in penetration detection due to the difference between the target object and the background environment. The typical low contrast between the two is greatly limited.

Based on the above problems, in order to avoid the exposure of the detection system, researchers need to further complicate the structural design, which leads to problems such as high complexity and not easy portability of the traditional detection system. Therefore, the research on the development of portable new detection modes that can provide characterization of unknown objects at a long distance has attracted the attention of scholars at home and abroad.

Due to the limitations of detection methods in many environments, there has been renewed interest in acoustic signals for detection applications. In a similar process to electromagnetic wave detection, the scattered sound field produced by the interaction of the incident sound wave with the object contains information about the scattering object. But for acoustic systems, the high reflectivity of most materials prevents the coupling of acoustic energy into the object, while for electromagnetic systems, electromagnetic losses and metallic objects prevent the coupling of electromagnetic energy into the interior of the object. Furthermore, acoustic detection of the desired reflected signal may not be possible in an acoustically saturated environment with a lot of background clutter, or when dealing with acoustic shear waves that do not produce a direct reflected signal. Therefore, it is difficult for both independent detection systems to obtain feature information about objects.

SUMMARY OF THE INVENTION

The present invention proposes a method and system for fast detection of indoor targets based on acousto-electromagnetic intermodulation, so as to solve the above-mentioned defects of the prior art.

In one aspect, the present invention provides a fast detection method for indoor targets based on acousto-electromagnetic intermodulation, the method comprising the following steps:

S1: Acoustic excitation of a plurality of indoor targets by acoustic wave signals induces mechanical vibration, so as to generate acoustic excitation signals, and at the same time, the electromagnetic wave signals and the acoustic excitation signals are used to perform acousto-electromagnetic mutual modulation to generate acoustic and electromagnetic echoes;

S2: Receive an acousto-electromagnetic echo of each of the multiple indoor targets, and in each frequency band, perform feature extraction on the acousto-electromagnetic echo based on the two dimensions of the Doppler effect and amplitude modulation, respectively, Thereby, the Doppler features and amplitude features of each of the acousto-electromagnetic echoes located in different frequency bands are extracted;

S3: Use the splicing method to perform feature fusion on the Doppler features located in different frequency bands to obtain multi-band Doppler fusion features of each target, and use the splicing method to fuse the amplitudes of the different frequency bands. The feature is characterized by feature fusion to obtain the multi-band amplitude fusion feature of each target;

S4: Construct a plurality of machine learning models based on different classifiers according to the multi-band Doppler fusion feature and the multi-band amplitude fusion feature, and use the different machine learning models to analyze each of the acousto-electromagnetic echoes respectively. Perform classification to obtain classification results corresponding to different machine learning models, and then perform decision fusion on the classification results to determine the type of the target corresponding to each of the acousto-electromagnetic echoes.

The above method combines the acoustic detection system and the electromagnetic detection system to characterize the difference between the object and its surrounding environment through the Doppler frequency shift and amplitude modulation caused by the interaction of the propagating electromagnetic wave and the wave sound source in the object, that is, the acousto-electromagnetic detection method . For indoor targets, the contrast with the environment is enhanced by mechanical vibration. The synthesized reflected and scattered electromagnetic signals are modulated with properties specific to the structure and composition of the vibrating object, thereby providing information about the target structure and distinguishing it from clutter. For the extracted target information, deep learning is used for feature modeling, and then the main features of various targets are classified. By using two orthogonal detection systems of acoustics and electromagnetics, the present invention can better obtain the information of the object compared with the situation that the single-mode detection system may prevent the sound wave from penetrating the interior of the object or cannot receive the reflected wave. Using different characteristics of echo signals, the accuracy of target detection is improved, the problem of single-method detection is solved, and the target detection efficiency is improved.

In a specific embodiment, the sound source level of the sound wave signal satisfies one-way penetration.

In a specific embodiment, the S1 specifically includes: using an acoustic wave signal to acoustically excite a plurality of indoor targets to induce mechanical vibration, so as to generate an acoustic excitation signal, and adding the acoustic excitation signal to the electromagnetic wave signal after passing through a power amplifier The electromagnetic wave signal is modulated to obtain an acoustic electromagnetic echo.

In a specific embodiment, the S2 specifically includes:

receiving the acousto-electromagnetic echoes of each of the plurality of indoor targets, pre-processing the acousto-electromagnetic echoes and then performing Fourier transform to obtain the frequency spectrum of each of the acousto-electromagnetic echoes;

In each frequency band, the features of the acousto-electromagnetic echoes are extracted based on the two dimensions of the Doppler effect and the amplitude modulation, so as to extract the Doppler features and Amplitude characteristics.

In a specific embodiment, the feature extraction of the acoustic electromagnetic echo based on the Doppler effect includes:

The phase modulation of the electromagnetic wave signal caused by the Doppler effect in the acousto-electromagnetic echo is analyzed, and the phase modulation amplitude generated by the phase modulation is obtained. The Doppler effect, which determines the interaction of the propagating electromagnetic wave with the wave sound source will cause the Doppler frequency shift of the electromagnetic wave. The vibrational motion can change the distance of the object relative to the electromagnetic wave source, and the electromagnetic wave source modulates the phase of the reflected radio frequency signal, so as to The target's unique object characteristics are modulated in phase, producing target information.

In a specific embodiment, the feature extraction of the acoustic electromagnetic echo based on the amplitude modulation includes:

The amplitude modulation of the electromagnetic wave signal caused by special relativity, path loss and radar cross section in the acoustic electromagnetic echo is analyzed respectively, and the amplitude modulation amplitude caused by special relativity, the amplitude modulation amplitude caused by path loss and Amplitude modulation amplitude due to radar cross section. Amplitude modulation, when analyzing electromagnetic scattering from a vibrating object, due to the overall contribution of the modulated sideband power is much smaller than the phase modulation produced by the Doppler effect, and is often below the measurable noise floor of the detection system. However, the dynamic range of the RF detection system can be improved by the analog cancellation technique, and the detection capability of low-frequency and low-level sideband modulation in the reflected signal can be significantly improved. Therefore, the effects of special relativity, road stiffness loss and RCS on amplitude modulation are reintroduced into vibration. In the analysis of RF signals scattered by objects.

In a specific embodiment, the S4 specifically includes:

S401: Build a micro-Doppler feature library and a multi-order amplitude feature library respectively according to the multi-band Doppler fusion feature and the multi-band amplitude fusion feature, and according to the micro-Doppler feature library and the multi-order amplitude feature The feature library performs preliminary matching and judgment on the received acousto-electromagnetic echoes of the target to be measured, and obtains a preliminary judgment result;

S402: Build a micro-Doppler feature library and a multi-order amplitude feature library according to the multi-band Doppler fusion feature and the multi-band amplitude fusion feature, respectively, to construct multiple sample data sets, and use each of the sample data sets to separate Perform ensemble learning based on different classifiers to build different machine learning models;

S403: Use the different machine learning models to classify and identify the Doppler feature and the amplitude feature of the acousto-electromagnetic echo of the target to be measured to obtain a corresponding classification result, and vote on the classification result decision to get the decision result;

The preliminary judgment result and the decision result obtained after performing the S401 to the S403 on the acousto-electromagnetic echoes received by different receiving points at the same time are combined to perform decision fusion, thereby judging each of the acoustic and electromagnetic echoes. The type of target that the electromagnetic echo corresponds to. Using the splicing method, the multi-band feature fusion is realized, and the target signal that is easier to be extracted is obtained, and the detection accuracy is improved; at the same time, for the complex multi-signal sources captured from different receiving antennas, the energy distribution within the range of the threshold is set. The multi-target separation and combination are performed to reduce the computational complexity of the multi-target detection algorithm.

According to a second aspect of the present invention, there is provided a computer-readable storage medium on which a computer program is stored, the computer program implementing the above method when executed by a computer processor.

According to a third aspect of the present invention, an indoor target rapid detection system based on acousto-electromagnetic intermodulation is proposed, the system comprising:

Acoustic electromagnetic echo generation module: configured to use acoustic wave signals to acoustically excite multiple indoor targets to induce mechanical vibrations, so as to generate acoustic excitation signals, and at the same time use the electromagnetic wave signals and the acoustic excitation signals to perform acousto-electromagnetic mutual modulation, generate acousto-electromagnetic echoes;

The acousto-electromagnetic echo feature extraction module: configured to receive the acousto-electromagnetic echo of each target in the plurality of indoor targets, and in each frequency band, based on the two dimensions of the Doppler effect and the amplitude modulation, respectively. Perform feature extraction on the acousto-electromagnetic echoes, so as to extract Doppler features and amplitude features of each of the acousto-electromagnetic echoes located in different frequency bands;

Feature fusion module: configured to perform feature fusion on the Doppler features located in different frequency bands by using the splicing method to obtain the multi-band Doppler fusion features of each target, and use the splicing method to perform feature fusion on the Doppler features located in different frequency bands. Perform feature fusion on the amplitude features of the frequency bands to obtain the multi-frequency band amplitude fusion features of each target;

Classification decision module: configured to construct a plurality of machine learning models based on different classifiers according to the multi-band Doppler fusion features and the multi-band amplitude fusion features, and use the different machine learning models for each The acousto-electromagnetic echoes are classified to obtain classification results corresponding to the different machine learning models, and then decision fusion is performed on the classification results to determine the type of the target corresponding to each of the acousto-electromagnetic echoes.

The invention combines the acoustic detection system and the electromagnetic detection system to characterize the difference between the object and its surrounding environment through the Doppler frequency shift and amplitude modulation caused by the interaction between the propagating electromagnetic wave and the wave sound source in the object, namely the acoustic electromagnetic detection method . For indoor targets, the contrast with the environment is enhanced by mechanical vibration. The synthesized reflected and scattered electromagnetic signals are modulated with properties specific to the structure and composition of the vibrating object, thereby providing information about the target structure and distinguishing it from clutter. For the extracted target information, deep learning is used for feature modeling, and then the main features of various targets are classified. By using two orthogonal detection systems of acoustics and electromagnetics, the present invention can better obtain the information of the object compared with the situation that the single-mode detection system may prevent the sound wave from penetrating the interior of the object or cannot receive the reflected wave. Using different characteristics of echo signals, the accuracy of target detection is improved, the problem of single-method detection is solved, and the target detection efficiency is improved.

Description of drawings

The accompanying drawings are included to provide a further understanding of the embodiments and are incorporated into and constitute a part of this specification. The drawings illustrate embodiments and together with the description serve to explain the principles of the invention. Other embodiments and many of the intended advantages of the embodiments will be readily recognized as they become better understood by reference to the following detailed description. Other features, objects and advantages of the present application will become more apparent by reading the detailed description of non-limiting embodiments made with reference to the following drawings:

1 is a flow chart of a method for fast detection of indoor targets based on acousto-electromagnetic intermodulation according to an embodiment of the present invention;

2 is an overall frame diagram of a method for fast detection of indoor targets based on acousto-electromagnetic intermodulation according to a specific embodiment of the present invention;

3 is a schematic diagram of attenuation channel modeling of an acoustic signal according to a specific embodiment of the present invention;

Fig. 4 is the simulation schematic diagram of the relationship between target deformation amount and detection signal power spectral density of a specific embodiment of the present invention;

Fig. 5 is the simulation schematic diagram of the influence of different acoustic excitation effects of a specific embodiment of the present invention;

6 is a schematic diagram of multi-scene feature fusion based on ensemble learning according to a specific embodiment of the present invention;

7 is a schematic diagram of an indoor target recognition scheme according to a specific embodiment of the present invention;

FIG. 8 is a frame diagram of an indoor target fast detection system based on acousto-electromagnetic intermodulation according to an embodiment of the present invention.

Detailed ways

The present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the related invention, but not to limit the invention. In addition, it should be noted that, for the convenience of description, only the parts related to the related invention are shown in the drawings.

It should be noted that the embodiments in the present application and the features of the embodiments may be combined with each other in the case of no conflict. The present application will be described in detail below with reference to the accompanying drawings and in conjunction with the embodiments.

A method for fast detection of indoor targets based on acousto-electromagnetic intermodulation according to an embodiment of the present invention, FIG. 1 shows a flowchart of a method for fast detection of indoor targets based on acousto-electromagnetic intermodulation according to an embodiment of the present invention . As shown in Figure 1, the method includes the following steps:

S1: Acoustic excitation of a plurality of indoor targets by acoustic wave signals induces mechanical vibration, so as to generate acoustic excitation signals, and at the same time, the electromagnetic wave signals and the acoustic excitation signals are used to perform acousto-electromagnetic mutual modulation to generate acoustic and electromagnetic echoes;

S2: Receive an acousto-electromagnetic echo of each of the multiple indoor targets, and in each frequency band, perform feature extraction on the acousto-electromagnetic echo based on the two dimensions of the Doppler effect and amplitude modulation, respectively, Thereby, the Doppler features and amplitude features of each of the acousto-electromagnetic echoes located in different frequency bands are extracted;

S3: Use the splicing method to perform feature fusion on the Doppler features located in different frequency bands to obtain multi-band Doppler fusion features of each target, and use the splicing method to fuse the amplitudes of the different frequency bands. The feature is characterized by feature fusion to obtain the multi-band amplitude fusion feature of each target;

S4: Construct a plurality of machine learning models based on different classifiers according to the multi-band Doppler fusion feature and the multi-band amplitude fusion feature, and use the different machine learning models to analyze each of the acousto-electromagnetic echoes respectively. Perform classification to obtain classification results corresponding to different machine learning models, and then perform decision fusion on the classification results to determine the type of the target corresponding to each of the acousto-electromagnetic echoes.

In a specific embodiment, the sound source level of the sound wave signal satisfies one-way penetration.

In a specific embodiment, the S1 specifically includes: using an acoustic wave signal to acoustically excite a plurality of indoor targets to induce mechanical vibration, so as to generate an acoustic excitation signal, and adding the acoustic excitation signal to the electromagnetic wave signal after passing through a power amplifier The electromagnetic wave signal is modulated to obtain an acoustic electromagnetic echo.

In a specific embodiment, the S2 specifically includes:

receiving the acousto-electromagnetic echoes of each of the plurality of indoor targets, pre-processing the acousto-electromagnetic echoes and then performing Fourier transform to obtain the frequency spectrum of each of the acousto-electromagnetic echoes;

In each frequency band, the features of the acousto-electromagnetic echoes are extracted based on the two dimensions of the Doppler effect and the amplitude modulation, so as to extract the Doppler features and Amplitude characteristics.

In a specific embodiment, the feature extraction of the acoustic electromagnetic echo based on the Doppler effect includes:

The phase modulation of the electromagnetic wave signal caused by the Doppler effect in the acousto-electromagnetic echo is analyzed, and the phase modulation amplitude generated by the phase modulation is obtained.

In a specific embodiment, the feature extraction of the acoustic electromagnetic echo based on the amplitude modulation includes:

The amplitude modulation of the electromagnetic wave signal caused by special relativity, path loss and radar cross section in the acoustic electromagnetic echo is analyzed respectively, and the amplitude modulation amplitude caused by special relativity, the amplitude modulation amplitude caused by path loss and Amplitude modulation amplitude due to radar cross section.

In a specific embodiment, the S4 specifically includes:

S401: Build a micro-Doppler feature library and a multi-order amplitude feature library respectively according to the multi-band Doppler fusion feature and the multi-band amplitude fusion feature, and according to the micro-Doppler feature library and the multi-order amplitude feature The feature library performs preliminary matching and judgment on the received acousto-electromagnetic echoes of the target to be measured, and obtains a preliminary judgment result;

S402: Build a micro-Doppler feature library and a multi-order amplitude feature library according to the multi-band Doppler fusion feature and the multi-band amplitude fusion feature, respectively, to construct multiple sample data sets, and use each of the sample data sets to separate Perform ensemble learning based on different classifiers to build different machine learning models;

S403: Use the different machine learning models to classify and identify the Doppler feature and the amplitude feature of the acousto-electromagnetic echo of the target to be measured to obtain a corresponding classification result, and vote on the classification result decision to get the decision result;

The preliminary judgment result and the decision result obtained after performing the S401 to the S403 on the acousto-electromagnetic echoes received by different receiving points at the same time are combined to perform decision fusion, thereby judging each of the acoustic and electromagnetic echoes. The type of target that the electromagnetic echo corresponds to.

The following uses a specific embodiment to fully describe the flow of this scheme:

1. A fast detection method for indoor targets based on acousto-electromagnetic intermodulation

FIG. 2 is an overall frame diagram of a method for fast detection of indoor targets based on acousto-electromagnetic intermodulation according to a specific embodiment of the present invention.

3 is a schematic diagram of the attenuation channel modeling of acoustic signals according to a specific embodiment of the present invention; due to the need to stimulate indoor targets, sound waves pass through the free space path from the sound source to penetrate the target environment, and the excitation of the detection target to generate signal characteristics requires certain sound source level. Therefore, a sound energy attenuation model is established for the attenuation of sound waves in different propagation environments.

FIG. 4 is a schematic diagram of a simulation of the relationship between the target deformation amount and the power spectral density of the detection signal according to a specific embodiment of the present invention.

According to Figures 2, 3 and 4, the present invention mainly integrates a multi-modal indoor target detection method integrating acoustic detection and electromagnetic detection, designs a sound wave signal whose sound source level satisfies one-way penetrability, and interacts with electromagnetic waves by exciting indoor target objects The scattered electromagnetic field is generated, and the acoustic and electromagnetic integrated detection system is built to collect the echo signal of the indoor target. After preprocessing, Fourier transform is performed to obtain the frequency domain representation of a single signal source, and the Doppler effect and amplitude modulation are used as the Select features, perform feature extraction, and perform multi-scene data feature fusion and multi-recognition model decision fusion.

In order to solve the serious attenuation effect of the acoustic channel in practice, the attenuation model of the acoustic signal is simulated under different propagation media, and the sound wave signal that can meet the unidirectional penetration is designed through the effective experimental conclusion, so as to alleviate the sound wave passing through. The phenomenon of insufficient target deformation caused by power drop after channel attenuation improves the accuracy of later target recognition. Electromagnetic wave modulation is induced by acoustically induced medium fluctuations, based on the signal propagating in the acoustically excited environment and reflected from the surface of the object, to study the phase and amplitude modulation that occurs in the comparative combined effect.

For the task of indoor target detection and recognition, the information fusion technology is further extended to the target classification process of "data-feature-recognition" to improve the indoor target detection method. The multi-scene indoor target fusion classification model structure based on ensemble learning is divided into three fusion levels according to the degree of information fusion: data-level fusion, feature-level fusion, and decision-level fusion.

Under normal circumstances, under the same magnitude and homogeneous physical field of the ship, the original data is collected from the distributed acoustic and magnetic sensors for horizontal multi-domain feature fusion, and the principal component analysis method and wavelet sub-band fusion are used. method to reduce the amount of data.

Under the premise of data-level fusion, the ship sound field signal extracts the power spectrum of ship noise, the fundamental frequency of the ship, the field strength change rate, the sound pressure level, the characteristic spectrum and other feature quantities, and then comprehensively analyzes and processes the feature vectors. Get the fused feature vector. Using the D-S reasoning algorithm with high fusion efficiency, it can identify the source of the feature level, reduce the loss of main feature data, and improve the detection accuracy.

Under the premise of feature-level fusion, combined with the time-frequency feature confidence, the signal source derived from the main eigenvalues of single-period and single-sensor is used to calculate the basic probability assignment (BPA) of multiple sources, and used as the input of decision-level fusion, and The signal source is judged, and the target signal detection result is output.

2. Comparison of different acoustic and electromagnetic modulation effects

As a means of improving the distinction between an object and its surroundings, sound waves are used to induce mechanical vibrations on the object, thereby modulating the scattered electromagnetic signal and enhancing the contrast of the excited object. The resulting reflected and scattered RF signals are modulated with properties specific to the structure and composition of the vibrating object, providing information about the target structure and distinguishing it from clutter.

Analytical solution of modulation effects introduced when high-power acoustic waves propagating through a medium induce vibrations on a structure, and compare the analytical results with the main modulation effects that arise as a direct result of object vibrations, including Doppler phase modulation and special relativity induced Amplitude modulation, path loss and radar cross section (RCS). By combining all the effects, the modulated received signal looks like this:

s(t)=α ₀ A _Acoust (t)A _PL (t)A _RCS (t)γ(t)cos[ω _RF t-φ _Doppler (t)] (1)

Among them, α ₀ is the transmission power; A _PL (t) is the amplitude modulation caused by the change of path loss; A _RCS (t) is the amplitude modulation caused by RCS; γ(t) is the amplitude modulation caused by special relativity; φ _Doppler (t ) is the phase modulation that produces the Doppler effect of the reflected signal, and its expressions are:

The amplitude modulation due to medium vibration is therefore complementary to other modulation effects, and when the vibration is caused by sound, ie ω _V = ω _A , the modulated tone appears at the same frequency.

为多普勒效应的相位调制振幅，

为狭义相对论引起的幅度调制振幅，

为路径损耗变化引起的幅度调制振幅，

为RCS引起的幅度调制振幅，

为声幅度调制振幅。To further compare the results of the modulation process, the different modulation amplitudes caused by the acoustically induced medium fluctuations are given below,

is the phase modulation amplitude of the Doppler effect,

is the amplitude modulation amplitude caused by special relativity,

is the amplitude modulation amplitude due to path loss variation,

is the amplitude modulation amplitude caused by RCS,

Modulates the amplitude for the sound amplitude.

Please refer to FIG. 5. FIG. 5 is a schematic diagram of the simulation of the influence of different acoustic excitation effects in a specific embodiment of the present invention; it can be seen that among these intermodulation effects, the amplitude modulation of the acoustic modulation and the phase modulation of the Doppler effect affect the acousto-electromagnetic interaction. The impact is the largest, which also supports the basis of subsequent feature optimization from theoretical simulation.

3. Multi-scene feature fusion based on ensemble learning

Please refer to FIG. 6 and FIG. 7 , FIG. 6 is a schematic diagram of multi-scene feature fusion based on ensemble learning according to a specific embodiment of the present invention; FIG. 7 is a schematic diagram of an indoor target recognition scheme according to a specific embodiment of the present invention.

As shown in Figure 6 and Figure 7, for the echo signals generated by the acousto-electromagnetic intermodulation effect, it is necessary to further establish a micro-Doppler feature library and a multi-order amplitude feature library through phase modulation analysis and amplitude modulation analysis. Different typical indoor scenes are fused with multi-features, and the initial matching is performed according to the spectral features and the micro-Doppler feature library. Finally, the material identification results are obtained by merging with the matching results of the multi-order amplitude sample library. At the classification decision level, a machine learning model based on ensemble learning is constructed for decision fusion.

The scene echo data of different measuring points measured at the same time are fused, and N data sets are obtained by sampling with replacement from the scene echo data set using the bootstrap method, and a model is learned on each data set. The prediction result is obtained by using the outputs of the N indoor sample classification models (as shown in Figure 6), that is, the voting method is predicted by using the N indoor sample classification models. The specific process is as follows:

(1) Select n samples from the data samples of the acousto-electromagnetic echo data by resampling (with repetitions);

(2) Establish a classifier (ID3, C4.5, CART, SVM, Logistic regression, etc.) for these n samples on the typical sample attributes in all indoor target sample libraries;

(3) Repeat the above two steps m times;

(4) Put the data on the m classifiers, and finally decide which category the acousto-electromagnetic echo data belongs to according to the voting results of the m classifiers.

The beneficial effects of the invention are as follows: the invention combines the multi-modal system fusion of acoustic detection and electromagnetic detection, fully stimulates the hidden feature information in the target object, can generate the typical contrast between the target object and the background environment as much as possible, and utilizes the target object It can overcome the defect that the single target signal is submerged by the background clutter and improve the accuracy of target detection. At the same time, the present invention filters and decomposes a large number of signals, reduces unnecessary calculation amount, improves detection efficiency, and at the same time can reduce the power consumption of target penetration wall detection equipment.

FIG. 8 shows a frame diagram of an indoor target fast detection system based on acousto-electromagnetic intermodulation according to an embodiment of the present invention. The system includes an acousto-electromagnetic echo generation module 801 , an acousto-electromagnetic echo feature extraction module 802 , a feature fusion module 803 and a classification decision module 804 .

In a specific embodiment, the acoustic electromagnetic echo generation module 801 is configured to use acoustic wave signals to acoustically excite a plurality of indoor objects to induce mechanical vibration, so as to generate acoustic excitation signals, and use the electromagnetic wave signals and the acoustic excitation signals at the same time. Perform acousto-electromagnetic mutual modulation to generate acousto-electromagnetic echoes;

The acousto-electromagnetic echo feature extraction module 802 is configured to receive the acousto-electromagnetic echo of each target in the plurality of indoor targets, and in each frequency band, based on the two dimensions of the Doppler effect and the amplitude modulation, respectively. Perform feature extraction on the acousto-electromagnetic echoes, so as to extract Doppler features and amplitude features of each of the acousto-electromagnetic echoes located in different frequency bands;

The feature fusion module 803 is configured to perform feature fusion on the Doppler features located in different frequency bands by using the splicing method to obtain the multi-band Doppler fusion features of each target, and use the splicing method to perform feature fusion on the Doppler features located in different frequency bands. Perform feature fusion on the amplitude features of the frequency bands to obtain the multi-frequency band amplitude fusion features of each target;

The classification decision module 804 is configured to construct a plurality of machine learning models based on different classifiers according to the multi-band Doppler fusion features and the multi-band amplitude fusion features, and use the different machine learning models for each The acousto-electromagnetic echoes are classified to obtain classification results corresponding to the different machine learning models, and then decision fusion is performed on the classification results to determine the type of the target corresponding to each of the acousto-electromagnetic echoes.

This system combines the acoustic detection system and the electromagnetic detection system to characterize the difference between the object and its surrounding environment through the Doppler frequency shift and amplitude modulation caused by the interaction of the propagating electromagnetic wave and the wave sound source in the object, that is, the acousto-electromagnetic detection method . For indoor targets, the contrast with the environment is enhanced by mechanical vibration. The synthesized reflected and scattered electromagnetic signals are modulated with properties specific to the structure and composition of the vibrating object, thereby providing information about the target structure and distinguishing it from clutter. For the extracted target information, deep learning is used for feature modeling, and then the main features of various targets are classified. By using two orthogonal detection systems of acoustics and electromagnetics, the present invention can better obtain the information of the object compared with the situation that the single-mode detection system may prevent the sound wave from penetrating the interior of the object or cannot receive the reflected wave. Using different characteristics of echo signals, the accuracy of target detection is improved, the problem of single-method detection is solved, and the target detection efficiency is improved.

Embodiments of the present invention also relate to a computer-readable storage medium having stored thereon a computer program that, when executed by a computer processor, implements the above method. The computer program contains program code for carrying out the method shown in the flowchart. It should be noted that the computer-readable medium of the present application may be a computer-readable signal medium or a computer-readable medium, or any combination of the above two.

The invention combines the acoustic detection system and the electromagnetic detection system to characterize the difference between the object and its surrounding environment through the Doppler frequency shift and amplitude modulation caused by the interaction between the propagating electromagnetic wave and the wave sound source in the object, namely the acoustic electromagnetic detection method . For indoor targets, the contrast with the environment is enhanced by mechanical vibration. The synthesized reflected and scattered electromagnetic signals are modulated with properties specific to the structure and composition of the vibrating object, thereby providing information about the target structure and distinguishing it from clutter. For the extracted target information, deep learning is used for feature modeling, and then the main features of various targets are classified. By using two orthogonal detection systems of acoustics and electromagnetics, the present invention can better obtain the information of the object compared with the situation that the single-mode detection system may prevent the sound wave from penetrating the interior of the object or cannot receive the reflected wave. Using different characteristics of echo signals, the accuracy of target detection is improved, the problem of single-method detection is solved, and the target detection efficiency is improved.

The above description is only a preferred embodiment of the present application and an illustration of the applied technical principles. Those skilled in the art should understand that the scope of the invention involved in this application is not limited to the technical solution formed by the specific combination of the above technical features, and should also cover the above technical features or Other technical solutions formed by any combination of its equivalent features. For example, a technical solution is formed by replacing the above-mentioned features with the technical features disclosed in this application (but not limited to) with similar functions.

Claims (9)

1. an indoor target fast detection method based on acousto-electromagnetic intermodulation, is characterized in that, comprises the following steps: S1: Acoustic excitation of a plurality of indoor targets by acoustic wave signals induces mechanical vibration, so as to generate acoustic excitation signals, and at the same time, the electromagnetic wave signals and the acoustic excitation signals are used to perform acousto-electromagnetic mutual modulation to generate acoustic and electromagnetic echoes; S2: Receive an acousto-electromagnetic echo of each of the multiple indoor targets, and in each frequency band, perform feature extraction on the acousto-electromagnetic echo based on the two dimensions of the Doppler effect and amplitude modulation, respectively, Thereby, the Doppler features and amplitude features of each of the acousto-electromagnetic echoes located in different frequency bands are extracted; S3: Use the splicing method to perform feature fusion on the Doppler features located in different frequency bands to obtain multi-band Doppler fusion features of each target, and use the splicing method to fuse the amplitudes of the different frequency bands. The feature is characterized by feature fusion to obtain the multi-band amplitude fusion feature of each target; S4: Construct a plurality of machine learning models based on different classifiers according to the multi-band Doppler fusion feature and the multi-band amplitude fusion feature, and use the different machine learning models to analyze each of the acousto-electromagnetic echoes respectively. Perform classification to obtain classification results corresponding to different machine learning models, and then perform decision fusion on the classification results to determine the type of the target corresponding to each of the acousto-electromagnetic echoes. 2 . The method according to claim 1 , wherein the sound source level of the sound wave signal satisfies one-way penetration. 3 . 3 . The method according to claim 1 , wherein the S1 specifically comprises: using acoustic wave signals to perform acoustic excitation on a plurality of indoor targets to induce mechanical vibration, so as to generate acoustic excitation signals, and use the acoustic excitation signals to generate acoustic excitation signals. 4 . After the power amplifier is added to the electromagnetic wave signal, the electromagnetic wave signal is modulated to obtain the acoustic electromagnetic echo. 4. The method according to claim 1, wherein the S2 specifically comprises: receiving the acousto-electromagnetic echoes of each of the plurality of indoor targets, pre-processing the acousto-electromagnetic echoes and then performing Fourier transform to obtain the frequency spectrum of each of the acousto-electromagnetic echoes; In each frequency band, the features of the acousto-electromagnetic echoes are extracted based on the two dimensions of the Doppler effect and the amplitude modulation, so as to extract the Doppler features and Amplitude characteristics. 5. The method according to claim 1, wherein the feature extraction of the acoustic electromagnetic echo based on the Doppler effect comprises: The phase modulation of the electromagnetic wave signal caused by the Doppler effect in the acousto-electromagnetic echo is analyzed, and the phase modulation amplitude generated by the phase modulation is obtained. 6. The method according to claim 1, wherein the feature extraction of the acousto-electromagnetic echo based on the amplitude modulation comprises: The amplitude modulation of the electromagnetic wave signal caused by special relativity, path loss and radar cross section in the acoustic electromagnetic echo is analyzed respectively, and the amplitude modulation amplitude caused by special relativity, the amplitude modulation amplitude caused by path loss and Amplitude modulation amplitude due to radar cross section. 7. The method according to claim 1, wherein the S4 specifically comprises: S401: Build a micro-Doppler feature library and a multi-order amplitude feature library respectively according to the multi-band Doppler fusion feature and the multi-band amplitude fusion feature, and according to the micro-Doppler feature library and the multi-order amplitude feature The feature library performs preliminary matching and judgment on the received acousto-electromagnetic echoes of the target to be measured, and obtains a preliminary judgment result; S402: Build a micro-Doppler feature library and a multi-order amplitude feature library according to the multi-band Doppler fusion feature and the multi-band amplitude fusion feature, respectively, to construct multiple sample data sets, and use each of the sample data sets to separate Perform ensemble learning based on different classifiers to build different machine learning models; S403: Use the different machine learning models to classify and identify the Doppler feature and the amplitude feature of the acousto-electromagnetic echo of the target to be measured to obtain a corresponding classification result, and vote on the classification result decision to get the decision result; The preliminary judgment result and the decision result obtained after performing the S401 to the S403 on the acousto-electromagnetic echoes received by different receiving points at the same time are combined to perform decision fusion, thereby judging each of the acoustic and electromagnetic echoes. The type of target the electromagnetic echo corresponds to. 8. A computer-readable storage medium on which a computer program is stored, characterized in that, when the computer program is executed by a computer processor, the method according to any one of claims 1 to 7 is implemented. 9. An indoor target rapid detection system based on acousto-electromagnetic intermodulation, characterized in that, comprising: Acoustic electromagnetic echo generation module: configured to use acoustic wave signals to acoustically excite multiple indoor targets to induce mechanical vibrations, so as to generate acoustic excitation signals, and at the same time use the electromagnetic wave signals and the acoustic excitation signals to perform acousto-electromagnetic mutual modulation, generate acousto-electromagnetic echoes; The acousto-electromagnetic echo feature extraction module: configured to receive the acousto-electromagnetic echo of each target in the plurality of indoor targets, and in each frequency band, based on the two dimensions of the Doppler effect and the amplitude modulation, respectively. Perform feature extraction on the acousto-electromagnetic echoes, so as to extract Doppler features and amplitude features of each of the acousto-electromagnetic echoes located in different frequency bands; Feature fusion module: configured to perform feature fusion on the Doppler features located in different frequency bands by using the splicing method to obtain the multi-band Doppler fusion features of each target, and use the splicing method to perform feature fusion on the Doppler features located in different frequency bands. Perform feature fusion on the amplitude features of the frequency bands to obtain the multi-frequency band amplitude fusion features of each target; Classification decision module: configured to construct a plurality of machine learning models based on different classifiers according to the multi-band Doppler fusion features and the multi-band amplitude fusion features, and use the different machine learning models for each The acousto-electromagnetic echoes are classified to obtain classification results corresponding to the different machine learning models, and then decision fusion is performed on the classification results to determine the type of the target corresponding to each of the acousto-electromagnetic echoes.

Images