Object authentication method, watermark embedding method and device for wireless sensing model

Abstract

The present invention provides an object authentication method, a watermark embedding method and a device for a wireless sensing model, which can be applied to the technical field of signal processing. The method includes: obtaining an intermediate-frequency received signal, where the intermediate-frequency received signal is obtained by mixing a transmitted signal and an echo signal; performing phase offset compensation based on a fixed weight on the intermediate-frequency received signal to generate a wireless signal spectrogram; inputting the wireless signal spectrogram into a wireless sensing model to output a target wireless signal spectrogram, where the wireless sensing model includes multiple feature extraction layers for extracting features from the wireless signal spectrogram, and at least one feature extraction layer is embedded with a pass authentication module, and the pass authentication module is used to perform phase offset compensation based on a dynamic weight on the current input to authenticate the pass of the target object; determining the authentication result of the target object according to the target wireless signal spectrogram and a preset spectrogram.

Description

Technical Field

[0001] The present invention relates to the technical field of signal processing, and particularly to an object authentication method, a watermark embedding method and a device for a wireless sensing model. Background Art

[0002] Wireless sensing refers to the process of sensing without interfering with the normal activities of the target object. A wireless sensing model is a model that uses the propagation characteristics of wireless signals, such as channel state information, etc., to realize the sensing of the target object.

[0003] In the process of implementing the present invention, it is found that in the related art, the wireless sensing model is mainly protected by embedding watermarks. However, it is difficult to cope with the attack of forged watermarks by embedding watermarks, resulting in low security of the wireless sensing model. In addition, the wireless sensing model can also be protected by using judicial means afterwards, but the efficiency of post-processing is low, and it is difficult to achieve active protection of the wireless sensing model. Summary of the Invention

[0004] In view of the above problems, the present invention provides an object authentication method, a watermark embedding method and a device for a wireless sensing model.

[0005] According to a first aspect of the present invention, there is provided an object authentication method for a wireless sensing model, including: obtaining an intermediate-frequency received signal, where the intermediate-frequency received signal is obtained by mixing a transmitted signal and an echo signal; performing phase offset compensation based on a fixed weight on the intermediate-frequency received signal to generate a wireless signal spectrogram; inputting the wireless signal spectrogram into the wireless sensing model to output a target wireless signal spectrogram, where the wireless sensing model includes a plurality of feature extraction layers for extracting features from the wireless signal spectrogram, and at least one feature extraction layer is embedded with a pass authentication module, and the pass authentication module is used to perform phase offset compensation based on a dynamic weight on the current input to authenticate the pass of the target object; determining an authentication result of the target object according to the target wireless signal spectrogram and a preset spectrogram.

[0006] According to the object authentication method of the wireless sensing model provided by the present invention, by performing phase shift compensation with a fixed weight on the obtained intermediate frequency received signal, a wireless signal spectrogram can be generated. Then, by using the wireless sensing model to process the wireless signal spectrogram, a target wireless signal spectrogram can be obtained. By comparing the target wireless signal spectrogram with a preset spectrogram, the authentication result of the target object can be obtained. Since at least one of the multiple feature extraction layers in the wireless sensing model embeds a passage authentication module, and the passage authentication module performs phase shift compensation with a dynamic weight on the current input, the pass of the target object can be authenticated. When the quality of the target wireless signal spectrogram is higher than the preset spectrogram, the target object is an authorized object; when the quality of the target wireless signal spectrogram is lower than the preset spectrogram, the target object is an unauthorized object. Therefore, when the target object is an authorized object, the quality of the target wireless signal spectrogram is improved, thereby improving the sensing accuracy of the wireless sensing model. Since the authentication of the target object can be achieved through the quality of the output target wireless signal spectrogram, active protection of the wireless sensing model is realized. Brief Description of the Drawings

[0007] Through the following description of the embodiments of the present invention with reference to the drawings, the above content and other objects, features, and advantages of the present invention will become clearer.

[0008] Figure 1 The application scenario diagram of the object authentication method of the wireless sensing model according to the embodiment of the present invention is shown.

[0009] Figure 2 The flowchart of the object authentication method of the wireless sensing model according to the embodiment of the present invention is shown.

[0010] Figure 3 The architecture diagram of the passage authentication module according to the embodiment of the present invention is shown.

[0011] Figure 4A The wireless signal spectrogram according to the embodiment of the present invention is shown.

[0012] Figure 4B The first target wireless signal spectrogram according to the embodiment of the present invention is shown.

[0013] Figure 4C The second target wireless signal spectrogram according to the embodiment of the present invention is shown.

[0014] Figure 5 The flowchart of the watermark embedding method of the wireless sensing model according to the embodiment of the present invention is shown.

[0015] Figure 6 The generation example diagram of the target string watermark according to the embodiment of the present invention is shown.

[0016] Figure 7 Shows an example diagram of the generation of a target pass according to an embodiment of the present invention.

[0017] Figure 8 Shows an example diagram of the sensing process of a wireless sensing model according to an embodiment of the present invention.

[0018] Figure 9A Shows an architecture diagram of a wireless sensing model according to an embodiment of the present invention.

[0019] Figure 9B Shows a standard module of a wireless sensing model according to an embodiment of the present invention.

[0020] Figure 9C Shows a protection module of a wireless sensing model according to an embodiment of the present invention.

[0021] Figure 10 Shows the sensing process of a wireless sensing model according to another embodiment of the present invention.

[0022] Figure 11 Shows a structural block diagram of an object authentication device of a wireless sensing model according to an embodiment of the present invention.

[0023] Figure 12 Shows a block diagram of an electronic device suitable for implementing an object authentication method of a wireless sensing model according to an embodiment of the present invention. Detailed implementation manners

[0024] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are merely exemplary and are not intended to limit the scope of the present invention. In the following detailed description, for the sake of explanation, many specific details are set forth to provide a comprehensive understanding of the embodiments of the present invention. However, obviously, one or more embodiments can also be implemented without these specific details. In addition, in the following description, descriptions of well-known structures and technologies are omitted to avoid unnecessarily obscuring the concepts of the present invention.

[0025] The terms used herein are only for describing specific embodiments and are not intended to limit the present invention. The terms "including", "comprising" and the like used herein indicate the presence of the described features, steps, operations and / or components, but do not exclude the presence or addition of one or more other features, steps, operations or components.

[0026] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification and should not be interpreted in an idealized or overly rigid manner.

[0027] In the case of using expressions such as "at least one of A, B, and C", generally, it should be interpreted according to the meaning that those skilled in the art usually understand this expression (for example, "a system having at least one of A, B, and C" should include, but not be limited to, a system having only A, only B, only C, having A and B, having A and C, having B and C, and / or having A, B, and C, etc.).

[0028] In the technical solution of the present invention, the user information involved (including but not limited to user personal information, user image information, user device information, such as location information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) are all information and data that have been authorized by the user or fully authorized by all parties. Moreover, the processing of relevant data, such as collection, storage, use, processing, transmission, provision, disclosure, and application, all comply with relevant laws, regulations, and standards, adopt necessary confidentiality measures, do not violate public order and good customs, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0029] In the process of implementing the present invention, it is found that a high-quality dataset is the key to improving the performance of learning-based wireless sensing. However, in the related technology of the wireless sensing field, there is a lack of large-scale and high-quality public data resources. Data collection often relies on developers to collect and label by themselves, resulting in a relatively high development cost of the wireless sensing model. In addition, since the wireless sensing model needs to be jointly deployed with sensing devices in Internet of Things devices, the parameters of its wireless sensing model are more likely to suffer from leakage risks, thus bringing potential economic losses to developers. Model watermarking technology is a commonly used method for protecting deep models, which can prevent unauthorized copying and abuse. By adding a watermark to the parameters of the wireless sensing model, that is, by embedding a specific identifier or information, the unique identification of the wireless sensing model can be achieved. And the watermark is usually an invisible tiny parameter change, which has little impact on the overall performance of the wireless sensing model, but is sufficient to identify the source of the wireless sensing model. When the wireless sensing model is copied or abused, the watermark can trace its source, thus providing a powerful means of copyright protection for developers. However, the watermarking technology in the related technology is difficult to effectively cope with the forged watermark attack. The forged watermark attack means that the attacker embeds his own watermark again on the wireless sensing model with a legal watermark embedded, resulting in both the original owner and the attacker of the wireless sensing model claiming to own the copyright of the wireless sensing model. In addition, it is also possible to rely on ex post judicial means to protect the wireless sensing model. However, once the parameters of the wireless sensing model are leaked, developers need to use the watermark as proof of ownership to initiate legal proceedings for liability. But this method is not efficient in actual scenarios, the judicial process and the cost of tort tracing are high, and for developers, the ex post passive protection often cannot effectively avoid losses.

[0030] In view of this, an embodiment of the present invention provides an object authentication method for a wireless sensing model, including: obtaining an intermediate-frequency received signal, where the intermediate-frequency received signal is obtained by mixing a transmitted signal and an echo signal; performing phase-offset compensation based on a fixed weight on the intermediate-frequency received signal to generate a wireless signal spectrogram; inputting the wireless signal spectrogram into the wireless sensing model to output a target wireless signal spectrogram, where the wireless sensing model includes multiple feature extraction layers for extracting features from the wireless signal spectrogram, and at least one feature extraction layer is embedded with a pass authentication module, and the pass authentication module is used to perform phase-offset compensation based on a dynamic weight on the current input to authenticate the pass of the target object; determining the authentication result of the target object according to the target wireless signal spectrogram and a preset spectrogram.

[0031] Figure 1 FIG. shows an application scenario diagram of an object authentication method for a wireless sensing model according to an embodiment of the present invention.

[0032] As Figure 1 shown, the application scenario 100 according to this embodiment may include a first terminal device 101, a second terminal device 102, a third terminal device 103, a network 104, and a server 105. The network 104 is used to provide a medium for communication links between the first terminal device 101, the second terminal device 102, the third terminal device 103, and the server 105. The network 104 may include various connection types, such as wired, wireless communication links, or fiber optic cables, etc.

[0033] Users can use the first terminal device 101, the second terminal device 102, and the third terminal device 103 to interact with the server 105 through the network 104 to receive or send messages, etc. Various communication client applications may be installed on the first terminal device 101, the second terminal device 102, and the third terminal device 103, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients, social platform software, etc. (only as examples).

[0034] The first terminal device 101, the second terminal device 102, and the third terminal device 103 may be various electronic devices with a display screen and supporting web browsing, including but not limited to smart phones, tablet computers, laptop portable computers, and desktop computers, etc.

[0035] The server 105 may be a server providing various services, such as a background management server (only as an example) that supports websites browsed by users using the first terminal device 101, the second terminal device 102, and the third terminal device 103. The background management server may analyze and process data such as received user requests, and feedback the processing results (such as web pages, information, or data obtained or generated according to user requests) to the terminal device.

[0036] It should be noted that the object authentication method of the wireless sensing model provided by the embodiments of the present invention can generally be executed by the server 105. Correspondingly, the object authentication device of the wireless sensing model provided by the embodiments of the present invention can generally be set in the server 105. The object authentication method of the wireless sensing model provided by the embodiments of the present invention can also be executed by a server or a server cluster different from the server 105 and capable of communicating with the first terminal device 101, the second terminal device 102, the third terminal device 103, and / or the server 105. Correspondingly, the object authentication device of the wireless sensing model provided by the embodiments of the present invention can also be set in a server or a server cluster different from the server 105 and capable of communicating with the first terminal device 101, the second terminal device 102, the third terminal device 103, and / or the server 105.

[0037] It should be understood that Figure 1 the numbers of the first terminal device, the second terminal device, the third terminal device, the network, and the server in

[0038] Figure 2 FIG. shows a flowchart of an object authentication method of a wireless sensing model according to an embodiment of the present invention.

[0039] As Figure 2 shown, the object authentication method 200 of the wireless sensing model in this embodiment includes operations S210 to S240.

[0040] In operation S210, an intermediate frequency received signal is acquired.

[0041] In operation S220, phase offset compensation based on fixed weights is performed on the intermediate frequency received signal to generate a wireless signal spectrogram.

[0042] In operation S230, the wireless signal spectrogram is input into the wireless sensing model to output a target wireless signal spectrogram.

[0043] In operation S240, an authentication result of the target object is determined according to the target wireless signal spectrogram and a preset spectrogram.

[0044] According to an embodiment of the present invention, the intermediate frequency received signal can be obtained by mixing the transmitted signal and the echo signal. The transmitted signal is the signal transmitted by the wireless access point using the array antenna. The transmitted signal can be a wireless signal. The echo signal is the signal reflected back to the array antenna by the transmitted signal when it encounters the target object. The wireless access point can refer to a device that allows wireless devices to connect to a wired network. For example, the wireless device can be a mobile phone, a laptop computer, a tablet computer, etc. The array antenna can represent an aggregate composed of multiple antenna elements. By controlling the phase and amplitude of the antenna elements, directional radiation of the beam can be achieved. Each antenna element can be regarded as an antenna, and the array antenna can be composed of multiple antennas.

[0045] According to an embodiment of the present invention, the intermediate frequency received signal can include K frequency points. By mixing the transmitted signal and the echo signal, the intermediate frequency received signal can be obtained as shown in the following formula (1).

[0046] (1)

[0047] Wherein, represents the intermediate frequency received signal of the m-th antenna at the k-th frequency point, represents the complex reflection coefficient of the transmitted signal by the target object located at (x, y), and j represents the imaginary unit, represents the round-trip transmission distance of the transmitted signal, represents the wavelength of the intermediate frequency received signal at the k-th frequency point.

[0048] According to an embodiment of the present invention, by coherently superimposing different antennas of the array antenna and the intermediate frequency received signals at different frequency points, phase offset compensation based on fixed weights can be achieved for the intermediate frequency received signal, thereby generating a wireless signal spectrogram.

[0049] According to an embodiment of the present invention, the fixed weights can be determined based on the echo signal and the transmitted signal. In the case where the fixed weights are determined, by performing phase offset compensation based on the fixed weights on the intermediate frequency received signal, a wireless signal spectrogram can be generated.

[0050] According to an embodiment of the present invention, the wireless sensing model can represent a model for sensing without interfering with the normal activities of the target object. For example, when the wireless sensing model senses the human body contour, the target object can be a human body. The wireless sensing model can include multiple feature extraction layers for extracting features from the wireless signal spectrogram. At least one feature extraction layer is embedded with a passage authentication module, and the passage authentication module is used to perform phase offset compensation based on dynamic weights on the current input to authenticate the pass of the target object.

[0051] For example, the wireless sensing model can be a neural network model. The feature extraction layer can be a convolutional layer, a fully connected layer, a recurrent neural network layer, an activation function, a normalization layer, etc. In addition, according to the task requirements, these feature extraction layers can be combined to form a wireless sensing model.

[0052] According to an embodiment of the present invention, by inputting the wireless signal spectrogram into the wireless sensing model, the access authentication module performs phase offset compensation based on dynamic weights on the current input to authenticate the pass of the target object, thereby outputting the target wireless signal spectrogram. The pass can be a digital matrix, for example, a two-dimensional code, a digital picture, and so on. There can be multiple target objects, and the passes of each target object are different.

[0053] According to an embodiment of the present invention, the preset spectrogram can be set according to experience or can be a wireless signal spectrogram. By comparing the quality of the target wireless signal spectrogram and the preset spectrogram, the authentication result of the target object can be determined, and the authentication result of the target object can indicate whether the target object has the permission to use the wireless sensing model.

[0054] According to an embodiment of the present invention, by performing phase offset compensation with fixed weights on the obtained intermediate-frequency received signal, a wireless signal spectrogram can be generated. Then, by using the wireless sensing model to process the wireless signal spectrogram, a target wireless signal spectrogram can be obtained. By comparing the target wireless signal spectrogram and the preset spectrogram, the authentication result of the target object can be obtained. Since at least one of the multiple feature extraction layers in the wireless sensing model embeds the access authentication module, and the access authentication module performs phase offset compensation based on dynamic weights on the current input to authenticate the pass of the target object. When the quality of the target wireless signal spectrogram is higher than the preset spectrogram, the target object is an authorized object; when the quality of the target wireless signal spectrogram is lower than the preset spectrogram, the target object is an unauthorized object. Therefore, when the target object is an authorized object, the quality of the target wireless signal spectrogram is improved, thereby improving the sensing accuracy of the wireless sensing model. Since the authentication of the target object can be achieved through the quality of the output target wireless signal spectrogram, the active protection of the wireless sensing model is realized.

[0055] According to an embodiment of the present invention, inputting the wireless signal spectrogram into the wireless sensing model and outputting the target wireless signal spectrogram includes: for the target feature extraction layer, determining the dynamic weight based on the preset pass; and performing phase offset compensation on the current input based on the dynamic weight to generate the output of the target feature extraction layer.

[0056] According to an embodiment of the present invention, the target feature extraction layer may be a feature extraction layer embedded with a pass authentication module in the wireless sensing model, and the target feature extraction layer may be one or more of multiple feature extraction layers. For example, in the case where the multiple feature extraction layers of the wireless sensing model are respectively a convolutional layer, a fully connected layer, a recurrent neural network layer, an activation function, and a normalization layer, the target feature extraction layer may be the convolutional layer.

[0057] According to an embodiment of the present invention, there may be multiple preset passes. For example, there may be 2 preset passes. For the target feature extraction layer, based on the preset passes, determining the dynamic weight includes: inputting the preset passes into the target feature extraction layer to output the first target feature; and determining the dynamic weight based on the first target feature.

[0058] According to an embodiment of the present invention, the first target feature may represent the feature obtained by the target feature extraction layer extracting features from the preset passes. By inputting the preset passes into the target feature extraction layer, the target feature extraction layer extracts features from the preset passes to obtain the first target feature. The dynamic weight may be determined based on the first target feature, and the dynamic weight has periodicity.

[0059] According to an embodiment of the present invention, the input of the first feature extraction layer is the wireless signal spectrogram, the output of the last feature extraction layer is the target wireless signal spectrogram, and the output of the previous layer of the target feature extraction layer is the current input of the target feature extraction layer.

[0060] According to an embodiment of the present invention, phase shift compensation may be performed on the current input based on the dynamic weight, thereby generating the output of the target feature extraction layer.

[0061] According to an embodiment of the present invention, the preset passes may be input into the target feature extraction layer to obtain the first target feature, and then based on the first target feature, the dynamic weight may be determined. Based on the dynamic weight, the interference of the intermediate frequency received signal can be better suppressed, the super-resolution of the signal can be completed, and a target wireless signal spectrogram with higher quality can be obtained.

[0062] According to an embodiment of the present invention, based on the dynamic weight, performing phase shift compensation on the current input to generate the output of the target feature extraction layer includes: using the target feature extraction layer to extract features from the current input to obtain the second target feature; performing inverse Fourier transform on the second target feature to obtain the original signal; and generating the output of the target feature extraction layer according to the original signal and the dynamic weight.

[0063] According to an embodiment of the present invention, the target feature extraction layer can be used to extract features from the current input, and a second target feature can be obtained. The process of extracting features from the current input can be to perform an Inverse Fast Fourier Transform (IFFT) on the current input.

[0064] According to an embodiment of the present invention, by performing a Fast Fourier Transform (FFT) on the second target feature, the original signal can be obtained.

[0065] Figure 3 The architecture diagram of the access authentication module according to an embodiment of the present invention is shown.

[0066] As Figure 3 shown, when the target feature extraction layer is a convolutional layer, the current input and the preset pass can be input into the convolutional layer, and the convolutional layer extracts features from the current input to obtain a second target feature , and then perform an Inverse Fast Fourier Transform (IFFT) on the second target feature to obtain the original signal . In addition, the convolutional layer can also extract features from the preset pass a and the preset pass p to obtain a first target feature corresponding to the preset pass a and a first target feature corresponding to the preset pass p . A dynamic weight can be generated according to the first target feature corresponding to the preset pass a and the first target feature corresponding to the preset pass p . After multiplying the dynamic weight and the original signal and performing a Fast Fourier Transform (FFT) on the result, the output of the target feature extraction layer can be obtained. According to an embodiment of the present invention, the output of the target feature extraction layer can be generated based on the original signal and the dynamic weight, including: determining the target feature of the echo signal according to the original signal and the dynamic weight; performing a Fourier transform on the target feature of the echo signal to obtain the output of the target feature extraction layer.

[0067] According to an embodiment of the present invention, when the target feature extraction layer is a convolutional layer,

[0068] denotes a convolution kernel. By performing convolution processing on the current input and the convolution kernel , a second target feature can be obtained . , as shown in the following formula (2). When there are 2 preset passes, for example, the 2 preset passes are preset pass a and preset pass p respectively. By performing convolution processing on preset pass a and the convolution kernel , the first target feature corresponding to preset pass a can be obtained , as shown in the following formula (3). By performing convolution processing on preset pass p and the convolution kernel , the first target feature corresponding to preset pass p can be obtained , as shown in the following formula (4).

[0069] (2)

[0070] (3)

[0071] (4)

[0072] According to the embodiments of the present invention, the original signal and the dynamic weight can be multiplied according to the original signal and the dynamic weight, so as to obtain the target feature of the echo signal, as shown in the following formula (5). By performing Fourier transform on the target feature of the echo signal, the output of the target feature extraction layer can be obtained, as shown in the following formula (6).

[0073] (5)

[0074] (6)

[0075] Wherein, s represents the target feature of the echo signal, represents the dynamic weight, represents the original signal, represents the inverse Fourier transform, represents the output of the target feature extraction layer, represents the Fourier transform.

[0076] According to the embodiments of the present invention, by using the target feature extraction layer to extract features from the current input, the second target feature can be obtained. Then, by performing Fourier transform on the second target feature, the original signal can be obtained. By multiplying the original signal and the dynamic weight, the target feature of the echo signal can be obtained. Then, by performing Fourier transform on the target feature of the echo signal, the output of the target feature extraction layer can be obtained, reducing the interference of the intermediate frequency received signal, realizing the phase shift compensation of the echo signal based on the dynamic weight, and improving the quality of the output of the target feature extraction layer.

[0077] According to an embodiment of the present invention, phase offset compensation based on a fixed weight is performed on an intermediate frequency received signal to generate a wireless signal spectrogram, including: determining a fixed weight according to the wavelength of the echo signal and the round-trip transmission distance of the transmitted signal; and coherently superposing the intermediate frequency received signal based on the fixed weight to generate a wireless signal spectrogram.

[0078] According to an embodiment of the present invention, the round-trip transmission distance of the transmitted signal can represent the transmission distance of the transmitted signal emitted by the array antenna after being reflected by the target object and then received by the array antenna.

[0079] According to an embodiment of the present invention, a fixed weight can be determined according to the wavelength of the echo signal and the round-trip transmission distance of the transmitted signal, and then, based on the fixed weight, the intermediate frequency received signal is coherently superposed, and a wireless signal spectrogram can be generated, as shown in the following formula (7).

[0080] (7)

[0081] Wherein, represents the wireless signal spectrogram, K represents the number of frequency points, M represents the number of antennas in the array antenna, represents the fixed weight.

[0082] According to an embodiment of the present invention, a fixed weight can be determined according to the wavelength of the echo signal and the round-trip transmission distance of the transmitted signal, and the intermediate frequency received signal is coherently superposed based on the fixed weight, so that the intermediate frequency received signal from the target position is enhanced, and the intermediate frequency received signals at other positions are suppressed, realizing the separation of the intermediate frequency received signal from the target position, thereby generating a wireless signal spectrogram.

[0083] According to an embodiment of the present invention, determining an authentication result of a target object according to the target wireless signal spectrogram and a preset spectrogram includes: determining a first authentication result of the target object when the target wireless signal spectrogram is a first target wireless signal spectrogram; and determining a second authentication result of the target object when the target wireless signal spectrogram is a second target wireless signal spectrogram.

[0084] According to an embodiment of the present invention, the preset spectrogram can be a spectrogram set according to requirements or a wireless signal spectrogram. The quality of the first target wireless signal spectrogram is higher than that of the preset spectrogram, and the first authentication result can represent that the target object is an authorized object.

[0085] According to an embodiment of the present invention, when the target wireless signal spectrogram is a first target wireless spectrogram, the quality of the first target wireless signal spectrogram and the preset spectrogram can be compared, so that the first authentication result of the target object can be determined.

[0086] According to an embodiment of the present invention, the quality of the second target wireless signal spectrogram is lower than the quality of the preset spectrogram, and the second authentication result can characterize that the target object is an unauthorized object. When the target wireless signal spectrogram is the second target wireless signal spectrogram, the quality of the second target wireless signal spectrogram and the preset spectrogram can be compared, so as to determine the second authentication result of the target object.

[0087] According to an embodiment of the present invention, by comparing the first target wireless signal spectrogram, the second target wireless signal spectrogram with the preset spectrogram, the first authentication result of the target object or the second authentication result of the target object can be obtained, so as to determine whether the target object is an authorized object, realizing the active protection of the wireless sensing model and improving the protection flexibility of the wireless sensing model.

[0088] Figure 4A The wireless signal spectrogram according to an embodiment of the present invention is shown. Figure 4B The first target wireless signal spectrogram according to an embodiment of the present invention is shown. Figure 4C The second target wireless signal spectrogram according to an embodiment of the present invention is shown.

[0089] Combined with Figures 4A - 4C , as Figure 4B shown, the quality of the first target wireless signal spectrogram is higher than the wireless signal spectrogram shown as Figure 4A , and the quality of the second target wireless signal spectrogram shown as Figure 4C is lower than the wireless signal spectrogram shown as Figure 4A , that is, the target object corresponding to the first target wireless signal spectrogram is an authorized object, and the target object corresponding to the second target wireless signal spectrogram is an unauthorized object.

[0090] Figure 5 The flowchart of the watermark embedding method of the wireless sensing model according to an embodiment of the present invention is shown.

[0091] As Figure 5 shown, the watermark embedding method 500 of the wireless sensing model in this embodiment includes operation S510 to operation S560.

[0092] In operation S510, according to the preset pass and the target feature extraction layer, the first target feature is determined.

[0093] In operation S520, according to the first target feature, the dynamic weight is determined.

[0094] In operation S530, according to the dynamic weight, a pass corresponding to the target object is generated.

[0095] In operation S540, the first target feature is averaged to obtain the mean value of the first target feature.

[0096] In operation S550, a string watermark is determined according to the mean value of the first target feature.

[0097] In operation S560, the string watermark is adjusted according to a preset watermark to obtain a target string watermark, and the target string watermark is embedded into the wireless sensing model.

[0098] According to an embodiment of the present invention, the target feature extraction layer may be a feature extraction layer in the wireless sensing model embedded with a pass authentication module. The target extraction layer may be one layer or multiple layers. For example, in the case where the wireless sensing model includes a convolutional layer, a fully connected layer, a recurrent neural network layer, an activation function, and a normalization layer, if the pass authentication module is embedded into the convolutional layer, then the convolutional layer is the target feature extraction layer.

[0099] According to an embodiment of the present invention, by inputting a preset pass into the target feature extraction layer, a first target feature can be obtained, and then a dynamic weight can be determined. Since the dynamic weight has periodicity, a pass corresponding to the target object can be determined.

[0100] According to an embodiment of the present invention, by averaging the first target feature, the mean value of the first target feature can be obtained. When the preset pass includes a preset pass a and a preset pass p, the mean values of two first target features can be obtained, as shown in the following formulas (8) and (9). Taking the positive or negative sign of the mean value of the first target feature to form a string watermark. The preset watermark is set according to requirements. Based on the preset watermark, the parameters of the wireless sensing model can be adjusted, so as to adjust the generated string watermark to obtain a target string watermark. The embedding of the target string watermark is realized by adding a watermark embedding loss function in the subtask module, as shown in the following formula (10).

[0101] (8)

[0102] (9)

[0103] (10)

[0104] Wherein, R represents a real number matrix, W represents the width of the real number matrix, H represents the height of the real number matrix, C represents the number of channels of the real number matrix, represents the mean value of the first target feature corresponding to the preset pass a of, represents the mean value of the first target feature corresponding to the preset pass p of, represents the index of the real number matrix on the width W, represents the index of the real number matrix on the height, represents the first target feature At the values of the width index and the height index, represents the first target feature At the values of the width index and the height index, represents the watermark embedding loss function, represents a control parameter, and the value of the control parameter can be 0.1, , B represents a preset watermark, , P represents the mean value of the first target feature, represents the mean value of the i-th first target feature, represents the i-th preset watermark.

[0105] Figure 6 Shows an example diagram of the generation of the target string watermark according to an embodiment of the present invention.

[0106] As Figure 6 shown, when the target feature extraction layer is a convolutional layer, inputting the preset pass a and the preset pass p into the convolutional layer can obtain the first target feature corresponding to the preset pass a and the first target feature corresponding to the preset pass p , respectively, for the first target feature corresponding to the preset pass a and the first target feature corresponding to the preset pass p take the average, and the mean value of the first target feature corresponding to the preset pass a can be obtained and the mean value of the first target feature corresponding to the preset pass p and the mean value of the first target feature corresponding to the preset pass p take the positive and negative signs, and the target string watermark 1, -1, 1,..., 1, 1, -1, -1,..., 1 can be obtained. Since the target string watermark is obtained by taking the positive and negative signs of the mean value of the first target feature corresponding to the preset pass a and the mean value of the first target feature corresponding to the preset pass p , the length of the target string is twice the number of channels of the real number matrix. and the mean value of the first target feature corresponding to the preset pass p take the positive and negative signs, so According to an embodiment of the present invention, the mean value of the first target feature corresponding to the preset pass a can also be called the mean value of the amplitude, and the mean value of the first target feature corresponding to the preset pass p

[0107] ​​​​​​​It can also be called the mean of the phase.

[0108] According to an embodiment of the present invention, based on a preset watermark, the parameters of the wireless sensing model are adjusted, so as to adjust the generated string watermark to obtain a target string watermark. Furthermore, based on the wireless sensing model with adjusted parameters, a target pass corresponding to the target object can be generated.

[0109] According to an embodiment of the present invention, different target passes can be assigned to different target objects based on the periodicity of the dynamic weight. The periodicity of the dynamic weight satisfies the following formula (11). By utilizing the periodicity of the dynamic weight, a target pass can be generated, and at the same time, the value of the dynamic weight is not changed. The target pass satisfies the following formula (12). The target pass and the preset pass have the same function. A pass can be randomly initialized for the target object, and then the parameters of the wireless sensing model are adjusted by using the gradient descent algorithm, that is, the operation of the following formula (13) is repeatedly executed until the value of the pass does not change, and it can be determined that the pass is the target pass, that is, the target pass can be generated by the wireless sensing model with adjusted parameters.

[0110] (11)

[0111] (12)

[0112] (13)

[0113] Where k represents a positive integer matrix, represents the learning rate, represents the target pass of the target object.

[0114] Figure 7 Shows an example diagram of the generation of the target pass according to an embodiment of the present invention.

[0115] Such as Figure 7 shown, when the target feature extraction layer is a convolutional layer, the preset pass a and the preset pass p are input into the convolutional layer, and the first target feature corresponding to the preset pass a can be obtained and the first target feature corresponding to the preset pass p , and then the dynamic weight can be obtained. Since the dynamic weight has periodicity, the dynamic weight satisfies . By using the gradient descent algorithm, when tends to 0, the target pass can be obtained.

[0116] According to an embodiment of the present invention, by adjusting the parameters of the target feature extraction layer in the wireless sensing model, the adjustment of the string watermark can be achieved to obtain the target string watermark, and the target string watermark is embedded into the wireless sensing model based on the watermark embedding loss function of the above formula (10), so as to generate a target pass corresponding to the target object based on the wireless sensing model embedded with the target string watermark.

[0117] According to an embodiment of the present invention, by inputting a preset pass into the target feature extraction layer, the first target feature can be determined, thereby the dynamic weight can be obtained, and then a pass corresponding to the target object is generated. By taking the average of the first target features, the mean value of the first target features can be obtained. Taking the mean value of the first target features to form a string watermark, and adjusting the string watermark based on a preset watermark, thereby adjusting the parameters of the wireless sensing model to obtain a target string, and embedding the target string watermark into the wireless sensing model, so that the wireless sensing model embedded with the target string watermark generates a target pass, thereby making the target object with the target pass a privileged object, realizing the active protection of the wireless sensing model.

[0118] Figure 8 An example diagram of the sensing process of the wireless sensing model according to an embodiment of the present invention is shown.

[0119] As Figure 8 shown, by preprocessing the echo signal, a wireless sensing signal spectrogram can be obtained. Among them, the preprocessing of the echo signal can be mixing the echo signal with the transmitted signal, and its purpose is to obtain an intermediate-frequency received signal. By performing phase offset compensation on the intermediate-frequency received signal based on a fixed weight, a wireless signal spectrogram can be obtained. Then, the wireless sensing signal spectrogram is input into the wireless sensing model. The wireless sensing model generates different dynamic weights for different subtask modules, such as the 3D pose estimation subtask module and the human body contour generation subtask module, and performs phase offset compensation on the wireless signal spectrogram based on different dynamic weights, thereby generating a target wireless signal spectrogram, and further determining whether the target object using the wireless sensing model is a privileged object.

[0120] Figure 9A An architecture diagram of the wireless sensing model according to an embodiment of the present invention is shown. Figure 9B A standard module of the wireless sensing model according to an embodiment of the present invention is shown. Figure 9C A protection module of the wireless sensing model according to an embodiment of the present invention is shown.

[0121] In one embodiment, the subtask module of the wireless sensing model can be 3D pose estimation. 3D pose estimation is a structure prediction task, and ConvNext is used as the backbone network. As Figure 9AAs shown, where the wireless signal can be an intermediate-frequency received signal, ConvNext can contain 12 basic modules. The standard ConvNext uses 12 standard modules, such as Figure 9B shown. The first 9 modules of the protected ConvNext use standard modules, and the last 3 use protected modules embedded with multi-user target pass authentication, such as Figure 9C shown. The backbone network ConvNext is used to extract the features of the intermediate-frequency received signal, and then the extracted features are input into the 3D pose estimation subtask module to complete pose estimation. The performance metric is the average joint error. The dataset contains 47,304 samples in the training set and 22,192 samples in the test set. The wireless sensing model is trained on the training set, and the performance of the wireless sensing model is evaluated on the test set. The performance results show that the error of the 3D pose estimation model using the standard ConvNext as the backbone network is 90.89 mm, and the error of the 3D pose estimation model using the protected ConvNext as the backbone network is 87.46 mm. The pass authentication module not only does not affect the performance of the wireless sensing model, but can also improve the performance of the wireless sensing model because the pass authentication module can provide super-resolved features. In the case where the target object is an object without permission, that is, when a wrong pass is given, the error of the 3D pose estimation model is 983.45 mm, thus demonstrating the effectiveness of object authentication. When and only when the target object is an authorized object, that is, when the correct pass is given, the wireless sensing model can work properly, while with a wrong pass, the performance of the wireless sensing model drops significantly. When the target pass is input, the error of the 3D pose estimation model is 87.18 mm, and the performance is consistent with the preset pass, and there is a slight performance improvement, that is, the generation process of the multi-user target pass is lossless. When a forged pass attack is carried out on the 3D pose estimation model, that is, assuming that the attacker knows the parameters of the 3D pose estimation model and can obtain the complete training data, by freezing the parameters of the 3D pose estimation model, randomly initializing the pass, and then using the complete dataset and updating the pass using gradients to obtain a forged pass, the error of the 3D pose estimation model is 147.53 mm. Therefore, the pass authentication module can well resist reverse engineering or forged pass attacks. A pruning attack is carried out on the 3D pose estimation model, and then the matching degree of the watermark is tested. Assuming that the attacker obtains the complete parameters of the model and obtains a new model by pruning the redundant parameters of the model, under different pruning rates, the pose error and watermark accuracy are as shown in Table 1 below.

[0122] Table 1 Pose error and watermark accuracy table under different pruning rates

[0123]

[0124] As shown in Table 1, even when the pruning rate is 90%, the watermark accuracy rate is still as high as 85.10%.

[0125] In one embodiment, a fine-tuning attack is adopted on the wireless sensing model. That is, the backbone network parameters of the wireless sensing model whose subtask is changed to human contour generation are used to initialize the backbone network of the 3D pose estimation model, and then further fine-tuning is performed using the dataset of 3D pose estimation. The watermark accuracy rate is 100%, and the error of the 3D pose estimation model is 99.59 mm. That is, even when the model is used for other subtasks, the success rate of detecting the embedded watermark is still very high. Therefore, the 3D pose estimation model can be well protected.

[0126] In one embodiment, the subtask module of the wireless sensing model can be human contour generation. Human contour generation is a dense prediction task, and ConvNext is used as the backbone network. As Figure 9A shown, ConvNext contains 12 basic modules. The standard ConvNext uses 12 standard modules. As Figure 9B shown, for the protected ConvNext, the first 9 modules use standard modules, and the last 3 use protection modules with multi-user pass authentication. As Figure 9CAs shown in the figure. The backbone network ConvNext is used to extract the features of the intermediate-frequency received signal, and then input into the decoder module to generate the human body contour. The performance measurement index is the Intersection over Union (IoU). The dataset contains 17,520 samples in the training set and 4,675 samples in the test set. The wireless sensing model is trained on the training set, and the performance of the wireless sensing model is evaluated on the test set. The performance results show that the performance of the human body contour generation model using the standard ConvNext as the backbone network is 0.683 IoU, and the error of the human body contour generation model using the protected ConvNext as the backbone network is 0.684 IoU. Therefore, the access authentication module not only does not affect the performance of the human body contour generation model, but can instead improve the performance of the human body contour generation model because the pass authentication module can provide super-resolved features. In the case where the target object is an object without permission, that is, when an incorrect pass is given, the performance of the human body contour generation model is 0.283 IoU. Therefore, it can be shown that the object authentication is effective. Only when the target object is an authorized object can the human body contour generation model work properly, and with an incorrect pass, the performance of the human body contour generation model drops significantly. When the target pass is input, the performance of the human body contour generation model is 0.685 IoU, and the performance is consistent with the initial pass, and there is a slight performance improvement, that is, the generation process of the target passes for multiple users is lossless. By launching a forged pass attack on the human body contour generation model, that is, assuming that the attacker knows the details of the human body contour generation model and can obtain the complete training data, by freezing the parameters of the human body contour generation model, randomly initializing the pass, and then using the complete dataset and updating the pass using the gradient to obtain a forged pass, the performance of the human body contour generation model is 0.536 IoU, which indicates that the pass authentication module can well resist reverse engineering or forged pass attacks. A pruning attack is carried out on the human body contour generation model, and then the matching degree of the watermark is tested. Assuming that the attacker obtains the complete parameters of the wireless sensing model and obtains a new model by pruning the redundant parameters of the model, in the case of different pruning rates, the contour generation and watermark accuracy rates are shown in Table 2 below.

[0127] Table 2 Contour generation and watermark accuracy rates under different pruning rates

[0128]

[0129] As shown in Table 2, even when the pruning rate is 90%, the watermark accuracy rate is still as high as 99.50%.

[0130] In one embodiment, a fine-tuning attack is adopted on the human body contour generation model, that is, the subtask is changed to the backbone network parameters of 3D pose estimation to initialize the backbone network of the human body contour generation model, and then further fine-tuning is performed using the dataset of contour generation. The watermark accuracy rate is 100%, and the performance of the human body contour generation model is 0.642 IoU. Therefore, it shows that even when the human body contour generation model is used for other tasks, the success rate of detecting the embedded watermark is still very high.

[0131] Figure 10 FIG. 4 shows the sensing process of the wireless sensing model according to another embodiment of the present invention.

[0132] As Figure 10 shown, by inputting the wireless sensing signal spectrogram into the wireless sensing model, when a target object with permission uses the wireless sensing model, a target wireless signal spectrogram with a quality higher than the preset spectrogram can be generated, that is, the performance of the wireless sensing model is normal. When a target object without permission uses the wireless sensing model, a target wireless signal spectrogram with a quality lower than the preset spectrogram can be generated, that is, the performance of the wireless sensing model deteriorates.

[0133] Based on the above object authentication method of the wireless sensing model, the present invention also provides an object authentication device for the wireless sensing model. The following will be combined with Figure 11 to describe this device in detail.

[0134] Figure 11 FIG. 5 shows a structural block diagram of an object authentication device for a wireless sensing model according to an embodiment of the present invention.

[0135] As Figure 11 shown, the object authentication device 1100 of the wireless sensing model in this embodiment includes an acquisition module 1110, a first compensation module 1120, a second compensation module 1130, and a determination module 1140.

[0136] The acquisition module 1110 is configured to acquire an intermediate frequency received signal, where the intermediate frequency received signal is obtained by mixing a transmitted signal and an echo signal. In one embodiment, the acquisition module 1110 can be used to perform the operation S210 described above, which will not be elaborated here.

[0137] The first compensation module 1120 is configured to perform phase offset compensation based on a fixed weight on the intermediate frequency received signal to generate a wireless signal spectrogram. In one embodiment, the first compensation module 1120 can be used to perform the operation S220 described above, which will not be elaborated here.

[0138] The second compensation module 1130 is configured to input the wireless signal spectrogram into the wireless sensing model and output the target wireless signal spectrogram. The wireless sensing model includes multiple feature extraction layers for extracting features from the wireless signal spectrogram, and at least one feature extraction layer is embedded with a pass authentication module. The pass authentication module is configured to perform phase shift compensation based on dynamic weights on the current input to authenticate the pass of the target object. In one embodiment, the second compensation module 1130 can be used to execute the operation S230 described above, which will not be elaborated here.

[0139] The determination module 1140 is configured to determine the authentication result of the target object according to the target wireless signal spectrogram and the preset spectrogram. In one embodiment, the determination module 1140 can be used to execute the operation S240 described above, which will not be elaborated here.

[0140] According to an embodiment of the present invention, the second compensation module 1130 includes: a first compensation sub-module and a second compensation sub-module.

[0141] The first compensation sub-module is configured to determine dynamic weights for the target feature extraction layer based on the preset pass, where the target feature extraction layer is the feature extraction layer in the wireless sensing model embedded with the pass authentication module.

[0142] The second compensation sub-module is configured to perform phase shift compensation on the current input based on the dynamic weights to generate the output of the target feature extraction layer. The input of the first feature extraction layer is the wireless signal spectrogram, the output of the last feature extraction layer is the target wireless signal spectrogram, and the output of the previous layer of the target feature extraction layer is the current input of the target feature extraction layer.

[0143] According to an embodiment of the present invention, the first compensation sub-module includes: a first compensation unit and a second compensation unit.

[0144] The first compensation unit is configured to input the preset pass into the target feature extraction layer and output the first target feature, where the first target feature represents the feature obtained by the target feature extraction layer extracting features from the preset pass.

[0145] The second compensation unit is configured to determine dynamic weights based on the first target feature.

[0146] According to an embodiment of the present invention, the second compensation sub-module includes: a third compensation unit, a fourth compensation unit, and a fifth compensation unit.

[0147] The third compensation unit is configured to extract features from the current input using the target feature extraction layer to obtain the second target feature.

[0148] The fourth compensation unit is configured to perform an inverse Fourier transform on the second target feature to obtain the original signal.

[0149] A fifth compensation unit, configured to generate an output of a target feature extraction layer according to an original signal and a dynamic weight.

[0150] According to an embodiment of the present invention, the fifth compensation unit includes: a first compensation subunit and a second compensation subunit.

[0151] The first compensation subunit is configured to determine a target feature of an echo signal according to the original signal and the dynamic weight.

[0152] The second compensation subunit is configured to perform a Fourier transform on the target feature of the echo signal to obtain an output of the target feature extraction layer.

[0153] According to an embodiment of the present invention, the first compensation module 1120 includes: a third compensation sub-module and a fourth compensation sub-module.

[0154] The third compensation sub-module is configured to determine a fixed weight according to the wavelength of the echo signal and the round-trip transmission distance of the transmitted signal, where the round-trip transmission distance of the transmitted signal represents the transmission distance of the transmitted signal emitted by the array antenna after being reflected by the target object and then received by the array antenna.

[0155] The fourth compensation sub-module is configured to perform coherent superposition on the intermediate-frequency received signal based on the fixed weight to generate a wireless signal spectrogram.

[0156] According to an embodiment of the present invention, the determination module 1140 includes: a first determination sub-module and a second determination sub-module.

[0157] The first determination sub-module is configured to determine a first authentication result of the target object when the target wireless signal spectrogram is a first target wireless signal spectrogram, where the quality of the first target wireless signal spectrogram is higher than the quality of a preset spectrogram, and the first authentication result represents that the target object is an authorized object.

[0158] The second determination sub-module is configured to determine a second authentication result of the target object when the target wireless signal spectrogram is a second target wireless signal spectrogram, where the quality of the second target wireless signal spectrogram is lower than the quality of the preset spectrogram, and the second authentication result represents that the target object is an unauthorized object.

[0159] According to an embodiment of the present invention, any of the acquisition module 1110, the first compensation module 1120, the second compensation module 1130, and the determination module 1140 can be combined and implemented in one module, or any one of them can be split into multiple modules. Alternatively, at least part of the functions of one or more of these modules can be combined with at least part of the functions of other modules and implemented in one module. According to an embodiment of the present invention, at least one of the acquisition module 1110, the first compensation module 1120, the second compensation module 1130, and the determination module 1140 can be at least partially implemented as a hardware circuit, such as a field programmable gate array (FPGA), a programmable logic array (PLA), a system on chip, a system on a substrate, a system in a package, an application specific integrated circuit (ASIC), or any other reasonable way of integrating or packaging circuits, etc., implemented by hardware or firmware, or implemented in any one of the three implementation manners of software, hardware, and firmware, or in an appropriate combination of any several of them. Alternatively, at least one of the acquisition module 1110, the first compensation module 1120, the second compensation module 1130, and the determination module 1140 can be at least partially implemented as a computer program module, which can perform corresponding functions when the computer program module is run.

[0160] Figure 12 A block diagram of an electronic device suitable for implementing an object authentication method for a wireless sensing model according to an embodiment of the present invention is shown.

[0161] As Figure 12 shown, the electronic device 1200 according to an embodiment of the present invention includes a processor 1201, which can perform various appropriate actions and processes according to a program stored in a read only memory (ROM) 1202 or a program loaded from a storage section 1208 into a RAM (Random Access Memory). The processor 1201 can include, for example, a general microprocessor (such as a CPU), an instruction set processor, and / or a related chipset, and / or a dedicated microprocessor (such as an application specific integrated circuit (ASIC)), etc. The processor 1201 can also include on-board memory for caching purposes. The processor 1201 can include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of the present invention.

[0162] In the RAM 1203, various programs and data required for the operation of the electronic device 1200 are stored. The processor 1201, the ROM 1202, and the RAM 1203 are connected to each other via a bus 1204. The processor 1201 performs various operations of the method flow according to the embodiments of the present invention by executing programs in the ROM 1202 and / or the RAM 1203. It should be noted that the programs may also be stored in one or more memories other than the ROM 1202 and the RAM 1203. The processor 1201 may also perform various operations of the method flow according to the embodiments of the present invention by executing programs stored in the one or more memories.

[0163] According to an embodiment of the present invention, the electronic device 1200 may further include an input / output (I / O) interface 1205, and the input / output (I / O) interface 1205 is also connected to the bus 1204. The electronic device 1200 may further include one or more of the following components connected to the input / output (I / O) interface 1205: an input portion 1206 including a keyboard, a mouse, etc.; an output portion 1207 including, for example, a cathode ray tube (CRT), a liquid crystal display (LCD), etc. and a speaker, etc.; a storage portion 1208 including a hard disk, etc.; and a communication portion 1209 including a network interface card such as a LAN card, a modem, etc. The communication portion 1209 performs communication processing via a network such as the Internet. A drive 1210 is also connected to the input / output (I / O) interface 1205 as needed. A removable medium 1211, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 1210 as needed so that a computer program read from it can be installed into the storage portion 1208 as needed.

[0164] The present invention also provides a computer-readable storage medium, which may be included in the device / apparatus / system described in the above embodiments; or may exist separately without being assembled into the device / apparatus / system. The above computer-readable storage medium carries one or more programs, and when the one or more programs are executed, the method according to the embodiments of the present invention is implemented.

[0165] According to an embodiment of the present invention, the computer-readable storage medium may be a non-volatile computer-readable storage medium, for example, it may include but is not limited to: portable computer disks, hard disks, RAM (Random Access Memory), ROM (Read Only Memory), erasable programmable read-only memory (EPROM or flash memory), portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the above. In the present invention, the computer-readable storage medium may be any tangible medium that contains or stores a program, and this program can be used by or in combination with an instruction execution system, apparatus, or device. For example, according to an embodiment of the present invention, the computer-readable storage medium may include one or more memories other than the above-described ROM 1202 and / or RAM 1203 and / or ROM 1202 and RAM 1203.

[0166] An embodiment of the present invention further includes a computer program product, which includes a computer program that contains program code for executing the method shown in the flowchart. When the computer program product runs in a computer system, the program code is used to enable the computer system to implement the object authentication method of the wireless sensing model provided by the embodiment of the present invention.

[0167] When the computer program is executed by the processor 1201, it executes the above functions defined in the system / apparatus of the embodiment of the present invention. According to an embodiment of the present invention, the above-described systems, apparatuses, modules, units, etc. can be implemented by computer program modules.

[0168] In one embodiment, the computer program may rely on tangible storage media such as optical storage devices and magnetic storage devices. In another embodiment, the computer program may also be transmitted and distributed in the form of a signal on a network medium, and is downloaded and installed through the communication part 1209, and / or installed from the removable medium 1211. The program code contained in the computer program can be transmitted by any suitable network medium, including but not limited to: wireless, wired, etc., or any suitable combination of the above.

[0169] In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 1209, and / or installed from the removable medium 1211. When the computer program is executed by the processor 1201, it executes the above functions defined in the system of the embodiment of the present invention. According to an embodiment of the present invention, the above-described systems, devices, apparatuses, modules, units, etc. can be implemented by computer program modules.

[0170] According to embodiments of the present invention, program code for executing the computer programs provided by the embodiments of the present invention can be written in any combination of one or more programming languages. Specifically, these computing programs can be implemented using high-level procedural and / or object-oriented programming languages, and / or assembly / machine languages. Programming languages include, but are not limited to, such as Java, C++, Python, the "C" language, or similar programming languages. The program code can be executed entirely on the user's computing device, partially on the user's device, partially on a remote computing device, or entirely on a remote computing device or server. In the case of a remote computing device, the remote computing device can be connected to the user's computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (e.g., by using an Internet service provider to connect through the Internet).

[0171] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architectures, functions, and operations of systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in the flowchart or block diagram may represent a module, a program segment, or a part of code, and the above-mentioned module, program segment, or part of code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than that marked in the accompanying drawings. For example, two consecutive blocks shown may actually be executed substantially in parallel, and they may sometimes be executed in the reverse order, depending on the functions involved. It should also be noted that each block in the block diagram or flowchart, and the combination of blocks in the block diagram or flowchart, can be implemented by a dedicated hardware-based system for performing the specified functions or operations, or can be implemented by a combination of dedicated hardware and computer instructions.

[0172] Those skilled in the art can understand that the features described in the various embodiments of the present invention can be combined and / or combined in various ways, even if such combinations or combinations are not explicitly described in the present invention. In particular, without departing from the spirit and teachings of the present invention, the features described in the various embodiments of the present invention can be combined and / or combined in various ways. All such combinations and / or combinations fall within the scope of the present invention.

[0173] The above describes the embodiments of the present invention. However, these embodiments are only for illustrative purposes and not for limiting the scope of the present invention. Although the embodiments are described separately above, this does not mean that the measures in the various embodiments cannot be used advantageously in combination. Without departing from the scope of the present invention, those skilled in the art can make various substitutions and modifications, and all such substitutions and modifications should fall within the scope of the present invention.

Claims

1. An object authentication method for a wireless sensing model, characterized in that, The method includes: Obtaining an intermediate-frequency received signal, where the intermediate-frequency received signal is obtained by mixing a transmitted signal and an echo signal; Performing phase-offset compensation based on a fixed weight on the intermediate-frequency received signal to generate a wireless signal spectrogram; Inputting the wireless signal spectrogram into a wireless sensing model, and the output of the target wireless signal spectrogram includes: for a target feature extraction layer, determining a dynamic weight based on a preset pass, where the target feature extraction layer is a feature extraction layer in the wireless sensing model embedded with a pass authentication module; using the target feature extraction layer to perform feature extraction on the current input to obtain a second target feature; performing an inverse Fourier transform on the second target feature to obtain an original signal; determining the target feature of the echo signal according to the original signal and the dynamic weight; performing a Fourier transform on the target feature of the echo signal to obtain the output of the target feature extraction layer, where the wireless sensing model includes multiple feature extraction layers for performing feature extraction on the wireless signal spectrogram, and at least one of the feature extraction layers is embedded with a pass authentication module, and the pass authentication module is used to perform phase-offset compensation based on a dynamic weight on the current input to authenticate the pass of the target object; Determining the authentication result of the target object according to the target wireless signal spectrogram and a preset spectrogram; When the authentication result of the target object indicates that the target object is an authorized object, the pass corresponding to the target object is calculated according to the following operations: When the target feature extraction layer is a convolutional layer, input the preset pass a and the preset pass p into the convolutional layer to obtain the first target feature corresponding to the preset pass a and the first target feature corresponding to the preset pass p , and then obtain the dynamic weight . The dynamic weight satisfies . When tends to 0, obtain the pass corresponding to the target object , where , j represents the imaginary unit represents the convolutional kernel, and k represents a positive integer matrix 2. The method according to claim 1, characterized in that, The input of the first feature extraction layer is the wireless signal spectrogram, the output of the last feature extraction layer is the target wireless signal spectrogram, and the output of the layer before the target feature extraction layer is the current input of the target feature extraction layer.

3. The method according to claim 2, characterized in that, The determining the dynamic weight based on a preset pass for the target feature extraction layer includes: Inputting the preset pass into the target feature extraction layer to output a first target feature, where the first target feature represents the feature obtained by the target feature extraction layer performing feature extraction on the preset pass; Determining the dynamic weight based on the first target feature.

4. The method according to claim 1, wherein The performing phase-offset compensation based on a fixed weight on the intermediate-frequency received signal to generate a wireless signal spectrogram includes: Determining the fixed weight according to the wavelength of the echo signal and the round-trip transmission distance of the transmitted signal, where the round-trip transmission distance of the transmitted signal represents the transmission distance that the transmitted signal emitted by the array antenna is reflected by the target object and then received by the array antenna; Performing coherent superposition on the intermediate-frequency received signal based on the fixed weight to generate the wireless signal spectrogram.

5. The method according to claim 1, characterized in that, The determining the authentication result of the target object according to the target wireless signal spectrogram and a preset spectrogram includes: When the target wireless signal spectrogram is a first target wireless signal spectrogram, determining a first authentication result of the target object, where the quality of the first target wireless signal spectrogram is higher than the quality of the preset spectrogram, and the first authentication result indicates that the target object is an authorized object; When the target wireless signal spectrogram is the second target wireless signal spectrogram, determine the second authentication result of the target object, where the quality of the second target wireless signal spectrogram is lower than the quality of the preset spectrogram, and the second authentication result indicates that the target object is an unauthorized object.

6. A watermark embedding method for a wireless sensing model, characterized in that, The method includes: When the target feature extraction layer is a convolutional layer, input the preset pass a and the preset pass p into the convolutional layer to obtain the first target feature corresponding to the preset pass a and the first target feature corresponding to the preset pass p , and further obtain the dynamic weight , where the target feature extraction layer is the feature extraction layer in the wireless sensing model embedded with a pass authentication module; The dynamic weight satisfies , in when approaching 0, a pass corresponding to the target object is obtained , where , j represents the imaginary unit represents the convolution kernel, and k represents a positive integer matrix; For the first target feature corresponding to the preset pass a and the first target feature corresponding to the preset pass p For each of the first target features, average the first target feature to obtain the mean value of the first target feature; Determine a string watermark according to the mean value of the first target feature; Adjust the string watermark according to a preset watermark to obtain a target string watermark, and embed the target string watermark into the wireless sensing model to obtain the wireless sensing model according to any one of claims 1-5.

7. An object authentication device for a wireless sensing model, characterized in that The device includes: An acquisition module, configured to acquire an intermediate frequency received signal, where the intermediate frequency received signal is obtained by mixing a transmitted signal and an echo signal; A first compensation module, configured to perform phase offset compensation based on a fixed weight on the intermediate frequency received signal to generate a wireless signal spectrogram; A second compensation module, configured to input the wireless signal spectrogram into a wireless sensing model and output a target wireless signal spectrogram, where the wireless sensing model includes a plurality of feature extraction layers for extracting features from the wireless signal spectrogram, and at least one of the feature extraction layers is embedded with a passage authentication module, and the passage authentication module is configured to perform phase offset compensation based on a dynamic weight on the current input to authenticate the pass of the target object; The second compensation module includes: a first compensation sub-module and a second compensation sub-module; The first compensation sub-module is configured to determine a dynamic weight based on a preset pass for a target feature extraction layer, where the target feature extraction layer is a feature extraction layer in the wireless sensing model embedded with a passage authentication module; The second compensation sub-module is configured to perform phase offset compensation on the current input based on the dynamic weight to generate a target feature; The second compensation sub-module includes: a third compensation unit, a fourth compensation unit, and a fifth compensation unit; The third compensation unit is configured to extract features from the current input by using the target feature extraction layer to obtain a second target feature; The fourth compensation unit is configured to perform an inverse Fourier transform on the second target feature to obtain an original signal; The fifth compensation unit is configured to generate an output of the target feature extraction layer according to the original signal and the dynamic weight; The fifth compensation unit includes: a first compensation sub-unit and a second compensation sub-unit; The first compensation sub-unit is configured to determine a target feature of the echo signal according to the original signal and the dynamic weight; The second compensation sub-unit is configured to perform a Fourier transform on the target feature of the echo signal to obtain an output of the target feature extraction layer; A determination module, configured to determine an authentication result of the target object according to the target wireless signal spectrogram and a preset spectrogram; When the authentication result of the target object indicates that the target object is an authorized object, the pass corresponding to the target object is calculated according to the following operations: When the target feature extraction layer is a convolutional layer, input the preset pass a and the preset pass p into the convolutional layer to obtain the first target feature corresponding to the preset pass a and the first target feature corresponding to the preset pass p , and then obtain the dynamic weight . The dynamic weight satisfies . When approaches 0, obtain the pass corresponding to the target object , where , j represents the imaginary unit represents the convolution kernel, and k represents a positive integer matrix 8. An electronic device, including: One or more processors; A memory, configured to store one or more computer programs, It is characterized in that the one or more processors execute the one or more computer programs to implement the steps of the method according to any one of claims 1 to 5 or 6.

Images