Data security transmission method and system of vehicle-mounted OBU (On Board Unit)

Abstract

The invention relates to the field of data security transmission methods, in particular to a data security transmission method and system for a vehicle-mounted OBU, and the method comprises the steps: collecting a sound wave signal in a vehicle in real time through a vehicle-mounted OBU microphone array, loading the sound wave signal to a first analysis model, and obtaining a target sound wave feature; comparing and judging the target sound wave characteristics with a preset pet sound wave resonance characteristic template stored in a local on-board unit (OBU) through a second analysis model, performing interference signal filtering on the pet sound waves to obtain filtered OBU data when the target sound wave characteristics accord with the preset pet sound wave resonance characteristic template, and performing encryption processing on the filtered OBU data, the encrypted data is safely transmitted to the background server in a wireless communication mode, so that the problem of data safe transmission caused by voiceprint false triggering is solved.

Description

Technical Field

[0001] The present invention relates to the field of data security transmission methods, in particular to a data security transmission method and system for an on-vehicle OBU. Background Art

[0002] In intelligent transportation systems, the on-board OBU serves as the core device for communication between vehicles and roadside units (RSUs), cloud platforms, and other vehicles. The security of its data transmission is directly related to the real-time and reliability of emergency response. Especially in the event of sudden traffic incidents, the OBU needs to receive and process key warning information with extremely low latency, usually within 50 milliseconds, to trigger automatic braking, path avoidance, and other operations. However, traditional on-board OBU data security transmission methods focus on network and protocol layer protection and lack protection against physical layer acoustic interference. In pet transportation scenarios, the problem of pet sound wave resonance leads to false triggering of voiceprints, such as causing door opening and emergency braking. Therefore, how to solve the false triggering of pet voiceprints during data transmission and ensure the secure transmission of on-board OBU data has become an urgent problem to be solved. Summary of the Invention

[0003] The purpose of the present invention is to provide a method and system for secure data transmission of an on-board OBU to solve the problem of false triggering of voiceprints caused by acoustic resonance of pets in pet transport vehicle scenarios.

[0004] To achieve the above object, on the one hand, the present invention provides a method for securely transmitting data of an on-board OBU, the method comprising:

[0005] The vehicle's internal sound wave signals are collected in real time through the on-board OBU microphone array, and the sound wave signals are loaded into the first analysis model to obtain target sound wave characteristics, which include frequency characteristics, time domain envelope characteristics, energy distribution characteristics and duration.

[0006] The target sound wave feature is compared with the preset pet sound wave resonance feature template stored locally in the vehicle-mounted OBU through a second analysis model. When the target sound wave feature meets the preset pet sound wave resonance feature template, the pet sound wave is filtered for interference signals to obtain filtered OBU data.

[0007] The filtered OBU data is encrypted and securely transmitted to the backend server via wireless communication.

[0008] Furthermore, the method of comparing and judging the target sound wave feature with a preset pet sound wave resonance feature template stored locally in the vehicle-mounted OBU through a second analysis model includes:

[0009] The similarity S is calculated based on the frequency characteristics of the sound wave signal and the preset pet sound wave frequency characteristics:

[0010]

[0011] Among them F r (i) is the eigenvalue corresponding to the i-th frequency component in the frequency characteristics of the real-time acquired sound wave signal, F p (i) is the eigenvalue corresponding to the i-th frequency component in the preset frequency characteristics of the pet sound wave, n is the total number of frequency components, and S is the similarity.

[0012] The correlation coefficient C is calculated based on the time domain envelope characteristics of the sound wave signal and the preset time domain envelope characteristics of the pet sound wave:

[0013]

[0014] Among them E r (t) Real-time acquisition of the time domain envelope value of the sound wave at the tth time point, is the mean value of the real-time acquisition of the acoustic wave time domain envelope, E p (t) is the time domain envelope eigenvalue of the preset pet sound wave at the time point, is the mean of the preset pet sound wave time domain envelope feature, N is the total number of time points, and C is the correlation coefficient.

[0015] The Euclidean distance is calculated based on the energy distribution characteristics of the sound wave and the preset energy distribution characteristics of the pet sound wave:

[0016]

[0017] Among them, P r (j) is the energy value of the real-time collected sound wave in the jth frequency band, P P (j) is the energy value of the preset pet sound wave in the jth frequency band, and D is the Euclidean distance.

[0018] When the similarity, correlation coefficient, and Euclidean distance all meet the preset judgment threshold and the duration of the real-time collected sound wave signal is within the preset pet sound wave duration range, it is determined that the current sound wave meets the preset pet sound wave resonance characteristics.

[0019] Furthermore, the method of loading the acoustic wave signal into the first analysis model to obtain the target acoustic wave feature includes:

[0020] The acoustic wave signal is segmented to obtain the original frame sequence, and the original frame sequence is windowed by the window function to obtain the windowed acoustic wave data:

[0021] x ′ (n) = x(n)·w(n).

[0022] Where x(n) is the original frame sequence, w(n) is the window function, x ′ (n) is the windowed sound wave data.

[0023] The windowed sound wave data is input into the feature extraction model to obtain frequency features, time domain envelope features, energy distribution features and duration as target sound wave features.

[0024] Furthermore, the method of inputting the windowed sound wave data into a feature extraction model to obtain frequency features, time domain envelope features, energy distribution features, and duration as target sound wave features includes:

[0025] The windowed sound wave data is transformed by fast Fourier transform to obtain the original frequency sequence X(k):

[0026]

[0027] Where j is an imaginary unit, satisfying j 2 = -1, n is the sampling point number, N is the number of FFT points, k is the frequency number, is a complex exponential function used to convert the time domain signal to the frequency domain, and X(k) is the original frequency sequence.

[0028] The amplitude spectrum Y(k) calculated according to the original frequency sequence is:

[0029]

[0030] Where Re[X(k)] is the real part of the original frequency sequence, Im[X(k)] is the imaginary part of the original spectrum sequence, and Y(k) is the amplitude spectrum.

[0031] Select all amplitude values ​​within the frequency range [200Hz, 12000Hz] from the amplitude spectrum to obtain the frequency feature F r for.

[0032]

[0033] Where k is the frequency index, N is the number of FFT points, and f s is the sampling frequency, F r is the frequency characteristic.

[0034] The windowed sound wave data is interpolated and reconstructed to obtain a continuous time signal, and the continuous time signal is subjected to Hilbert transform to obtain the transformed signal H[x ′ (t)] is:

[0035]

[0036] where x ′ (τ) is a continuous time signal, τ is the integral variable. t is the time variable of the continuous time signal, H[x′ (t)] is the transformed signal.

[0037] The analytical signal z(t) is constructed based on the transformed signal:

[0038] z(t)=x′(t)+j·H[x′(t)].

[0039] where z(t) is the analytical signal.

[0040] Extract the amplitude of the analytical signal to obtain the time domain envelope feature E r (t) is:

[0041] E r (t)=|z(t)|.

[0042] Among them E r (E) is the time domain envelope feature.

[0043] According to the frequency characteristics, the frequency band is divided into sub-bands according to the preset number of frequency band divisions, and the frequency peak of each sub-band is extracted to obtain the energy distribution characteristics:

[0044]

[0045] Where n is the preset number of frequency band divisions, j is the frequency band index, Y(k) is the amplitude spectrum, and P r (j) is the peak frequency of the jth frequency band.

[0046] The start and end time points of each pet sound wave are identified through the sound endpoint detection algorithm, and the difference is calculated to obtain the duration feature.

[0047] The frequency feature, time domain envelope feature, energy distribution feature and duration feature are combined as the target sound wave feature.

[0048] Furthermore, the method for presetting the pet sound wave resonance feature template includes:

[0049] The preset pet sound wave frequency characteristics, time domain envelope characteristics, energy distribution characteristics and duration characteristics are obtained in the same way as loading the sound wave signal into the first analysis model to obtain the target sound wave characteristics.

[0050] The frequency features, time domain envelope features, energy distribution features and duration features are combined to form a high-dimensional feature vector of the sound wave. The high-dimensional feature vector of the sound wave is standardized and reduced in dimension through principal component analysis. The principal components whose cumulative contribution rate exceeds the preset ratio are selected as the features after dimensionality reduction.

[0051] According to the features after dimensionality reduction, clustering is performed through Gaussian mixture and the optimal cluster number is automatically determined by the Bayesian Information Criterion to obtain the acoustic wave feature clusters. According to each acoustic wave feature cluster, the acoustic wave resonance feature pattern is constructed by calculating the center point and covariance matrix.

[0052] The acoustic wave resonance characteristic pattern is associated with a specific pet type to obtain a preset pet acoustic wave resonance characteristic template.

[0053] Furthermore, when the target sound wave feature meets the preset pet sound wave resonance feature template, the method for filtering the pet sound wave for interference signal to obtain filtered OBU data is:

[0054] Interference signals are filtered using an adaptive suppression algorithm based on the original frame sequence, specifically:

[0055] y(n)=x(n)―βw(n)s(n).

[0056] Where y(n) is the filtered output signal, x(n) is the original frame sequence, w(n) is the filter coefficient, n is the sampling point number, s(n) is the signal in the preset pet sound wave resonance feature template, and β is the suppression coefficient. Different suppression coefficients β are set for the main frequency bands of different pet types: for cat sound waves, the suppression coefficient in the 400Hz-800Hz frequency band is set to 0.95-0.99. For dog sound waves, the suppression coefficient in the 500Hz-1500Hz frequency band is set to 0.93-0.98. For bird sound waves, the suppression coefficient in the 2000Hz-6000Hz frequency band is set to 0.90-0.97.

[0057] The filter coefficients are iteratively updated, and the update expression is:

[0058]

[0059] in, is the adaptive step size factor, μ0 is the initial step size, ranging from 0.01 to 0.1, δ is a small positive number to prevent division by zero, and its value is 10 ―6 , e(n) is the error signal.

[0060] Set the error signal e(n) to:

[0061] e(n)=x(n)―y(n).

[0062] The system continuously adjusts w(n) to stop the iteration when the output e(n) is less than the error threshold and uses y(n) as the final filtered output signal when the iteration stops.

[0063] Furthermore, the method further comprises:

[0064] A new sample set is formed by recording the pet acoustic resonance features detected during the operation of the vehicle-mounted OBU. When the number of samples in the new sample set reaches the preset sample number threshold, the feature library is updated. Specifically:

[0065] By calculating the Mahalanobis distance between the new sample feature and the existing preset feature and comparing it with the preset judgment threshold, when the preset judgment threshold is exceeded, the new sample feature is added to the preset feature library and the cluster center is updated.

[0066] Furthermore, the method of encrypting the filtered OBU data and securely transmitting the encrypted data to the backend server via wireless communication includes:

[0067] Perform CRC-32 check and pass signal-to-noise ratio (SNR) > 20dB and bit error rate (BER) < 1×10 ―6 Verify data quality.

[0068] The encryption key generated by the security element of the on-board OBU is used to encrypt the OBU data packet using the AES-256 algorithm.

[0069] By embedding the pet sound wave filtering status binary flag in the OBU data packet, the flag is encapsulated together with the encrypted data to construct a secure data packet format. According to the secure data packet format, the data packets are transmitted to the background server through the 802.11p protocol and the server confirmation response is received.

[0070] Based on the same inventive concept, the present invention provides a data security transmission system for an on-vehicle OBU, which includes a storage module, an acoustic wave acquisition module, an acoustic wave feature analysis module, a signal processing module, a data encryption module, and a wireless communication module connected in sequence.

[0071] The storage module is used to store preset pet sound wave resonance feature templates, including a pet sound wave feature database corresponding to the sound wave frequencies of dogs, cats, and birds, as well as sound wave resonance feature patterns generated by principal component features after dimensionality reduction through principal component analysis and Gaussian mixture model clustering.

[0072] The sound wave acquisition module is used to collect the vehicle's internal sound wave signals in real time through the vehicle-mounted OBU microphone array.

[0073] The acoustic wave feature analysis module is used to load the acoustic wave signal into a first analysis model to obtain target acoustic wave features, where the target acoustic wave features include frequency features, time domain envelope features, energy distribution features, and duration.

[0074] The signal processing module is used to compare and judge the target sound wave characteristics with the preset pet sound wave resonance characteristic template stored locally in the vehicle-mounted OBU through a second analysis model. When the target sound wave characteristics meet the preset pet sound wave resonance characteristic template, the pet sound wave is subjected to interference signal filtering to obtain filtered OBU data.

[0075] The data encryption module is used for encrypting the filtered OBU data.

[0076] The wireless communication module is used to securely transmit the encrypted data to the backend server via wireless communication.

[0077] Compared with the prior art, the present invention has the following beneficial effects: the present invention realizes the recognition of pet sound wave resonance characteristics of vehicle-borne sound wave signals within a specific frequency range through the first analysis model and the second analysis model, further filters the interference signal caused by pet sound wave resonance through the suppression algorithm and then combines it with the encryption algorithm for encryption processing, thereby solving the problem of data security transmission caused by false triggering of voiceprints as a whole. BRIEF DESCRIPTION OF THE DRAWINGS

[0078] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without paying any creative work.

[0079] Figure 1 A flow chart of a method for securely transmitting data from an on-board OBU according to embodiment 1 of the present invention is provided;

[0080] Figure 2 This is a schematic diagram of the module composition of a method and system for secure data transmission of an on-board OBU according to embodiment 2 of the present invention. DETAILED DESCRIPTION

[0081] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0082] Example 1: Figure 1 As shown, this embodiment provides a method for secure data transmission of an on-board OBU, the method comprising:

[0083] S1. Collect the vehicle's internal sound wave signal in real time through the vehicle-mounted OBU microphone array, and load the sound wave signal into the first analysis model to obtain target sound wave characteristics, wherein the target sound wave characteristics include frequency characteristics, time domain envelope characteristics, energy distribution characteristics and duration.

[0084] S2. Compare and judge the target sound wave feature with the preset pet sound wave resonance feature template stored locally in the vehicle-mounted OBU through a second analysis model. When the target sound wave feature meets the preset pet sound wave resonance feature template, filter the pet sound wave for interference signals to obtain filtered OBU data.

[0085] S3. Encrypt the filtered OBU data and securely transmit the encrypted data to the backend server via wireless communication.

[0086] For example, taking the sound waves of a dog on a vehicle as an example, the vehicle-mounted OBU microphone array continuously monitors the sound environment inside the vehicle. After the collected sound wave signals inside the vehicle are input into the first analysis model, the target sound wave features are extracted, including frequency features, time domain envelope features, energy distribution features and duration. These features are compared with the locally stored dog sound wave resonance feature template, and the calculated similarity is 0.91, the time domain envelope correlation coefficient is 0.89, and the energy distribution Euclidean distance is 11.2. When the judgment thresholds are set to similarity greater than 0.85, correlation coefficient greater than 0.8, and Euclidean distance less than 15.0, and the duration is between 2 and 5 seconds, the judgment thresholds are met, confirming that the dog sound wave is detected. The OBU device starts the interference signal filtering program and sets the suppression coefficient to 0.96. After 45 iterations, the error is reduced to 0.0018, and the dog's voice is successfully filtered. The filtered 2.6KB data packet is encrypted with AES-256 and securely transmitted to the background server.

[0087] It should be noted that the method of comparing the target sound wave feature with the preset pet sound wave resonance feature template stored locally in the vehicle-mounted OBU through the second analysis model includes:

[0088] The similarity S is calculated based on the frequency characteristics of the sound wave signal and the preset pet sound wave frequency characteristics:

[0089]

[0090] Among them F r (i) is the eigenvalue corresponding to the i-th frequency component in the frequency characteristics of the real-time acquired sound wave signal, F p (i) is the eigenvalue corresponding to the i-th frequency component in the preset frequency characteristics of the pet sound wave, n is the total number of frequency components, and S is the similarity.

[0091] The correlation coefficient C is calculated based on the time domain envelope characteristics of the sound wave signal and the preset time domain envelope characteristics of the pet sound wave:

[0092]

[0093] Among them E r (t) Real-time acquisition of the time domain envelope value of the sound wave at the tth time point, is the mean value of the real-time acquisition of the acoustic wave time domain envelope, E p (t) is the time domain envelope eigenvalue of the preset pet sound wave at the tth time point, is the mean of the preset pet sound wave time domain envelope feature, N is the total number of time points, and C is the correlation coefficient.

[0094] The Euclidean distance is calculated based on the energy distribution characteristics of the sound wave and the preset energy distribution characteristics of the pet sound wave:

[0095]

[0096] Among them, P r (j) is the energy value of the real-time collected sound wave in the jth frequency band, P P (j) is the energy value of the preset pet sound wave in the jth frequency band, and D is the Euclidean distance.

[0097] When the similarity, correlation coefficient, and Euclidean distance all meet the preset judgment threshold and the duration of the real-time collected sound wave signal is within the preset pet sound wave duration range, it is determined that the current sound wave meets the preset pet sound wave resonance characteristics.

[0098] For example, eight key frequency components were selected for comparison. The input data is: the frequency eigenvalues ​​of the dog's barking sound collected in real time are [0.82, 0.75, 0.91, 0.68, 0.54, 0.43, 0.37, 0.29], and the corresponding frequency eigenvalues ​​of the dog template are [0.85, 0.72, 0.88, 0.71, 0.52, 0.41, 0.35, 0.31]. After comparing each of these components using the similarity calculation formula, the similarity contributions of the eight frequency components are [0.96, 0.96, 0.97, 0.96, 0.96, 0.95, 0.94, 0.94], resulting in an average similarity S of 0.96.

[0099] The envelope data from 280 time points were extracted, and the mean of the time domain envelope sequence was 0.42, while the mean of the canine template envelope sequence was 0.38. After covariance calculation and standard deviation normalization, the correlation coefficient C was 0.89.

[0100] Taking 24 sub-bands as an example, the energy values ​​of the dog barking are [0.12, 0.18, 0.35, 0.67, 0.84, 0.92, 0.78, 0.61, 0.45, 0.32, 0.24, 0.18, 0.13, 0.09, 0.07, 0.05, 0.04, 0.03, 0.02, 0.02, 0.01, 0.01, 0.01, 0.01]. The corresponding energy values ​​of the plates are [0.15, 0.21, 0.38, 0.71, 0.88, 0.95, 0.82, 0.64, 0.48, 0.35, 0.26, 0.19, 0.14, 0.10, 0.07, 0.05, 0.04, 0.03, 0.02, 0.02, 0.01, 0.01, 0.01, 0.01], and the Euclidean distance D is calculated to be 11.2.

[0101] Since the dog's barking lasted for 2.8 seconds, which is within the set duration range of 2.0-5.0 seconds, it was finally determined that the current sound wave met the preset dog sound wave resonance characteristics.

[0102] It should be noted that the method of loading the acoustic wave signal into the first analysis model to obtain the target acoustic wave feature includes:

[0103] The acoustic wave signal is segmented to obtain the original frame sequence, and the original frame sequence is windowed by the window function to obtain the windowed acoustic wave data:

[0104] x ′ (n) = x(n)·w(n).

[0105] Where x(n) is the original frame sequence, w(n) is the window function, x ′ (n) is the windowed sound wave data.

[0106] The windowed sound wave data is input into the feature extraction model to obtain frequency features, time domain envelope features, energy distribution features and duration as target sound wave features.

[0107] For example, framing and windowing a raw audio signal with a sampling frequency of 44100 Hz. A 2.8-second continuous sound wave signal is divided into 25-millisecond frames, resulting in 112 frames of data, each containing 1102 sampling points. When windowing the 50th frame, the amplitude of the input raw frame sequence ranges from -0.8 to 0.7. A Hanning window with a window length of N = 1102 is used as the window function. After windowing, the amplitude of the sound wave data is smoothly adjusted to the range of -0.4 to 0.35.

[0108] It should be noted that the method of inputting the windowed sound wave data into the feature extraction model to obtain frequency features, time domain envelope features, energy distribution features and duration as target sound wave features includes:

[0109] The windowed sound wave data is transformed by fast Fourier transform to obtain the original frequency sequence X(k):

[0110]

[0111] Where j is an imaginary unit, satisfying j 2 = -1, n is the sampling point number, N is the number of FFT points, k is the frequency number, is a complex exponential function used to convert the time domain signal to the frequency domain, and X(k) is the original frequency sequence.

[0112] The amplitude spectrum Y(k) calculated according to the original frequency sequence is:

[0113]

[0114] Where Re[X(k)] is the real part of the original frequency sequence, Im[X(k)] is the imaginary part of the original spectrum sequence, and Y(k) is the amplitude spectrum.

[0115] Select all amplitude values ​​within the frequency range [200Hz, 12000Hz] from the amplitude spectrum to obtain the frequency feature F r for.

[0116]

[0117] Where k is the frequency index, N is the number of FFT points, and f s is the sampling frequency, F r is the frequency characteristic.

[0118] For example, in the Fourier transform process, the time domain data of 1102 sampling points is input, and the complex spectrum is obtained after 2048-point FFT expansion. If the frequency number k=39 corresponding to the key frequency 850Hz, the real part Re[X(39)] of the FFT output is 0.73, and the imaginary part Im[X(39)] is -0.42, then the frequency characteristic of this frequency point can be further calculated to be 0.84.

[0119] The windowed sound wave data is interpolated and reconstructed to obtain a continuous time signal, and the continuous time signal is subjected to Hilbert transform to obtain the transformed signal H[x ′ (t)] is:

[0120]

[0121] where x ′(τ) is a continuous time signal, τ is the integral variable. t is the time variable of the continuous time signal, H[x ′ (t)] is the transformed signal.

[0122] The analytical signal z(t) is constructed based on the transformed signal:

[0123] z(t)=x′(t)+j·H[x′(t)].

[0124] where z(t) is the analytical signal.

[0125] Extract the amplitude of the analytical signal to obtain the time domain envelope feature E r (t) is:

[0126] E r (t)=|z(t)|.

[0127] Among them E r (t) is the time domain envelope feature.

[0128] For example, after Hilbert transforming the windowed sound wave data, at time point t = 1.2 seconds, the original signal value x ′ (1.2) = 0.65, Hilbert transform result H[x ′ (1.2)] = -0.38. The analytical signal z(1.2) = 0.65 + j (-0.38]) is constructed, and the time domain envelope characteristic at this time point is calculated to be 0.75.

[0129] According to the frequency characteristics, the frequency band is divided into sub-bands according to the preset number of frequency band divisions, and the frequency peak of each sub-band is extracted to obtain the energy distribution characteristics:

[0130]

[0131] Where m is the preset number of frequency band divisions, j is the frequency band index, Y(k) is the amplitude spectrum, and P r (j) is the peak frequency of the jth frequency band.

[0132] For example, in the energy distribution calculation, the 200Hz-12000Hz frequency band is evenly divided into 24 sub-bands, each with a width of 491.7Hz. The fifth frequency band is the 1966.8Hz-2458.5Hz band, and the maximum amplitude value in this band is 0.92. Similarly, the energy distribution characteristic vectors of the 24 frequency bands can be obtained.

[0133] The start and end time points of each pet sound wave are identified through the sound endpoint detection algorithm, and the difference is calculated to obtain the duration feature.

[0134] For example, if the short-term energy threshold is set to 0.05 and the zero-crossing rate threshold is set to 0.3, the start time of Dahuang's barking is detected at the 529th sampling point at 0.12 seconds and the end time is at the 1282nd sampling point at 2.92 seconds. The calculated duration is 2.8 seconds.

[0135] The frequency feature, time domain envelope feature, energy distribution feature and duration feature are combined as the target sound wave feature.

[0136] It should be noted that the method for presetting the pet sound wave resonance feature template includes:

[0137] The preset pet sound wave frequency characteristics, time domain envelope characteristics, energy distribution characteristics and duration characteristics are obtained in the same way as loading the sound wave signal into the first analysis model to obtain the target sound wave characteristics.

[0138] The frequency features, time domain envelope features, energy distribution features and duration features are combined to form a high-dimensional feature vector of the sound wave. The high-dimensional feature vector of the sound wave is standardized and reduced in dimension through principal component analysis. The principal components whose cumulative contribution rate exceeds the preset ratio are selected as the features after dimensionality reduction.

[0139] According to the features after dimensionality reduction, clustering is performed through Gaussian mixture and the optimal cluster number is automatically determined by the Bayesian Information Criterion to obtain the acoustic wave feature clusters. According to each acoustic wave feature cluster, the acoustic wave resonance feature pattern is constructed by calculating the center point and covariance matrix.

[0140] The acoustic wave resonance characteristic pattern is associated with a specific pet type to obtain a preset pet acoustic wave resonance characteristic template.

[0141] For example, 50 pet vocalization samples were collected. Each sample was subjected to the same feature extraction process, resulting in a 128-dimensional feature vector: 64-dimensional frequency features, 32-dimensional time envelope features, 24-dimensional energy distribution features, and 8-dimensional duration features. After inputting the principal component analysis algorithm, the cumulative contribution of the first 32 principal components reached 95.3%, and these 32 principal components were selected as the feature representation after dimensionality reduction. The 32-dimensional feature vectors of the 50 samples were clustered using a Gaussian mixture model, and the Bayesian Information Criterion determined the optimal number of clusters to be 3. The first cluster, consisting of 18 dog samples, had the first three components of the central feature vector in the range [0.72, -0.31, 0.54]. The second cluster, consisting of 21 cat samples, had the first three components of the central feature vector in the range [0.68, 0.43, -0.29]. The third cluster, consisting of 11 bird samples, had the first three components of the central feature vector in the range [0.51, 0.17, 0.39].

[0142] It should be noted that, when the target sound wave feature meets the preset pet sound wave resonance feature template, the method for filtering the pet sound wave for interference signal to obtain filtered OBU data is:

[0143] Interference signals are filtered using an adaptive suppression algorithm based on the original frame sequence, specifically:

[0144] y(n)=x(n)―βw(n)s(n).

[0145] Where y(n) is the filtered output signal, x(n) is the original frame sequence, w(n) is the filter coefficient, n is the sampling point number, s(n) is the signal in the preset pet sound wave resonance feature template, and β is the suppression coefficient. Different suppression coefficients β are set for the main frequency bands of different pet types: for cat sound waves, the suppression coefficient in the 400Hz-800Hz frequency band is set to 0.95-0.99. For dog sound waves, the suppression coefficient in the 500Hz-1500Hz frequency band is set to 0.93-0.98. For bird sound waves, the suppression coefficient in the 2000Hz-6000Hz frequency band is set to 0.90-0.97.

[0146] The filter coefficients are iteratively updated, and the update expression is:

[0147]

[0148] in, is the adaptive step size factor, μ0 is the initial step size, ranging from 0.01 to 0.1, δ is a small positive number to prevent division by zero, and its value is 10 ―6 , e(n) is the error signal.

[0149] Set the error signal e(n) to:

[0150] e(n)=x(n)―y(n).

[0151] The system continuously adjusts w(n) to stop the iteration when the output e(n) is less than the error threshold and uses y(n) as the final filtered output signal when the iteration stops.

[0152] For example, after confirming that the sound wave belongs to a dog, the detected dog sound wave is subjected to adaptive suppression filtering, with the suppression coefficient β set to 0.96, the initial filter coefficient w(0) set to 0.1, the initial step size set to 0.05, and the anti-zero parameter set to 1×10 ―6At the 30th iteration, the original signal x(30) = 0.73, the preset canine template signal s(30) = 0.68, the current filter coefficient w(30) = 0.142, and the calculated filter output y(30) = 0.73-0.96×0.142×0.68 = 0.638. The error signal e(30) = 0.73-0.638 = 0.092, and the adaptive step size μ(30) = 0.05 / (10^-6+0.68 2 )=0.108. Update filter coefficient w(31)=0.142+0.108×0.092×0.68 / 0.68 2 =0.156. After 45 iterations, the error signal drops to 0.0018, which is less than the set error threshold of 0.002. The iteration is stopped and the final filtered signal is output as the filtered OBU data.

[0153] It should be noted that the method further includes:

[0154] A new sample set is formed by recording the pet acoustic resonance features detected during the operation of the vehicle-mounted OBU. When the number of samples in the new sample set reaches the preset sample number threshold, the feature library is updated. Specifically:

[0155] By calculating the Mahalanobis distance between the new sample feature and the existing preset feature and comparing it with the preset judgment threshold, when the preset judgment threshold is exceeded, the new sample feature is added to the preset feature library and the cluster center is updated.

[0156] For example, during the operation of the on-board OBU system, the detected pet sound wave characteristics are continuously monitored and recorded. When the preset sample quantity threshold is set to 15, the feature library update process is triggered when 15 new dog sound wave samples are accumulated. The Mahalanobis distance between the newly collected 12th sample feature vector and the existing dog cluster center feature vector is calculated, and the distance value is 2.47. If the preset judgment threshold set by the system is 2.30, since 2.47>2.30, the new sample is determined to represent a new characteristic pattern of dog sound waves and is added to the dog feature library, thereby improving the system's recognition accuracy of dog sound wave variations.

[0157] It should be noted that the method of encrypting the filtered OBU data and securely transmitting the encrypted data to the backend server via wireless communication includes:

[0158] Perform CRC-32 check and pass signal-to-noise ratio (SNR) > 20dB and bit error rate (BER) < 1×10 ―6 Verify data quality.

[0159] The encryption key generated by the security element of the on-board OBU is used to encrypt the OBU data packet using the AES-256 algorithm.

[0160] By embedding the pet sound wave filtering status binary flag in the OBU data packet, the flag is encapsulated together with the encrypted data to construct a secure data packet format. According to the secure data packet format, the data packets are transmitted to the background server through the 802.11p protocol and the server confirmation response is received.

[0161] For example, after filtering, the data is first subjected to CRC-32 check to generate a 32-bit check code of 0x8F4A2C1D to ensure data integrity. At the same time, the signal-to-noise ratio of the current communication environment is detected to be 23.4dB, which meets the requirement of SNR>20dB. The measured bit error rate is 3.2×10 ―7 , meeting BER < 1 × 10 ―6 The data quality meets the requirements. The HSM, a secure element built into the vehicle's OBU, then generates a 256-bit AES encryption key. For example, the key is "A1B2C3D4E5F67890FEDCBA0987654321A1B2C3D4E5F67890FEDCBA0987654321." The 2.6KB data packet is encrypted with AES-256. An 8-bit pet sound wave filtering status flag, "10110001," is embedded in the packet header. Bit 1 indicates whether pet sound waves are detected (1 indicates detection). Bits 2-4 indicate the pet type (011 indicates dog). Bits 5-8 indicate the filtering intensity level (0001 indicates mild filtering). The flag bit is encapsulated with the encrypted data to construct a secure data packet, which is transmitted to the background server via the 802.11p protocol in the 5.9GHz frequency band with a 10MHz channel bandwidth. The transmission takes 45ms, and the ACK confirmation response packet returned by the server is successfully received, thus completing the secure transmission process.

[0162] Example 2: Based on the same inventive concept, Figure 2 As shown, this embodiment provides a data security transmission system for an on-vehicle OBU, which includes a storage module, an acoustic wave acquisition module, an acoustic wave feature analysis module, a signal processing module, a data encryption module, and a wireless communication module connected in sequence.

[0163] The storage module is used to store preset pet sound wave resonance feature templates, including a pet sound wave feature database corresponding to the sound wave frequencies of dogs, cats, and birds, as well as sound wave resonance feature patterns generated by principal component features after dimensionality reduction through principal component analysis and Gaussian mixture model clustering.

[0164] The sound wave acquisition module is used to collect the vehicle's internal sound wave signals in real time through the vehicle-mounted OBU microphone array.

[0165] The acoustic wave feature analysis module is used to load the acoustic wave signal into a first analysis model to obtain target acoustic wave features, where the target acoustic wave features include frequency features, time domain envelope features, energy distribution features, and duration.

[0166] The signal processing module is used to compare and judge the target sound wave characteristics with the preset pet sound wave resonance characteristic template stored locally in the vehicle-mounted OBU through a second analysis model. When the target sound wave characteristics meet the preset pet sound wave resonance characteristic template, the pet sound wave is subjected to interference signal filtering to obtain filtered OBU data.

[0167] The data encryption module is used for encrypting the filtered OBU data.

[0168] The wireless communication module is used to securely transmit the encrypted data to the backend server via wireless communication.

[0169] It should be noted that, regarding the system in the above embodiment, the specific manner in which each module performs operations has been described in detail in Example 1 of the method, and will not be elaborated on here.

[0170] Finally, it should be noted that although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments, or make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A data security transmission method for an on-board OBU, characterized in that: The method comprises the following steps: Acquiring a vehicle interior acoustic wave signal in real time through an onboard OBU microphone array, and loading the acoustic wave signal into a first analysis model to obtain target acoustic wave characteristics, wherein the target acoustic wave characteristics include frequency characteristics, time domain envelope characteristics, energy distribution characteristics, and duration; The target sound wave feature is compared with the preset pet sound wave resonance feature template stored in the vehicle-mounted OBU through a second analysis model, and when the target sound wave feature meets the preset pet sound wave resonance feature template, the pet sound wave is subjected to interference signal filtering to obtain filtered OBU data; The filtered OBU data is encrypted and securely transmitted to the backend server via wireless communication.

2. The data security transmission method of a vehicle-mounted OBU according to claim 1, wherein The method of comparing and judging the target sound wave feature with a preset pet sound wave resonance feature template stored locally in the vehicle-mounted OBU through a second analysis model includes: The similarity S is calculated based on the frequency characteristics of the sound wave signal and the preset pet sound wave frequency characteristics: Among them F r (i) is the eigenvalue corresponding to the i-th frequency component in the frequency characteristics of the real-time acquired sound wave signal, F p (i) is the eigenvalue corresponding to the i-th frequency component in the preset frequency characteristics of the pet sound wave, n is the total number of frequency components, and S is the similarity; The correlation coefficient C is calculated based on the time domain envelope characteristics of the sound wave signal and the preset time domain envelope characteristics of the pet sound wave: Among them E r (t) Real-time acquisition of the time domain envelope value of the sound wave at the tth time point, is the mean value of the real-time acquisition of the acoustic wave time domain envelope, E p (t) is the time domain envelope eigenvalue of the preset pet sound wave at the tth time point, is the mean of the preset pet sound wave time domain envelope feature, N is the total number of time points, and C is the correlation coefficient; The Euclidean distance is calculated based on the energy distribution characteristics of the sound wave and the preset energy distribution characteristics of the pet sound wave: Among them, P r (j) is the energy value of the real-time collected sound wave in the jth frequency band, P P (j) is the energy value of the preset pet sound wave in the jth frequency band, and D is the Euclidean distance; When the similarity, correlation coefficient, and Euclidean distance all meet the preset judgment threshold and the duration of the real-time collected sound wave signal is within the preset pet sound wave duration range, it is determined that the current sound wave meets the preset pet sound wave resonance characteristics.

3. The data security transmission method of a vehicle-mounted OBU according to claim 1, wherein The method of loading the acoustic wave signal into the first analysis model to obtain the target acoustic wave feature includes: The acoustic wave signal is segmented to obtain the original frame sequence, and the original frame sequence is windowed by the window function to obtain the windowed acoustic wave data: x'(n)=x(n)·w(n); Where x(n) is the original frame sequence, w(n) is the window function, x'(n) is the windowed sound wave data; The windowed sound wave data is input into the feature extraction model to obtain frequency features, time domain envelope features, energy distribution features and duration as target sound wave features.

4. The data security transmission method of a vehicle-mounted OBU according to claim 3, wherein The method of inputting the windowed sound wave data into a feature extraction model to obtain frequency features, time domain envelope features, energy distribution features, and duration as target sound wave features includes: The windowed sound wave data is transformed by fast Fourier transform to obtain the original frequency sequence X(k): Where j is an imaginary unit, satisfying j 2 =-1, n is the sampling point number, N is the number of FFT points, k is the frequency number, is a complex exponential function used to convert the time domain signal to the frequency domain, X(k) is the original frequency sequence; The amplitude spectrum Y(k) calculated according to the original frequency sequence is: Where Re[X(k)] is the real part of the original frequency sequence, Im[X(k)] is the imaginary part of the original spectrum sequence, and Y(k) is the amplitude spectrum; Select all amplitude values ​​within the frequency range [200Hz, 12000Hz] from the amplitude spectrum to obtain the frequency feature F r for; Where k is the frequency index, N is the number of FFT points, and f s is the sampling frequency, F r is the frequency characteristic; The windowed sound wave data is interpolated and reconstructed to obtain a continuous-time signal. The Hilbert transform is performed on the continuous-time signal to obtain the transformed signal H[x'(t)]: Where X'(τ) is the continuous-time signal, τ is the integral variable; t is the time variable of the continuous-time signal, and H[x'(t)] is the transformed signal; The analytical signal z(t) is constructed based on the transformed signal: z(t)=x'(t)+j·H[x'(t)]; Where z(t) is the analytical signal; Extract the amplitude of the analytical signal to obtain the time domain envelope feature E r (t) is: E r (t)=|z(t)|; Among them E r (t) is the time domain envelope feature; According to the frequency characteristics, the frequency band is divided into sub-bands according to the preset number of frequency band divisions, and the frequency peak of each sub-band is extracted to obtain the energy distribution characteristics: Where m is the preset number of frequency band divisions, j is the frequency band index, Y(k) is the amplitude spectrum, and P r (j) is the peak frequency of the jth frequency band; The start and end time points of each pet sound wave are identified by the sound endpoint detection algorithm and the difference is calculated to obtain the duration feature; The frequency feature, time domain envelope feature, energy distribution feature and duration feature are combined as the target sound wave feature.

5. The data security transmission method of a vehicle-mounted OBU according to claim 3, wherein The method for presetting the pet sound wave resonance feature template includes: Obtaining preset pet sound wave frequency characteristics, time domain envelope characteristics, energy distribution characteristics, and duration characteristics by the same method as loading the sound wave signal into the first analysis model to obtain the target sound wave characteristics; The frequency feature, time domain envelope feature, energy distribution feature and duration feature are combined to form a high-dimensional feature vector of the sound wave. The high-dimensional feature vector of the sound wave is standardized and reduced in dimension through principal component analysis. The principal component with a cumulative contribution rate exceeding a preset ratio is selected as the feature after dimensionality reduction. According to the features after dimensionality reduction, clustering is performed through Gaussian mixture and the optimal cluster number is automatically determined by Bayesian information criterion to obtain acoustic wave feature clusters. According to each acoustic wave feature cluster, the acoustic wave resonance feature pattern is constructed by calculating the center point and covariance matrix; The acoustic wave resonance characteristic pattern is associated with a specific pet type to obtain a preset pet acoustic wave resonance characteristic template.

6. A method for securely transmitting data of an on-board OBU according to claim 4, characterized in that: The method for filtering the pet sound wave for interference signals to obtain filtered OBU data when the target sound wave feature meets the preset pet sound wave resonance feature template is as follows: Interference signals are filtered using an adaptive suppression algorithm based on the original frame sequence, specifically: y(n)=x(n)-βw(n)s(n); Where y(n) is the filtered output signal, x(n) is the original frame sequence, w(n) is the filter coefficient, n is the sampling point number, s(n) is the signal in the preset pet sound wave resonance feature template, β is the suppression coefficient, and different suppression coefficients β are set for the main frequency bands of different pet types: for cat sound waves, the suppression coefficient in the 400Hz-800Hz frequency band is set to 0.95-0.99; for dog sound waves, the suppression coefficient in the 500Hz-1500Hz frequency band is set to 0.93-0.98; for bird sound waves, the suppression coefficient in the 2000Hz-6000Hz frequency band is set to 0.90-0.97; The filter coefficients are iteratively updated, and the update expression is: in, is the adaptive step size factor, μ0 is the initial step size, ranging from 0.01 to 0.1, δ is a small positive number to prevent division by zero, and its value is 10 -6 , e(n) is the error signal; Set the error signal e(n) to: e(n)=x(n)-y(n); The system continuously adjusts w(n) to stop the iteration when the output e(n) is less than the error threshold and uses y(n) as the final filtered output signal when the iteration stops.

7. A method for securely transmitting data of an on-board OBU according to claim 1, characterized in that: The method further comprises: A new sample set is formed by recording the pet acoustic resonance features detected during the operation of the vehicle-mounted OBU. When the number of samples in the new sample set reaches the preset sample number threshold, the feature library is updated. Specifically: By calculating the Mahalanobis distance between the new sample feature and the existing preset feature and comparing it with the preset judgment threshold, when the preset judgment threshold is exceeded, the new sample feature is added to the preset feature library and the cluster center is updated.

8. A method for securely transmitting data of an on-board OBU according to claim 1, characterized in that: The method of encrypting the filtered OBU data and securely transmitting the encrypted data to the backend server via wireless communication comprises: Perform CRC-32 check and pass signal-to-noise ratio (SNR) > 20dB and bit error rate (BER) < 1×10 -6 Verify data quality; The OBU data packet is encrypted using the AES-256 algorithm using the encryption key generated by the OBU's secure element. By embedding the pet sound wave filtering status binary flag in the OBU data packet, the flag is encapsulated together with the encrypted data to construct a secure data packet format. According to the secure data packet format, the data packets are transmitted to the background server through the 802.11p protocol and the server confirmation response is received.

9. A data security transmission system for an on-board OBU, for executing a data security transmission method for an on-board OBU according to any one of claims 1 to 8, characterized in that: The system includes a storage module, an acoustic wave acquisition module, an acoustic wave feature analysis module, a signal processing module, a data encryption module, and a wireless communication module connected in sequence; The storage module is used to store preset pet sound wave resonance feature templates, including a pet sound wave feature database corresponding to the sound wave frequencies of dogs, cats, and birds, as well as sound wave resonance feature patterns generated by principal component features after dimensionality reduction through principal component analysis and Gaussian mixture model clustering; The sound wave acquisition module is used to collect the vehicle's internal sound wave signals in real time through the vehicle-mounted OBU microphone array; The acoustic wave feature analysis module is used to load the acoustic wave signal into a first analysis model to obtain target acoustic wave features, where the target acoustic wave features include frequency features, time domain envelope features, energy distribution features, and duration; The signal processing module is used to compare and judge the target sound wave feature with the preset pet sound wave resonance feature template stored in the vehicle-mounted OBU through a second analysis model, and when the target sound wave feature meets the preset pet sound wave resonance feature template, the pet sound wave is subjected to interference signal filtering to obtain filtered OBU data; Described data encryption module is used for OBU data after filtering being carried out encryption process; The wireless communication module is used to securely transmit the encrypted data to the backend server via wireless communication.

Images