A method for expressing amplitude-phase time-frequency diagrams based on LAB color space mapping

Abstract

The present invention discloses an amplitude-phase time-frequency diagram expression method based on LAB color space mapping, comprising the following steps: performing a short-time Fourier transform on a time domain signal and obtaining an amplitude value and a phase value, thereby obtaining an initial amplitude matrix and a phase matrix; performing piecewise linear compression and expansion processing on the initial amplitude matrix to obtain a first amplitude matrix; performing a differential sine-cosine transform on the initial phase matrix to obtain a first phase matrix and a second phase matrix, and obtaining a set of three-channel data consisting of the first amplitude matrix, the first phase matrix, and the second phase matrix; mapping the obtained three-channel data to the LAB domain to obtain a three-channel time-frequency domain matrix L, A, and B under the LAB color space; converting the three-channel time-frequency domain matrix L, A, and B to the RGB domain to generate a visualized time-frequency diagram. The time-frequency diagram generated by the present invention can suppress the influence of noise on the time-frequency diagram while having amplitude spectrum and phase spectrum information, strengthen the color expression of the useful signal part in the time-frequency diagram, and improve the visualization effect of the time-frequency diagram.

Description

Technical Field

[0001] The present invention relates to the technical field of radio signal processing, and more particularly to an amplitude-phase time-frequency diagram expression method based on LAB color space mapping. Background Art

[0002] A time-frequency diagram is a two-dimensional image obtained by visualizing a two-dimensional matrix in the time-frequency domain obtained by performing a time-frequency transform (such as a short-time Fourier transform or wavelet transform) on a time-domain signal. A time-frequency diagram describes the temporal relationship between the amplitude and phase of a signal's spectral components. Its color and texture variations provide a highly effective representation of the signal. In radio signal processing (communications, radar, navigation, etc.), detection and estimation theory combined with artificial intelligence techniques are used to model and analyze radio signal time-frequency diagrams. This can be used to solve a wide range of signal processing problems, including detection, identification, noise reduction, separation, and parameter estimation. The most typical time-frequency diagram is the amplitude (or energy) spectrum, which carries only amplitude (or energy) information. Its planar two-dimensional image is a single-layer (single-channel) grayscale image. Another common time-frequency diagram is a single-channel grayscale image of the phase spectrum, which carries only phase information. Furthermore, there are also methods that combine the amplitude spectrum, energy spectrum, and phase spectrum together or in part to produce (dual-channel) pseudo-color images or (triple-channel) color images in RGB representation.

[0003] The current mainstream method for generating time-frequency maps is to calculate the amplitude, energy, and phase of the complex matrix obtained after time-frequency transformation of the time-domain signal. This results in an amplitude matrix, an energy matrix, and a phase matrix. By directly mapping the element values ​​in these three matrices to grayscale values, a time-frequency grayscale map of a single amplitude spectrum, energy spectrum, or phase spectrum is obtained. Alternatively, these three matrices are independently quantized and mapped to the RGB channels to obtain a time-frequency color map in RGB mode.

[0004] Existing time-frequency grayscale images that generate single amplitude (energy) or phase spectra only consider the amplitude (energy) or phase information of the signal's time-frequency transform. While some methods map amplitude (energy) and phase information to the three RGB channels to generate time-frequency color images, the quantization and mapping of amplitude (energy) and phase information are typically separate and independent, rigidly assigning amplitude and phase to the three RGB channels. The equivalence between the three RGB channels results in a lack of physical interpretability in the presentation of the time-frequency images, and also fails to consider noise suppression and optimized visualization of the time-frequency images. Summary of the Invention

[0005] In order to solve the problems of insufficient and unreasonable utilization of the amplitude and phase information of the signal time-frequency transformation in the above-mentioned prior art during the generation of time-frequency diagrams, the present invention provides an amplitude-phase time-frequency diagram expression method based on LAB color space mapping, which enables the generated time-frequency diagram to have amplitude spectrum and phase spectrum information while suppressing the influence of noise on the time-frequency diagram, strengthening the color expression of the useful signal part in the time-frequency diagram, and improving the visualization effect of the time-frequency diagram.

[0006] In order to achieve the above-mentioned purpose of the present invention, the technical solutions adopted are as follows:

[0007] A method for expressing an amplitude-phase time-frequency diagram based on LAB color space mapping, the method comprising the following steps:

[0008] S1: Perform short-time Fourier transform on the time domain signal and obtain the amplitude and phase values ​​to obtain the initial amplitude matrix and phase matrix;

[0009] S2: Perform piecewise linear compression and expansion on the initial amplitude matrix to obtain a first amplitude matrix; perform differential sine-cosine transform on the initial phase matrix to obtain a first phase matrix and a second phase matrix, thereby obtaining a set of three-channel data consisting of the first amplitude matrix, the first phase matrix, and the second phase matrix;

[0010] S3: Map the obtained three-channel data to the LAB domain to obtain the three-channel time-frequency domain matrices L, A, and B in the LAB color space;

[0011] S4: Convert the three-channel time-frequency domain matrices L, A, and B to the RGB domain to generate a visual time-frequency diagram.

[0012] Preferably, S1, specifically,

[0013] The time domain signal of the radio is expressed as Where L is the sample length of the signal; the short-time Fourier transform of the time domain signal is expressed as a complex time-frequency matrix Where T and F correspond to the number of time frames and frequency points;

[0014] Then the initial amplitude matrix and phase matrix corresponding to the time-frequency matrix of the time domain signal are expressed as:

[0015]

[0016] P=∠S∈[-π,π] T×F

[0017] Where ∠ is the radian angle operation, M represents the initial amplitude matrix, and P represents the initial phase matrix.

[0018] Furthermore, the initial amplitude matrix is ​​subjected to piecewise linear compression and expansion processing to obtain a first amplitude matrix, which is as follows:

[0019] Perform statistics on the matrix elements of the initial amplitude matrix M and select the first threshold M max Make M max The matrix element values ​​of the initial amplitude matrix M are not less than 98%;

[0020] The matrix elements of the initial amplitude matrix M are normalized by limiting the amplitude.

[0021]

[0022] Among them, m t,f is the matrix element of the initial amplitude matrix M, t∈{1,…,T},f∈{1,…,F}; m′ t,f is the normalized amplitude matrix M normal ∈[0,1] T×F Matrix elements of ;

[0023] Determine the second threshold M for dividing the noise component and the useful signal component in the time-frequency transform amplitude value according to the result of the noise estimation th ;

[0024] The signal below the second threshold is determined as noise and linearly compressed; the signal above the second threshold is determined as useful signal and linearly expanded; thus, a first amplitude matrix after piecewise linear compression and expansion is obtained.

[0025] Furthermore, the functional formula of the piecewise linear compression and expansion process is as follows:

[0026]

[0027] Where k∈(0,1) is the slope of the linear compression part, m″ t,f Represents the first amplitude matrix M compand The matrix element in row t and column f in M compand ∈[0,1] T×F Represents the normalized amplitude matrix M normal The first amplitude matrix after piecewise linear companding.

[0028] Furthermore, the initial phase matrix is ​​subjected to differential sine-cosine transform, as follows:

[0029] The first-order forward differential transformation of the initial phase matrix P in the frequency dimension is used to increase the discrimination between the target signal and the noise in the phase spectrum, and the differential phase matrix ΔP∈[-2π,2π] is obtained. T×F ,Right now

[0030] Δpt,f =p t,f+1 -p t,F

[0031] Where Δp t,f is the matrix element of the differential phase matrix ΔP, p t,f is the matrix element of the initial phase matrix P, t∈{1,…,T},f∈{1,…,F};

[0032] The differential phase matrix is ​​trigonometrically transformed to eliminate the radian angle jumps of -2π and +2π. The phase matrix after trigonometric transformation is expressed as:

[0033] P sin =sin(ΔP)∈[-1,1] T×F

[0034] P cos =cos(ΔP)∈[-1,1] T×F

[0035] Among them, cos(·) represents trigonometric cosine transform, sin(·) represents trigonometric sine transform, P sin represents the first phase matrix obtained after triangular sine transform, P cos represents the second phase matrix obtained by trigonometric cosine transform.

[0036] Furthermore, the obtained three-channel data is mapped to the LAB domain as follows:

[0037] Combine the three channels L, A, and B, and the first amplitude matrix M compand and the first phase matrix P sin , the second phase matrix P cos The physical meaning of P sin 2 +P cos 2 =1 constraint relationship;

[0038] The first amplitude matrix M compand Mapped to the L channel to control the brightness of the time-frequency graph;

[0039] The first phase matrix P sin , the second phase matrix P cos They are mapped to the A and B channels respectively to control the chromaticity of the time-frequency diagram.

[0040] Furthermore, for the first amplitude matrix M compand , the first phase matrix P sin , the second phase matrix P cos Mapping is performed to obtain the three-channel time-frequency domain matrices L, A, and B in the LAB color space;

[0041] First, the first amplitude matrix M compand Linear mapping to ±1 to obtain the normalized L channel matrix L = 2M compand -1∈[-1,1] T×F ;

[0042] Definition t,f ∈[-1,1], a t,f ∈[-1,1], b t,f ∈[-1,1] corresponds to the matrix elements of the normalized L, A, and B channel time-frequency domain matrices L, A, and B, respectively, where t∈{1,…,T}, f∈{1,…,F}, then the LAB color space that maximizes chromaticity change is mapped to the bounding sphere surface of the corresponding LAB color space, and then:

[0043]

[0044] For any element l of the L-channel matrix t,f , the equation of the largest circle in the corresponding chromaticity plane is:

[0045]

[0046] Among them, the maximum radius of the chromaticity plane

[0047] definition They correspond to the first phase matrix P sin , the second phase matrix P cos , where t∈{1,…,T},f∈{1,…,F}, the values ​​of the elements of the A and B channel matrices are determined as follows:

[0048]

[0049]

[0050] Furthermore, the three-channel time-frequency domain matrices L, A, and B are converted to the RGB domain to generate a visual time-frequency diagram. Specifically,

[0051] Linearly map the element value range of the L channel matrix to [0,100], and linearly map the element value range of the A and B channel matrices A and B to [-110,110], and obtain the amplitude-phase time-frequency diagram Ψ=cat(3,L,A,B) based on the LAB color space mapping, where the cat function cat(dim,A1,A2,…,An) represents the concatenation of the A1,A2,…,An matrices along its dim dimension;

[0052] The classic LAB to sRGB conversion formula is used to convert the time-frequency diagram Ψ in the LAB color space to the RGB color space, and the visual time-frequency diagram Ψ′ in the RGB color space is obtained.

[0053] A computer system includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the method described above are implemented.

[0054] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the method described above.

[0055] The beneficial effects of the present invention are as follows:

[0056] The present invention applies piecewise linear companding to the initial amplitude matrix obtained from time-frequency signal transformation to enhance the useful signal while suppressing noise. The resulting time-frequency graphs have sharper contrast, richer texture and color details, and are highly capable of suppressing noise.

[0057] The time-frequency diagram generated by the present invention can effectively improve the accuracy of radio signal target detection and recognition.

[0058] The present invention enables the generated time-frequency diagram to simultaneously have amplitude spectrum and phase spectrum information, suppress the influence of noise on the time-frequency diagram, enhance the color expression of the useful signal part in the time-frequency diagram, and improve the visualization effect of the time-frequency diagram. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 This is a flowchart of the steps of the amplitude-phase time-frequency diagram expression method based on LAB color space mapping described in the present invention.

[0060] Figure 2 This is a diagram showing the specific working principle of the amplitude-phase time-frequency diagram expression method based on LAB color space mapping described in the present invention.

[0061] Figure 3 is the statistical histogram of the amplitude matrix elements and the first threshold M on the amplitude normalization max , the second threshold M of noise estimation th .

[0062] Figure 4 It is the function input-output relationship of piecewise linear compression and expansion processing.

[0063] Figure 5 It is a schematic diagram of color distribution in the LAB color space coordinate system.

[0064] Figure 6It is the amplitude spectrum time-frequency diagram of a single-carrier digitally modulated communication signal.

[0065] Figure 7 It is a color diagram of the RGB three channels that directly corresponds to the amplitude spectrum and phase spectrum of a single-carrier digitally modulated communication signal.

[0066] Figure 8 It is the amplitude-phase time-frequency diagram of a single-carrier digitally modulated communication signal based on LAB color space mapping.

[0067] Figure 9 It is the curve of mAP changing with training rounds. DETAILED DESCRIPTION

[0068] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0069] Example 1

[0070] like Figure 1 、 Figure 2 As shown, a method for expressing an amplitude-phase time-frequency diagram based on LAB color space mapping, the method comprising the following steps:

[0071] S1: Perform short-time Fourier transform on the time domain signal and obtain the amplitude and phase values ​​to obtain the initial amplitude matrix and phase matrix;

[0072] S2: Perform piecewise linear compression and expansion on the initial amplitude matrix to obtain a first amplitude matrix; perform differential sine-cosine transform on the initial phase matrix to obtain a first phase matrix and a second phase matrix, thereby obtaining a set of three-channel data consisting of the first amplitude matrix, the first phase matrix, and the second phase matrix;

[0073] S3: Map the obtained three-channel data to the LAB domain to obtain the three-channel time-frequency domain matrices L, A, and B in the LAB color space;

[0074] S4: Convert the three-channel time-frequency domain matrices L, A, and B to the RGB domain to generate a visual time-frequency diagram.

[0075] In a specific embodiment, S1, specifically,

[0076] The time domain signal of the radio is expressed as Where L is the sample length of the signal; the short-time Fourier transform of the time domain signal is expressed as a complex time-frequency matrix Where T and F correspond to the number of time frames and frequency points;

[0077] Then the initial amplitude matrix and phase matrix corresponding to the time-frequency matrix of the time domain signal are expressed as:

[0078]

[0079] P=∠S∈[-π,π] T×F

[0080] Where ∠ is the radian angle operation, M represents the initial amplitude matrix, and P represents the initial phase matrix.

[0081] In a specific embodiment, the initial amplitude matrix is ​​subjected to piecewise linear companding to obtain a first amplitude matrix, as follows:

[0082] Perform statistics on the matrix elements of the initial amplitude matrix M and select the first threshold M max Make M max The matrix element values ​​of the initial amplitude matrix M are not less than 98%;

[0083] The matrix elements of the initial amplitude matrix M are normalized by limiting the amplitude.

[0084]

[0085] Among them, m t,f is the matrix element of the initial amplitude matrix M, t∈{1,…,T},f∈{1,…,F}; m′ t,f is the normalized amplitude matrix M normal ∈[0,1] T×F Matrix elements of ;

[0086] like Figure 3 As shown, the second threshold M for dividing the noise component and the useful signal component in the time-frequency transform amplitude value is determined according to the result of the noise estimation. th ;

[0087] The signal below the second threshold is determined as noise and linearly compressed; the signal above the second threshold is determined as useful signal and linearly expanded; thus, a first amplitude matrix after piecewise linear compression and expansion is obtained.

[0088] In a specific embodiment, Figure 4 As shown, use M compand ∈[0,1] T×F Represents the normalized amplitude matrix M normal The first amplitude matrix after piecewise linear compression and expansion, m″ t,f ,t∈{1,…,T},f∈{1,…,F} represents the first amplitude matrix M compand The matrix element in the tth row and fth column in the t-th row and f-th column. The functional formula of the piecewise linear compression and expansion processing is as follows:

[0089]

[0090] Where k∈(0,1) is the slope of the linear compression part, m″ t,f Represents the first amplitude matrix M compand The matrix element in row t and column f in M compand ∈[0,1] T×F Represents the normalized amplitude matrix M normal The first amplitude matrix for piecewise linear compression and expansion of the line.

[0091] In a specific embodiment,

[0092] Perform differential sine-cosine transform on the initial phase matrix to obtain the first phase matrix and the second phase matrix, as follows:

[0093] Because the time-frequency matrix S compresses the elements of the continuous phase matrix (-∞, +∞) into the range [-π, +π] during the phase calculation process, the phase jumps between -π and +π. This means that even a small phase difference at the boundary of the -π, +π interval may result in a phase jump of 2π in the periodic phase representation, leading to phase jumps and phase ambiguity in the phase spectrum. Furthermore, the phase of the time-frequency matrix is ​​typically more susceptible to noise than the amplitude, resulting in low discrimination between the useful signal and noise in the phase matrix.

[0094] Therefore, considering the different distribution of phase changes between the target signal and the noise, the first-order forward difference transformation of the initial phase matrix P in the frequency dimension is used to increase the discrimination between the target signal and the noise in the phase spectrum, and the differential phase matrix ΔP∈[-2π,2π] is obtained. T×F ,Right now

[0095] Δp t,f =p t,f+1 -p t,f

[0096] Where Δp t,f is the matrix element of the differential phase matrix ΔP, p t,f are the matrix elements of the initial phase matrix P, t∈{1,…,T}, f∈{1,…,F}.

[0097] The differential transformation increases the discrimination between the target signal and noise on the phase spectrum, but the differential phase may still jump between -2π and +2π. Therefore, a triangular transformation is performed on the differential phase matrix to eliminate the radian angle jumps between -2π and +2π. The phase matrix after the triangular transformation is expressed as:

[0098] P sin =sin(ΔP)∈[-1,1] T×F

[0099] P cos=cos(ΔP)∈[-1,1] T×F

[0100] Among them, cos(·) represents trigonometric cosine transform, sin(·) represents trigonometric sine transform, P sin represents the first phase matrix obtained after triangular sine transform, P cos represents the second phase matrix obtained by trigonometric cosine transform.

[0101] In a specific embodiment, the obtained three-channel data is mapped to the LAB domain as follows:

[0102] In LAB color space, Figure 5 As shown, the influence of the three channels L, A, and B on color presentation is not equal.

[0103] Combine the L, A, and B channels, and the first amplitude matrix M compand and the first phase matrix P sin , the second phase matrix P cos The physical meaning of P sin 2 +P cos 2 =1 constraint relationship;

[0104] The first amplitude matrix M compand Mapped to the L channel to control the brightness of the time-frequency graph;

[0105] The first phase matrix P sin , the second phase matrix P cos They are mapped to the A and B channels respectively to control the chromaticity of the time-frequency diagram.

[0106] In a specific embodiment, the first amplitude matrix M compand , the first phase matrix P sin , the second phase matrix P cos Mapping is performed to obtain the three-channel time-frequency domain matrices L, A, and B in the LAB color space;

[0107] First, the first amplitude matrix M compand Linear mapping to ±1 to obtain the normalized L channel matrix L = 2M compand -1∈[-1,1] T×F ;

[0108] Definition t,f ∈[-1,1], a t,f ∈[-1,1], b t,f∈[-1,1] corresponds to the matrix elements of the normalized L, A, and B channel time-frequency domain matrices L, A, and B, respectively, where t∈{1,…,T}, f∈{1,…,F}, then the LAB color space that maximizes chromaticity change is mapped to the bounding sphere surface of the corresponding LAB color space, and then:

[0109]

[0110] For any element l of the L-channel matrix t,f , the equation of the largest circle in the corresponding chromaticity plane is:

[0111]

[0112] Among them, the radius of the largest circle in the chromaticity plane is

[0113] definition They correspond to the first phase matrix P sin , the second phase matrix P cos , where t∈{1,…,T},f∈{1,…,F}, the values ​​of the elements of the A and B channel matrices are determined as follows:

[0114]

[0115]

[0116] In a specific embodiment, the three-channel time-frequency domain matrices L, A, and B are converted to the RGB domain to generate a visual time-frequency diagram. Specifically,

[0117] Linearly map the element value range of the L channel matrix to [0,100], and linearly map the element value range of the A and B channel matrices A and B to [-110,110], and obtain the amplitude-phase time-frequency diagram Ψ=cat(3,L,A,B) based on the LAB color space mapping, where the cat function cat(dim,A1,A2,…,An) represents the concatenation of the A1,A2,…,An matrices along its dim dimension;

[0118] The classic LAB to sRGB conversion formula is used to convert the time-frequency diagram Ψ in the LAB color space to the RGB color space, and the visual time-frequency diagram Ψ′ in the RGB color space is obtained.

[0119] In order to further confirm the technical effect of the present invention.

[0120] This embodiment provides a comparison of the visual presentation effects of time-frequency graphs generated using different methods, as follows:

[0121] Figure 6 、 7, 8 are respectively a time-frequency grayscale image using a single amplitude spectrum, a color image in which the amplitude spectrum and phase spectrum directly correspond to the RGB three channels, and an amplitude-phase time-frequency image based on LAB color space mapping generated by the method adopted by the present invention. Figure 6 、 7 8 shows the time-frequency diagram of the communication signal of the simulated single-carrier digital modulation when the signal-to-noise ratio is 0 dB. Figure 6 、 7 , 8 contains two signal components, BPSK modulation (located on the left side of the time-frequency diagram) and OOK modulation (located on the right side of the time-frequency diagram), where the energy ratio between the BPSK signal and the OOK signal is 6dB. The signal length used to generate the time-frequency diagram is 2 14 , time-frequency transform uses short-time Fourier transform, window function uses Hamming window, window length is 512, sliding window step is 128, and FFT point number is 512.

[0122] contrast Figures 6 to 8 It can be seen that the time-frequency diagram generated by the present invention has more obvious contrast and richer texture and color details than the time-frequency diagrams generated by the other two methods, and has a strong ability to suppress noise presentation that the other two methods do not have.

[0123] This embodiment also provides a performance comparison of signal detection and recognition using time-frequency diagrams generated by different methods, as follows:

[0124] The IQ sample data set of 11 single-carrier modulation signals including BPSK, QPSK, OQPSK, 8PSK, 16QAM, 16APSK, 2ASK, 2FSK, DSB-AM, SSB-AM, and FM is synthesized by simulation, with a training sample size of 1600 and a test sample size of 400. The length of each sample is 2 14 , with a signal-to-noise ratio of -10dB to 10dB. Three time-frequency graph sample sets were obtained for the IQ samples using the amplitude grayscale method, the method where the amplitude spectrum and phase spectrum are directly mapped to the RGB channels, and the method proposed in this invention. The short-time Fourier transform selected a window length of 512, the window function used was the Hamming window, the FFT number of points was 512, and the sliding window step size was 128.

[0125] like Figure 9 As shown, the mean average precision (mAP) result curve is obtained by training and inference of the amplitude spectrum time-frequency graph sample set, the amplitude spectrum + phase spectrum directly corresponding to the RGB three-channel time-frequency graph sample set, and the time-frequency graph sample set generated by the present invention using the YOLO-v5s lightweight model of the image target detection algorithm based on deep learning.

[0126] from Figure 9It can be seen that the recognition accuracy of YOLO-v5s using the time-frequency graph sample set generated by the present invention as input is higher than the recognition accuracy of the amplitude spectrum time-frequency graph sample set and the amplitude spectrum + phase spectrum directly corresponding to the RGB three-channel time-frequency graph sample set. The time-frequency graph sample set generated by the present invention achieved a mAP of 91.43% after convergence, which is a 3.12% improvement over the 88.31% mAP of the amplitude spectrum time-frequency graph sample set and a 2.08% improvement over the 89.35% mAP of the amplitude spectrum + phase spectrum directly corresponding to the RGB three-channel time-frequency graph sample set. This shows that the time-frequency graph generated by the present invention can effectively improve the accuracy of radio signal target detection and recognition.

[0127] Example 2

[0128] A computer system includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the amplitude-phase time-frequency diagram expression method based on LAB color space mapping as described in Example 1 are implemented.

[0129] Example 3

[0130] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the amplitude-phase time-frequency diagram expression method based on LAB color space mapping as described in Example 1.

[0131] Obviously, the above embodiments of the present invention are merely examples for the purpose of illustrating the present invention, and are not intended to limit the embodiments of the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the claims of the present invention.

Claims

1. A method for expressing amplitude-phase time-frequency diagrams based on LAB color space mapping, characterized by: The method comprises the following steps: S1: Perform short-time Fourier transform on the time domain signal and obtain the amplitude and phase values ​​to obtain the initial amplitude matrix and phase matrix; S2: Perform piecewise linear compression and expansion on the initial amplitude matrix to obtain a first amplitude matrix; perform differential sine-cosine transform on the initial phase matrix to obtain a first phase matrix and a second phase matrix, thereby obtaining a set of three-channel data consisting of the first amplitude matrix, the first phase matrix, and the second phase matrix; S3: Map the obtained three-channel data to the LAB domain to obtain the three-channel time-frequency domain matrices L, A, and B in the LAB color space; S4: Convert the three-channel time-frequency domain matrices L, A, and B to the RGB domain to generate a visual time-frequency diagram; The obtained three-channel data is mapped to the LAB domain as follows: Combine the three channels L, A, and B, and the first amplitude matrix M compand and the first phase matrix P sin , the second phase matrix P cos The physical meaning of P sin 2 +P cos 2 =1 constraint relationship; The first amplitude matrix M compand Mapped to the L channel to control the brightness of the time-frequency graph; The first phase matrix P sin , the second phase matrix P cos Mapped to the A and B channels respectively to control the chromaticity of the time-frequency diagram; as well as, For the first amplitude matrix M compand , the first phase matrix P sin , the second phase matrix P cos Mapping is performed to obtain the three-channel time-frequency domain matrices L, A, and B in the LAB color space: First, the first amplitude matrix M compand Linear mapping to ±1 to obtain the normalized L channel matrix L = 2M compand -1∈[-1,1] T×F ; Definition t,f ∈[-1,1], a t,f ∈[-1,1], b t,f ∈[-1,1] corresponds to the matrix elements of the normalized L, A, and B channel time-frequency domain matrices L, A, and B, respectively, where t∈{1,…,T}, f∈{1,…,F}, then the LAB color space that maximizes chromaticity change is mapped to the bounding sphere surface of the corresponding LAB color space, and then: For any element l of the L-channel matrix t,f , the equation of the largest circle in the corresponding chromaticity plane is: Among them, the radius of the largest circle in the chromaticity plane is definition They correspond to the first phase matrix P sin , the second phase matrix P cos , where t∈{1,…,T},f∈{1,…,F}, the values ​​of the elements of the A and B channel matrices are determined as follows:

2. The amplitude-phase time-frequency diagram expression method based on LAB color space mapping according to claim 1 is characterized in that: S1, specifically, The time domain signal of the radio is expressed as Where L is the sample length of the signal; the short-time Fourier transform of the time domain signal is expressed as a complex time-frequency matrix Where T and F correspond to the number of time frames and frequency points; Then the initial amplitude matrix and phase matrix corresponding to the time-frequency matrix of the time domain signal are expressed as: P=∠S∈[-π,π] T×F Where ∠ is the radian angle operation, M represents the initial amplitude matrix, and P represents the initial phase matrix.

3. The amplitude-phase time-frequency diagram expression method based on LAB color space mapping according to claim 2 is characterized in that: The initial amplitude matrix is ​​subjected to piecewise linear compression and expansion processing to obtain the first amplitude matrix, which is as follows: Perform statistics on the matrix elements of the initial amplitude matrix M and select the first threshold M max Make M max The matrix element values ​​of the initial amplitude matrix M are not less than 98%; The matrix elements of the initial amplitude matrix M are normalized by limiting the amplitude. Among them, m t,f is the matrix element of the initial amplitude matrix M, t∈{1,…,T},f∈{1,…,F}; m′ t,f is the normalized amplitude matrix M normal ∈[0,1] T×F Matrix elements of ; Determine the second threshold M for dividing the noise component and the useful signal component in the time-frequency transform amplitude value according to the result of the noise estimation th ; The signal below the second threshold is determined as noise and linearly compressed; the signal above the second threshold is determined as useful signal and linearly expanded; thus, a first amplitude matrix after piecewise linear compression and expansion is obtained.

4. The amplitude-phase time-frequency diagram expression method based on LAB color space mapping according to claim 3 is characterized in that: The functional formula of the piecewise linear compression and expansion process is as follows: Where k∈(0,1) is the slope of the linear compression part, m″ t,f Represents the first amplitude matrix M compand The matrix element in row t and column f in M compand ∈[0,1] T×F Represents the normalized amplitude matrix M normal The first amplitude matrix for piecewise linear companding.

5. The amplitude-phase time-frequency diagram expression method based on LAB color space mapping according to claim 4 is characterized in that: Perform differential sine-cosine transform on the initial phase matrix to obtain the first phase matrix and the second phase matrix, as follows: The first-order forward differential transformation of the initial phase matrix P in the frequency dimension is used to increase the discrimination between the target signal and the noise in the phase spectrum, and the differential phase matrix ΔP∈[-2π,2π] is obtained. T×F ,Right now Δp t,f =p t,f+1 -p t,f Where Δp t,f is the matrix element of the differential phase matrix ΔP, p t,f is the matrix element of the initial phase matrix P, t∈{1,…,T},f∈{1,…,F}; The differential phase matrix is ​​trigonometrically transformed to eliminate the radian angle jumps of -2π and +2π. The phase matrix after trigonometric transformation is expressed as: P sin =sin(ΔP)∈[-1,1] T×F P cos =cos(ΔP)∈[-1,1] T×F Among them, cos(·) represents trigonometric cosine transform, sin(·) represents trigonometric sine transform, P sin represents the first phase matrix obtained after triangular sine transform, P cos represents the second phase matrix obtained by trigonometric cosine transform.

6. The amplitude-phase time-frequency diagram expression method based on LAB color space mapping according to claim 5, characterized in that: Convert the three-channel time-frequency domain matrices L, A, and B to the RGB domain to generate a visual time-frequency diagram. Specifically, Linearly map the element value range of the L channel matrix to [0,100], and linearly map the element value range of the A and B channel matrices A and B to [-110,110], and obtain the amplitude-phase time-frequency diagram Ψ=cat(3,L,A,B) based on the LAB color space mapping, where the cat function cat(dim,A1,A2,…,An) represents the concatenation of the A1,A2,…,An matrices along its dim dimension; The classic LAB to sRGB conversion formula is used to convert the time-frequency diagram Ψ in the LAB color space to the RGB color space, and the visual time-frequency diagram Ψ′ in the RGB color space is obtained.

7. A computer system comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images