Data acquisition and analysis method of RFID distance reading intelligent terminal

Abstract

The invention relates to the technical field of electronic information, in particular to a data acquisition and analysis method of an RFID (Radio Frequency Identification Device) distance reading intelligent terminal, which comprises the following steps of: decomposing a radio frequency signal into a high-order IMF (Inertial Mode Function) component containing a carrier fundamental frequency and a low-frequency residual component reflecting a multipath effect by introducing a variational mode decomposition technology, so as to better separate signal characteristics; by combining segmented windowing, Hilbert transform and phase gradient analysis, the time-frequency characteristics of the phase abrupt change signal segment can be accurately extracted; and finally, a quantitative basis is provided for system optimization by constructing a multipath interference suppression effect evaluation model. According to the method, the anti-interference capability of the RFID reading distance intelligent terminal is analyzed, so that technical support is provided for deployment and performance evaluation of an RFID system in a complex environment.

Description

Technical Field

[0001] The invention relates to the field of electronic information technology, and in particular to a data collection and analysis method for an RFID range reading intelligent terminal. Background Art

[0002] With the rapid development of Internet of Things technology, RFID (Radio Frequency Identification) technology, as a non-contact automatic identification technology, has been widely used in logistics management, intelligent warehousing, asset tracking and other fields. As the core equipment of the RFID system, the performance of the RFID reading intelligent terminal directly affects the recognition accuracy and stability of the system. However, in practical applications, due to factors such as multipath effect, environmental noise and equipment hardware limitations, the collection and analysis of RFID signals often face many challenges. The multipath effect refers to the phenomenon that the radio frequency signal propagates along multiple paths after encountering obstacles during propagation, which will cause signal superposition, phase distortion and power attenuation, seriously reducing the reading accuracy and recognition reliability of the RFID system. Although traditional signal processing methods, such as Fourier transform and wavelet transform, can separate signal features to a certain extent, they have limited effect on nonlinear and non-stationary RFID signal processing, and it is difficult to accurately extract phase mutation features under multipath interference. In addition, the existing technology for analyzing phase mutation signals mostly stays in the single-dimensional processing of time domain or frequency domain, lacks the ability of joint analysis of time and frequency, and cannot fully reflect the nonlinear dynamic characteristics of the signal. Therefore, developing an RFID reading range intelligent terminal data collection and analysis method that can effectively suppress multipath interference, accurately extract phase mutation characteristics and evaluate system performance has become the key to improving the overall performance of the RFID system. Summary of the invention

[0003] The present invention overcomes the shortcomings of the prior art and provides a data collection and analysis method for an RFID reading range intelligent terminal.

[0004] To achieve the above-mentioned purpose, the technical solution adopted by the present invention is: The first aspect of the present invention discloses a data collection and analysis method for an RFID reading range intelligent terminal, comprising the following steps: Acquire the radio frequency signal of the RFID reading intelligent terminal, perform variational modal decomposition processing on the radio frequency signal, and obtain the high-order IMF component including the carrier fundamental frequency and the low-frequency residual component reflecting the multipath effect; Perform segmented windowing and Hilbert transform analysis on each high-order IMF component to obtain the phase mutation signal segment of the RF signal; Performing first-order difference calculation on the instantaneous phase sequence of the phase mutation signal segment to obtain the phase gradient sequence of the phase mutation signal segment, and constructing a phase gradient change curve according to the phase gradient sequence combined with weighted sliding average; Performing characteristic analysis on the phase gradient change curve to obtain mutation characteristic parameters of the phase mutation signal segment, wherein the mutation characteristic parameters include a mutation starting point, an effective peak point, and a mutation duration; The multipath interference suppression effect level of the RFID reading intelligent terminal is evaluated according to the mutation characteristic parameters of the phase mutation signal segment.

[0005] Preferably, the radio frequency signal of the RFID reading intelligent terminal is obtained, and the radio frequency signal is subjected to variational modal decomposition processing to obtain a high-order IMF component including the carrier fundamental frequency and a low-frequency residual component reflecting the multipath effect, specifically: Setting a mode number K and a penalty factor α, initializing a variational mode decomposition framework based on the mode number K and the penalty factor α, and importing the radio frequency signal into the variational mode decomposition framework for iterative decomposition; In each iterative decomposition process, instantaneous frequency smoothness constraints and energy density gradient constraints are simultaneously imposed on each modal component; After each iterative decomposition, the center frequency interval between adjacent modes is detected. When the center frequency interval between any two adjacent modes is less than the preset minimum allowable interval, the bandwidth parameter linkage adjustment mechanism is activated to give priority to compressing the bandwidth upper limit of the high-frequency mode and recalculate the energy constraint condition. An adaptive stop threshold is set based on the energy decay rate of the decomposed residual. When the residual energy changes in three consecutive iterations are all lower than the threshold, the decomposition process is judged to be over and each modal component is output. According to the overlap between the center frequency of each modal component and the theoretical interval of the carrier fundamental frequency and the satisfaction of the dual constraints, the pseudo-modes caused by multipath interference are eliminated, and the effective carrier components that meet the instantaneous frequency stability and spectrum energy focusing are retained; Perform Hilbert marginal spectrum analysis on each effective carrier component to obtain the instantaneous frequency peak and fluctuation variance of each effective carrier component; If the instantaneous frequency peak of a certain effective carrier component is located in the theoretical fluctuation range of the carrier base frequency and its fluctuation variance is less than half of that of the adjacent component, the effective carrier component is determined to be a high-order IMF component; otherwise, the effective carrier component is determined to be a low-frequency residual component.

[0006] Preferably, Hilbert marginal spectrum analysis is performed on each effective carrier component to obtain the instantaneous frequency peak and fluctuation variance of each effective carrier component, specifically: The effective carrier component is preprocessed by segmented windowing, and an adaptive window length adjustment strategy is adopted to dynamically set the window function length according to the center frequency of the effective carrier component so that the signal in the window covers at least three base frequency cycles to suppress spectrum leakage; Performing complex analysis processing on the windowed component signal segment, generating a corresponding analytical signal through Hilbert transform, extracting the real part and the imaginary part of the analytical signal, and cross-arranging and combining the real part and the imaginary part of the analytical signal to form a complex sequence; The instantaneous frequency of the complex sequence is calculated point by point based on a phase difference algorithm, the instantaneous frequency of the complex sequence is fitted into a continuous phase sequence curve according to a cubic spline interpolation method, and the second-order derivative of the phase sequence curve is obtained as an instantaneous frequency sequence; Performing probability density estimation on the instantaneous frequency sequence to obtain a probability density function of the instantaneous frequency, mapping each frequency value and its corresponding probability density in the probability density function of the instantaneous frequency, and constructing a probability density curve of the instantaneous frequency by drawing points and connecting lines with the frequency value as the horizontal axis and the probability density as the vertical axis; The peak position of the probability density curve is taken as the instantaneous frequency peak, and the distribution variance of the probability density curve is taken as the fluctuation variance.

[0007] Preferably, each high-order IMF component is subjected to segmented windowing and Hilbert transform analysis to obtain a phase mutation signal segment of the RF signal, specifically: Implement time-varying adaptive segmented windowing for each high-order IMF component, and dynamically adjust the window function length according to the instantaneous frequency peak of the high-order IMF component to ensure that the signal in the window contains at least two complete phase fluctuation cycles to balance the time-frequency resolution; Each windowed high-order IMF component signal segment is converted into a complex analytical signal by Hilbert transform, and then the phase angle of the complex analytical signal is calculated point by point by inverse tangent operation to obtain the instantaneous phase sequence of each high-order IMF component signal segment; The phase difference values ​​of adjacent instantaneous phase sequences are calculated frame by frame, the calculated phase difference values ​​of adjacent instantaneous phase sequences are weighted, and each weighted result is used as the phase fluctuation intensity index of the corresponding high-order IMF component signal segment; The phase fluctuation intensity index of each high-order IMF component signal segment is compared with a preset index threshold, and the high-order IMF component signal segment whose phase fluctuation intensity index is greater than the preset index threshold is marked as a phase mutation signal segment.

[0008] Preferably, a first-order difference calculation is performed on the instantaneous phase sequence of the phase mutation signal segment to obtain a phase gradient sequence of the phase mutation signal segment, and a phase gradient change curve is constructed according to the phase gradient sequence combined with a weighted sliding average, specifically: Based on the sampling clock of the phase mutation signal segment, the phase mutation signal segment is processed in an equal time manner to generate an equal interval phase sequence based on timestamp alignment; Performing forward difference and backward difference on each equally spaced phase sequence at the same time, calculating the arithmetic mean of the forward difference and the backward difference as the initial value of the instantaneous phase gradient; and symmetrically weighting the initial value of the instantaneous phase gradient to obtain a weighted phase gradient sequence; The weighted phase gradient sequence is subjected to three-level progressive sliding average processing: the first level performs linear weighted accumulation in the window and outputs a rough smoothed gradient; the second level detects residual pulse noise based on the rough smoothing result and uses median filtering to replace abnormal points; the third level applies exponential decay weighting to the filtered data to enhance the contribution of recent data; The phase gradient values ​​after multi-level processing are remapped according to the timestamp to construct a complete phase gradient change curve; and the window length, weight and filter state parameters at each time point are synchronously recorded to form a metadata chain for gradient calculation.

[0009] Preferably, the phase gradient change curve is subjected to characteristic analysis to obtain the mutation characteristic parameters of the phase mutation signal segment, specifically: Calculating a global gradient mean of the phase gradient change curve, and using the global gradient mean as a baseline mean for mutation detection; Performing a forward search on the phase gradient change curve, obtaining a time point in the phase gradient change curve when the gradient exceeds the baseline mean as a preliminary judgment position of the starting point; Perform a first-order derivative sign consistency check on the neighborhood data of the initial position of the starting point. If the derivatives of several consecutive sampling points are positive and the gradient increment increases monotonically, it is confirmed as the mutation starting point. Otherwise, the search range is expanded until the condition is met. In the mutation section, the parabolic interpolation method is used to optimize the sub-sampling accuracy of discrete gradient extreme points, locate the peak gradient point, and calculate the gradient variance between the peak gradient point and the sampling points before and after it. If the gradient variance is less than the preset tolerance and the gradient slope along the descending edge is greater than 70% of the slope along the ascending edge, it is determined to be a valid peak point, otherwise resampling correction is triggered; The time it takes for the gradient to drop from the effective peak point to the baseline mean is traced back as the end point of the mutation, and the absolute time difference from the starting point to the end point is calculated as the duration of the mutation.

[0010] Preferably, the multipath interference suppression effect level of the RFID reading intelligent terminal is evaluated according to the mutation characteristic parameters of the phase mutation signal segment, specifically: Based on the theoretical model of RFID signal propagation in an ideal interference-free environment, combined with the carrier wavelength and the spatial geometric parameters of the range reader, a benchmark model of phase mutation parameters is established. Importing the mutation characteristic parameters of the phase mutation signal segment into the benchmark model: comparing the density of the mutation starting point with the benchmark model to generate a density deviation coefficient; calculating the percentage excess of the peak gradient relative to the theoretical upper limit; and calculating the proportion of abnormal mutations whose duration exceeds the theoretical range; The characteristic parameter weights are determined by the entropy weight method, and the density deviation coefficient, percentage over-limit and abnormal mutation ratio are standardized, and then weighted summation is performed to obtain the multipath interference suppression index. The multipath interference suppression index is input into a pre-trained three-branch convolutional neural network, and the multipath interference suppression effect level is output through a feature fusion layer.

[0011] The second aspect of the present invention discloses a data collection and analysis system for an RFID reading range intelligent terminal. The data collection and analysis system for an RFID reading range intelligent terminal includes a memory and a processor. The memory stores a data collection and analysis method program for the RFID reading range intelligent terminal. When the data collection and analysis method program for the RFID reading range intelligent terminal is executed by the processor, any step of the data collection and analysis method for the RFID reading range intelligent terminal is implemented.

[0012] The third aspect of the present invention discloses a computer-readable storage medium, which includes a data collection and analysis method program for an RFID reading range intelligent terminal. When the data collection and analysis method program for the RFID reading range intelligent terminal is executed by a processor, any step of the data collection and analysis method for the RFID reading range intelligent terminal is implemented.

[0013] The present invention solves the technical defects existing in the background technology, and has the following beneficial effects: by introducing variational mode decomposition technology, the radio frequency signal is decomposed into high-order IMF components containing the carrier fundamental frequency and low-frequency residual components reflecting the multipath effect, so that the signal characteristics can be better separated; combined with segmented windowing, Hilbert transform and phase gradient analysis, the time-frequency characteristics of the phase mutation signal segment can be accurately extracted; finally, by constructing a multipath interference suppression effect evaluation model, a quantitative basis is provided for system optimization. The present invention provides technical support for RFID system deployment and performance evaluation in complex environments by analyzing the anti-interference ability of RFID reading intelligent terminals. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, drawings of other embodiments can be obtained based on these drawings without paying creative work.

[0015] Figure 1 It is a flow chart of the overall method of data collection and analysis method of an RFID reading range intelligent terminal; Figure 2 It is a partial method flow chart of a data collection and analysis method of an RFID reading range intelligent terminal; Figure 3 The system block diagram of a data collection and analysis system for an RFID reading range intelligent terminal. DETAILED DESCRIPTION

[0016] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

[0017] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited to the specific embodiments disclosed below.

[0018] like Figure 1 As shown, the first aspect of the present invention discloses a data collection and analysis method for an RFID reading distance intelligent terminal, comprising the following steps: S102: Acquire a radio frequency signal of an RFID reading intelligent terminal, perform variational mode decomposition processing on the radio frequency signal, and acquire a high-order IMF component including a carrier fundamental frequency and a low-frequency residual component reflecting a multipath effect; S104: performing segmented windowing and Hilbert transform analysis processing on each high-order IMF component to obtain a phase mutation signal segment of the radio frequency signal; S106: performing first-order difference calculation on the instantaneous phase sequence of the phase mutation signal segment to obtain a phase gradient sequence of the phase mutation signal segment, and constructing a phase gradient change curve according to the phase gradient sequence combined with a weighted sliding average; S108: Performing feature analysis on the phase gradient change curve to obtain mutation feature parameters of the phase mutation signal segment, where the mutation feature parameters include a mutation starting point, an effective peak point, and a mutation duration; S110: Evaluate the multipath interference suppression effect level of the RFID reading intelligent terminal according to the mutation characteristic parameters of the phase mutation signal segment.

[0019] Preferably, the radio frequency signal of the RFID reading intelligent terminal is obtained, and the radio frequency signal is subjected to variational modal decomposition processing to obtain a high-order IMF component including the carrier fundamental frequency and a low-frequency residual component reflecting the multipath effect, specifically: Setting a mode number K and a penalty factor α, initializing a variational mode decomposition framework based on the mode number K and the penalty factor α, and importing the radio frequency signal into the variational mode decomposition framework for iterative decomposition; It should be noted that the modal number K is preliminarily estimated based on the spectral characteristics of the RF signal and the theoretical range of the carrier fundamental frequency to ensure that the K value can cover the main frequency components of the signal; secondly, based on the energy distribution characteristics of the signal, the initial value of the penalty factor α is determined by trial and error or an optimization algorithm (such as grid search) to ensure that the α value can balance the bandwidth of the modal component and the separation of the center frequency during the modal decomposition process; then, the variational modal decomposition framework is initialized using K and α to construct the objective function, including the bandwidth constraints and center frequency constraints of the modal components; then, the objective function is solved by an iterative optimization algorithm (such as ADMM), and the frequency and bandwidth of the modal component are updated in each iteration until the convergence conditions are met.

[0020] In each iterative decomposition process, instantaneous frequency smoothness constraints and energy density gradient constraints are simultaneously applied to each modal component; Among them, the instantaneous frequency smoothness constraint calculates the rate of change of the instantaneous frequency in adjacent time windows (such as 1μs window) in real time, dynamically sets the frequency trajectory fluctuation threshold (such as 10MHz / μs), and triggers the local smoothing processing based on secondary frequency domain regularization when a sudden change exceeding the threshold is detected, forcing the correction of the frequency time-varying trajectory of the modal component to suppress the irrational frequency jump caused by multipath reflection; The energy density gradient constraint analyzes the energy ratio of the spectrum main lobe (carrier fundamental frequency) and side lobe (multipath interference) in real time. When the main lobe to side lobe energy ratio is lower than the critical value (such as 3dB), the weight of the gradient penalty term is adaptively enhanced (such as increasing the coefficient α to 1.5 times the initial value). The side lobe energy diffusion is suppressed through frequency domain gradient operation, forcing the spectrum energy to converge to the main lobe. The synergistic effect of the two forms a dual optimization mechanism in the frequency domain - the former ensures the time-frequency continuity of the carrier signal, and the latter enhances the energy resolution of the spectral components, ultimately achieving improved physical isolation accuracy between the carrier baseband component and the multipath residual component.

[0021] After each iterative decomposition, the center frequency interval between adjacent modes is detected. When the center frequency interval between any two adjacent modes is less than the preset minimum allowable interval, the bandwidth parameter linkage adjustment mechanism is activated to give priority to compressing the bandwidth upper limit of the high-frequency mode and recalculate the energy constraint condition. It should be noted that the bandwidth parameter linkage adjustment mechanism is a dynamic adjustment strategy designed to address the problem of frequency overlap of adjacent modes in variational mode decomposition. When it is detected that the center frequency interval of adjacent modes is less than a preset threshold (such as 10MHz), the mechanism reduces the bandwidth upper limit of the high-frequency mode (such as from the initial 50MHz to 40MHz), forcing its spectrum main lobe to shift to low frequency to expand the frequency interval; at the same time, the penalty factor in the energy constraint condition is updated in linkage, and the energy convergence boundary of the mode decomposition is recalculated to prevent signal energy truncation and distortion caused by bandwidth compression.

[0022] An adaptive stop threshold is set based on the energy decay rate of the decomposed residual. When the residual energy changes in three consecutive iterations are all lower than the threshold, the decomposition process is judged to be over and each modal component is output. According to the overlap between the center frequency of each modal component and the theoretical interval of the carrier fundamental frequency and the satisfaction of the dual constraints, the pseudo-modes caused by multipath interference are eliminated, and the effective carrier components that meet the instantaneous frequency stability and spectrum energy focusing are retained; It should be noted that the overlap of the center frequencies of each mode is calculated based on the preset theoretical range of the carrier fundamental frequency (such as 902-928MHz), and the candidate modes with a deviation of less than 15% from the theoretical interval are screened out; then the candidate modes are quantitatively evaluated under dual constraints, and the proportion of their instantaneous frequency change rate exceeding the limit point is statistically no more than 5%, and the main lobe energy ratio must reach more than 85% of the initial decomposition value; then a weighted scoring model is constructed including interval overlap (weight 60%), frequency smoothing compliance rate (weight 20%) and energy focusing (weight 20%), and the modes with a comprehensive score higher than 0.8 enter the next stage; then the Hilbert time-frequency spectrum analysis is used to verify that the instantaneous frequency fluctuation standard deviation of the candidate mode is less than 2MHz and the energy ratio within the 3dB bandwidth of the main lobe exceeds 90%; finally, a multi-dimensional cross-decision is performed, and only the modes that meet the triple standards of scoring threshold, time-frequency stability and energy concentration are retained as effective carrier components, and the rest are determined to be multipath pseudo-modes and are eliminated.

[0023] Perform Hilbert marginal spectrum analysis on each effective carrier component to obtain the instantaneous frequency peak and fluctuation variance of each effective carrier component; If the instantaneous frequency peak of a certain effective carrier component is located in the theoretical fluctuation range of the carrier base frequency and its fluctuation variance is less than half of that of the adjacent component, the effective carrier component is determined to be a high-order IMF component; otherwise, the effective carrier component is determined to be a low-frequency residual component.

[0024] It should be noted that, firstly, by setting the number of modes and the penalty factor, and introducing the instantaneous frequency smoothness constraint and the energy density gradient constraint, the frequency trajectory of the modal component is ensured to be smooth and the spectrum energy is concentrated during the decomposition process, which effectively suppresses the influence of multipath interference on the signal; secondly, by dynamically detecting the center frequency interval of adjacent modes and activating the bandwidth parameter linkage adjustment mechanism, the modal aliasing problem is avoided, and the accuracy and robustness of the decomposition are improved; thirdly, the adaptive stop threshold is set based on the energy decay rate of the decomposition residual, the convergence efficiency of the decomposition process is optimized, and over-decomposition or under-decomposition is avoided; finally, the effective carrier component is screened by Hilbert marginal spectrum analysis (Hilbert marginal spectrum analysis), and the high-order IMF component containing the carrier base frequency and the low-frequency residual component reflecting the multipath effect are accurately separated. The present invention not only improves the accuracy and stability of RFID signal decomposition, but also provides a high-quality signal basis for subsequent phase mutation analysis, multipath interference suppression and system performance evaluation, thereby enhancing the anti-interference ability and recognition reliability of the RFID reading intelligent terminal in complex environments as a whole.

[0025] Preferably, Hilbert marginal spectrum analysis is performed on each effective carrier component to obtain the instantaneous frequency peak and fluctuation variance of each effective carrier component, such as Figure 2 As shown, specifically: S202: performing segmented windowing preprocessing on the effective carrier component, adopting an adaptive window length adjustment strategy, dynamically setting the window function length according to the center frequency of the effective carrier component, so that the signal in the window covers at least three baseband cycles to suppress spectrum leakage; It should be noted that when the effective carrier component is preprocessed by segmented windowing, the fundamental frequency period is first calculated based on its center frequency, and the window function length is dynamically set according to the signal sampling rate to ensure that the window contains at least three complete fundamental frequency periods; then, the Hanning window function is used for windowing processing, and the spectrum leakage is suppressed by the edge attenuation characteristics of the window function; during the window length adjustment process, the instantaneous frequency stability of the signal in the window is monitored in real time. If the frequency fluctuation exceeds the preset threshold, the window length is dynamically expanded to cover more periods to ensure the consistency of frequency characteristics within the signal segment; finally, the windowed signal segments are overlapped and segmented, and the weighted average method is used to smooth the segment boundaries to further reduce the spectral distortion introduced by the window function switching, thereby providing high-quality signal segmentation for subsequent Hilbert transform and frequency analysis.

[0026] S204: performing complex analysis processing on the windowed component signal segments, generating corresponding analytical signals through Hilbert transform, extracting the real part and the imaginary part of the analytical signal, and cross-arranging and combining the real part and the imaginary part of the analytical signal to form a complex sequence; It should be noted that when performing complex analysis on the windowed component signal segment, the real signal is first converted into a complex signal through Hilbert transform to generate the corresponding analytical signal; then, the real part (i.e., the original signal) and the imaginary part (i.e., the orthogonal component after Hilbert transform) of the analytical signal are extracted, and the real part and the imaginary part are cross-arranged and combined in chronological order to form a complex sequence; in the arrangement process, it is ensured that the real part and the imaginary part at each time point correspond one-to-one to form a complete complex sequence; finally, the complex sequence is normalized to eliminate the influence of amplitude fluctuations on subsequent analysis, thereby providing high-quality complex signal data for instantaneous frequency calculation.

[0027] S206: calculating the instantaneous frequency of the complex sequence point by point based on a phase difference algorithm, fitting the instantaneous frequency of the complex sequence into a continuous phase sequence curve according to a cubic spline interpolation method, and obtaining the second-order derivative of the phase sequence curve as an instantaneous frequency sequence; It should be noted that when the instantaneous frequency of a complex sequence is calculated point by point based on the phase difference algorithm, the phase angle of the complex sequence is first extracted by the inverse tangent function, and the differential value of adjacent phase angles is calculated to obtain a discrete sequence of instantaneous frequencies; then, the cubic spline interpolation method is used to perform curve fitting on the discrete instantaneous frequencies to generate a continuous and smooth phase sequence curve to ensure the continuity of frequency changes.

[0028] S208: performing probability density estimation on the instantaneous frequency sequence to obtain a probability density function of the instantaneous frequency, mapping each frequency value and its corresponding probability density in the probability density function of the instantaneous frequency, and constructing a probability density curve of the instantaneous frequency by drawing points and connecting lines, with the frequency value as the horizontal axis and the probability density as the vertical axis; S210: Taking the peak position of the probability density curve as the instantaneous frequency peak, and taking the distribution variance of the probability density curve as the fluctuation variance.

[0029] In summary, through the steps, the present invention improves the accuracy and stability of instantaneous frequency analysis, and also provides key data support for subsequent signal feature extraction, multipath interference suppression and system performance evaluation, thereby overall improving the signal processing capability and anti-interference performance of the RFID reading intelligent terminal in complex environments.

[0030] Preferably, each high-order IMF component is subjected to segmented windowing and Hilbert transform analysis to obtain a phase mutation signal segment of the RF signal, specifically: Implement time-varying adaptive segmented windowing for each high-order IMF component, and dynamically adjust the window function length according to the instantaneous frequency peak of the high-order IMF component to ensure that the signal in the window contains at least two complete phase fluctuation cycles to balance the time-frequency resolution; It should be noted that the instantaneous frequency sequence of the high-order IMF component is extracted by Hilbert transform, and its peak frequency is calculated; the fundamental frequency period is dynamically calculated according to the peak frequency, and the window function length is determined based on the sampling rate to ensure that the window contains at least two complete phase fluctuation periods; in the process of window length adjustment, the instantaneous frequency stability of the signal in the window is monitored in real time. If the frequency fluctuation exceeds the preset threshold, the window length is dynamically expanded to cover more periods to ensure the consistency of frequency characteristics within the signal segment; finally, the Hanning window function is used to window the signal segment, and the edge attenuation characteristics of the window function are used to suppress spectrum leakage, and the segment boundaries are smoothed in combination with the overlapping segmentation strategy, thereby providing high-quality time-frequency localized signal segmentation for subsequent Hilbert transform and phase mutation analysis.

[0031] Each windowed high-order IMF component signal segment is converted into a complex analytical signal by Hilbert transform, and then the phase angle of the complex analytical signal is calculated point by point by inverse tangent operation to obtain the instantaneous phase sequence of each high-order IMF component signal segment; It should be noted that the windowed signal segment is Hilbert transformed to generate orthogonal components, which are combined with the original signal segment into a complex analytical signal; then, the real part (i.e., the original signal) and the imaginary part (i.e., the orthogonal components) of the complex analytical signal are extracted, and the ratio of the real part to the imaginary part is calculated point by point; then, the phase angle at each time point is calculated by inverse tangent operation to obtain the instantaneous phase sequence; finally, the instantaneous phase sequence is phase unwrapping processed to eliminate phase jumps and ensure the continuity and consistency of the phase sequence, thereby providing high-precision instantaneous phase data for subsequent phase mutation detection.

[0032] The phase difference values ​​of adjacent instantaneous phase sequences are calculated frame by frame, the calculated phase difference values ​​of adjacent instantaneous phase sequences are weighted, and each weighted result is used as the phase fluctuation intensity index of the corresponding high-order IMF component signal segment; The phase fluctuation intensity index of each high-order IMF component signal segment is compared with a preset index threshold, and the high-order IMF component signal segment whose phase fluctuation intensity index is greater than the preset index threshold is marked as a phase mutation signal segment.

[0033] In summary, a time-varying adaptive segmented windowing strategy is adopted to dynamically adjust the window function length according to the instantaneous frequency peak of the high-order IMF component to ensure that the signal in the window contains at least two complete phase fluctuation cycles, effectively balancing the time-frequency resolution and avoiding the loss of frequency characteristics or insufficient time resolution caused by the fixed window length; secondly, the windowed signal segment is converted into a complex analytical signal through the Hilbert transform, and the instantaneous phase sequence is calculated point by point in combination with the inverse tangent operation, providing high-precision phase data for phase mutation detection; thirdly, the phase difference value of the adjacent instantaneous phase sequence is calculated frame by frame and weighted to generate a phase fluctuation intensity index, which can quantify the severity of the phase change and thus accurately identify the phase mutation area; finally, by comparing the phase fluctuation intensity index with the preset threshold, the phase mutation signal segment is marked, providing key data support for subsequent multipath interference analysis and system performance optimization.

[0034] Preferably, a first-order difference calculation is performed on the instantaneous phase sequence of the phase mutation signal segment to obtain a phase gradient sequence of the phase mutation signal segment, and a phase gradient change curve is constructed according to the phase gradient sequence combined with a weighted sliding average, specifically: Based on the sampling clock of the phase mutation signal segment, the phase mutation signal segment is processed in an equal time manner to generate an equal interval phase sequence based on timestamp alignment; It should be noted that the time interval is determined according to the sampling rate, and the signal segment is divided into equal-time segments at fixed time intervals; the phase data in each equal-time segment is processed by cubic spline interpolation to generate phase values ​​aligned with equally spaced timestamps to ensure the time consistency of the phase sequence; then, the interpolated phase sequence is detected for abnormal points. If an abnormality is found, the mean of the adjacent points is used to replace it to ensure the integrity of the data; finally, the equally spaced phase sequence is normalized to eliminate the influence of amplitude fluctuations on subsequent analysis, thereby providing high-quality equally spaced phase data for instantaneous phase gradient calculation.

[0035] Perform forward difference (difference between current point and next phase point) and backward difference (difference between current point and previous phase point) on each equally spaced phase sequence, calculate the arithmetic mean of the forward difference and the backward difference as the initial value of the instantaneous phase gradient; and symmetrically weight the initial value of the instantaneous phase gradient to obtain a weighted phase gradient sequence; The weighted phase gradient sequence is subjected to three-level progressive sliding average processing: the first level performs linear weighted accumulation in the window and outputs a rough smoothed gradient; the second level detects residual pulse noise based on the rough smoothing result and uses median filtering to replace abnormal points; the third level applies exponential decay weighting to the filtered data to enhance the contribution of recent data; It should be noted that when performing three-level progressive sliding average processing on the weighted phase gradient sequence, first, in the first level of processing, the Hanning window function is used to perform linear weighted accumulation on the gradient sequence to generate a coarse smoothed gradient sequence; then, in the second level of processing, the residual pulse noise is detected based on the coarse smoothing result, the abnormal points are identified by calculating the local variance, and the median filter is used to replace the abnormal points to ensure the smoothness of the gradient sequence; then, in the third level of processing, exponential decay weighting is applied to the filtered gradient sequence, with the current moment as the center, a higher weight coefficient is given to the recent data to enhance the contribution of the recent data; finally, the gradient sequence after the three-level processing is normalized to eliminate the amplitude deviation introduced by the weighting, thereby providing high-quality data support for constructing the phase gradient change curve.

[0036] The phase gradient values ​​after multi-level processing are remapped according to the timestamp to construct a complete phase gradient change curve; and the window length, weight and filter state parameters at each time point are synchronously recorded to form a metadata chain for gradient calculation.

[0037] In summary, this paper improves the accuracy and stability of phase gradient analysis, providing a reliable technical guarantee for signal processing and performance optimization of RFID reading intelligent terminals in complex environments.

[0038] Preferably, the phase gradient change curve is subjected to characteristic analysis to obtain the mutation characteristic parameters of the phase mutation signal segment, specifically: Calculating a global gradient mean of the phase gradient change curve, and using the global gradient mean as a baseline mean for mutation detection; It should be noted that the phase gradient change curve is segmented and the curve is divided into multiple equal-length intervals; then, the arithmetic mean of the gradient values ​​in each interval is calculated as the local gradient mean; then, all local gradient means are weighted averaged, where the weight coefficient is dynamically adjusted based on the gradient stability of each interval, and intervals with smaller gradient fluctuations are given higher weights; finally, the weighted average result is used as the global gradient mean and as the baseline mean for mutation detection, providing a reliable reference benchmark for subsequent initial judgment of the starting point and positioning of the peak gradient point.

[0039] Performing a forward search on the phase gradient change curve, obtaining a time point in the phase gradient change curve when the gradient exceeds the baseline mean as a preliminary judgment position of the starting point; It should be noted that when the phase gradient change curve is searched in the forward direction, first start from the starting point of the curve, compare the gradient value with the baseline mean point by point, and record the time point when the gradient first exceeds the baseline mean as the initial position of the starting point; then, perform a secondary verification in the neighborhood of the initial position, calculate the mean and variance of the gradient value in the neighborhood, if the mean is higher than the baseline mean and the variance is lower than the preset threshold, then confirm it as the initial position of the starting point; finally, record the timestamp and gradient value of the initial position to provide initial data support for the subsequent first-order derivative sign consistency detection.

[0040] Perform a first-order derivative sign consistency check on the neighborhood data of the initial position of the starting point. If the derivatives of a number of consecutive (such as 5, 8 or 10) sampling points are all positive and the gradient increment increases monotonically, it is confirmed as the mutation starting point. Otherwise, the search range is expanded until the condition is met; It should be noted that, starting from the initial judgment position, the first-order derivative of the gradient value is calculated point by point backward, and the sign of the derivative is detected; then, the derivative signs of several consecutive sampling points (preferably 8) are counted, and if they are all positive and the gradient increment increases monotonically, it is confirmed as the mutation starting point; if the conditions are not met, the search range is expanded, and several sampling points are moved forward or backward, and the above detection process is repeated until a mutation starting point that meets the conditions is found; finally, the timestamp and gradient value of the mutation starting point are recorded to provide accurate starting position information for the subsequent peak gradient point positioning.

[0041] In the mutation section, the parabolic interpolation method is used to optimize the sub-sampling accuracy of discrete gradient extreme points, locate the peak gradient point, and calculate the gradient variance between the peak gradient point and the sampling points before and after it. It should be noted that, in the mutation section, the discrete extreme points in the gradient sequence are first identified, and the gradient values ​​and corresponding timestamps of the extreme point and the two sampling points before and after it are extracted; then, a parabola model is constructed based on these three points, and the vertex coordinates of the parabola are calculated as the peak gradient point after sub-sampling accuracy optimization; then, the rationality of the parabola vertex is verified, and if the vertex is located in the neighborhood of the extreme point and the gradient value is higher than the surrounding points, it is confirmed as the peak gradient point; finally, the timestamp and gradient value of the peak gradient point are recorded to provide accurate peak position information for subsequent gradient variance calculation and falling edge slope analysis.

[0042] If the gradient variance is less than the preset tolerance and the gradient slope along the descending edge is greater than 70% of the slope along the ascending edge, it is determined to be a valid peak point, otherwise resampling correction is triggered; It should be noted that when calculating the gradient variance of a peak gradient point and the sampling points before and after it, the gradient values ​​of the peak gradient point and the two sampling points before and after it are first extracted, and the gradient variance of these five points is calculated; then, the gradient variance is compared with the preset tolerance. If the variance is less than the tolerance, the slope of the gradient sequence before and after the peak point, that is, the rising edge slope and the falling edge slope, is further calculated; then, the ratio of the falling edge slope to the rising edge slope is compared. If the falling edge slope is greater than 70% of the rising edge slope, it is determined to be a valid peak point; if the conditions are not met, a resampling correction is triggered, the sampling range is expanded or the interpolation model is adjusted, and the gradient variance and slope are recalculated until a valid peak point that meets the conditions is found.

[0043] The time it takes for the gradient to drop from the effective peak point to the baseline mean is traced back as the end point of the mutation, and the absolute time difference from the starting point to the end point is calculated as the duration of the mutation.

[0044] In summary, by calculating the global gradient mean as the baseline for mutation detection, a reliable reference benchmark is provided for the initial judgment of the starting point; the forward search combined with the first-order derivative sign consistency detection is used to accurately identify the mutation starting point, ensuring the robustness and accuracy of the starting point detection; the subsampling accuracy of the discrete gradient extreme points is optimized by the parabolic interpolation method, the peak gradient point is located, and its effectiveness is verified by combining the gradient variance and the falling edge slope, ensuring the accuracy and stability of the peak gradient point; by tracing the time for the gradient to drop from the effective peak point to the baseline mean, the mutation duration is calculated, providing complete time dimension information for the characteristic description of the phase mutation signal segment.

[0045] Preferably, the multipath interference suppression effect level of the RFID reading intelligent terminal is evaluated according to the mutation characteristic parameters of the phase mutation signal segment, specifically: Based on the theoretical model of RFID signal propagation in an ideal interference-free environment, combined with the carrier wavelength and the spatial geometric parameters of the range reader, a benchmark model of phase mutation parameters is established. It should be noted that according to the carrier wavelength, the spatial geometric parameters of the range-reading terminal and the signal propagation path, the theoretical phase mutation starting point density, peak gradient upper limit and mutation duration range are calculated; then, the phase mutation signal in an ideal interference-free environment is generated through simulation, and its mutation characteristic parameters are extracted as the benchmark value; then, combined with the multipath effect and noise interference in the actual environment, the benchmark value is dynamically corrected to ensure that the benchmark model can reflect the signal characteristics in the real scene; finally, the corrected benchmark value is stored as the benchmark model of the phase mutation parameter. In general, the benchmark model is a reference framework constructed based on the radio frequency identification signal propagation theory in an ideal interference-free environment, which includes the starting point density, peak gradient upper limit and mutation duration range unidirectional theoretical range. The benchmark model is obtained through prior training.

[0046] Importing the mutation characteristic parameters of the phase mutation signal segment into the benchmark model: comparing the density of the mutation starting point (the number of mutations per unit time) with the benchmark model to generate a density deviation coefficient; calculating the percentage excess of the peak gradient relative to the theoretical upper limit; and calculating the proportion of abnormal mutations whose duration exceeds the theoretical range; It should be noted that first, the actual density of the mutation starting points per unit time is calculated and compared with the theoretical density in the benchmark model to generate the density deviation coefficient; then, the actual peak gradient value is extracted and its percentage excess relative to the theoretical upper limit in the benchmark model is calculated; then, the proportion of abnormal mutations in the phase mutation signal segment whose mutation duration exceeds the theoretical range of the benchmark model is statistically analyzed; finally, the density deviation coefficient, percentage excess and abnormal mutation proportion are used as key indicators for evaluating the multipath interference suppression effect.

[0047] The characteristic parameter weights are determined by the entropy weight method, and the density deviation coefficient, percentage over-limit and abnormal mutation ratio are standardized, and then weighted summation is performed to obtain the multipath interference suppression index. It should be noted that when determining the weights of characteristic parameters by the entropy weight method, the information entropy of the density deviation coefficient, the percentage of excess limit and the proportion of abnormal mutations is first calculated, and the weight coefficient of each parameter is determined according to the difference in information entropy; then, the three characteristic parameters are standardized and converted into dimensionless values ​​to ensure comparison under the same dimension; then, the standardized characteristic parameters are multiplied by their corresponding weight coefficients, and weighted summation is performed to obtain the multipath interference suppression index; finally, the suppression index is normalized so that its value range is between 0 and 1, providing a unified quantitative indicator for the subsequent suppression effect level classification.

[0048] The multipath interference suppression index is input into a pre-trained three-branch convolutional neural network, and the multipath interference suppression effect level (including excellent / good / medium / poor) is output through a feature fusion layer.

[0049] It should be noted that when the multipath interference suppression index is input into the pre-trained three-branch convolutional neural network, the suppression index is first mapped into a feature vector through the input layer and input into three parallel convolution branches respectively; then, each convolution branch performs convolution and pooling operations on the feature vector to extract feature information at different levels; then, the output features of the three branches are concatenated and weighted fused in the feature fusion layer to generate a comprehensive feature vector; finally, through the fully connected layer and the Softmax classifier, the comprehensive feature vector is mapped into a multipath interference suppression effect level (excellent / good / medium / poor), providing intelligent decision support for the performance optimization of the RFID reading range intelligent terminal.

[0050] It should be noted that the specific method and steps for training the three-branch convolutional neural network are as follows: first, a large amount of RFID signal data containing different degrees of multipath interference is collected, and the corresponding suppression effect levels (excellent / good / medium / poor) are marked; then, the signal data is preprocessed, including normalization, segmentation and feature extraction, to generate training samples; then, a three-branch convolutional neural network architecture is constructed, each branch contains a convolution layer, a pooling layer and a fully connected layer, which are used to extract feature information at different levels; then, the training samples are input into the network, the network parameters are optimized by the back propagation algorithm, and cross-validation and early stopping strategies are used during the training process to prevent overfitting; finally, the network performance is evaluated on an independent test set, and the network structure and hyperparameters are adjusted until the predetermined classification accuracy and generalization ability are achieved, thereby completing the pre-training of the network.

[0051] In this embodiment, the data collection and analysis method of the RFID reading intelligent terminal may further include the following steps: Step 1: Based on the mutation starting point, peak gradient and duration parameters in the phase gradient change curve, the instantaneous gradient slope at the mutation starting point, the phase jump peak rate and the gradient saturation interval length are extracted as input parameters for gradient compensation; Step 2: construct a hyperbolic tangent compensation function according to the input parameters: adjust the curvature parameter of the hyperbolic tangent function to match the phase jump peak rate, so that the instantaneous gradient slope of its derivative at the mutation starting point and the original gradient slope form a negative offset; at the same time, set the saturation threshold of the compensation function according to the length of the gradient saturation interval to ensure that the compensation effect is only effective in the mutation active interval to avoid interference with the stable segment signal; Step 3: Apply the compensation function to the neighborhood of the mutation starting point of the original phase sequence to generate a corrected phase sequence; synchronously calculate the phase difference residuals before and after compensation, and if the sliding window mean of the absolute value of the residual exceeds the preset tolerance, reversely adjust the curvature and saturation threshold of the hyperbolic tangent function according to the residual distribution characteristics until the residual converges to the tolerance range; Step 4: after the mutation end point, define a stable segment interval, extract the phase difference between the corrected phase sequence at the start and end of the stable segment, and correct the residual accumulated phase offset introduced by the compensation operation according to the phase difference; Step 5: Compare the phase gradient change curves before and after correction, and calculate the peak gradient attenuation rate of the mutation segment, the phase fluctuation variance of the stable segment, and the Hilbert spectrum entropy value of the overall signal. If any indicator does not reach the preset optimization target, return to step 2 to adjust the compensation function parameters to form a closed-loop optimization link, and finally achieve the suppression of phase jumps and the global elimination of system errors.

[0052] It should be noted that the instantaneous gradient slope, phase jump peak rate and gradient saturation interval length of the mutation starting point are extracted based on the phase gradient change curve, which provides a scientific basis for the construction of the compensation function; by adjusting the curvature parameter and saturation threshold of the hyperbolic tangent function, the phase jump rate is dynamically matched and the compensation range is limited to ensure the efficiency and locality of the compensation effect; by real-time monitoring of the phase difference residual before and after compensation and reverse adjustment of the compensation parameters, the iterative optimization of the compensation accuracy is achieved; the phase difference value is extracted in the stable segment interval and the residual accumulated offset is corrected to eliminate the global error introduced by the compensation operation; finally, by comparing the phase gradient change curves before and after correction, the peak attenuation rate of the mutation segment gradient, the phase fluctuation variance of the stable segment and the Hilbert spectrum entropy value of the overall signal are statistically calculated to form a closed-loop optimization link to ensure the global optimality of the phase jump suppression effect and the elimination of system errors. This method constructs a closed-loop optimization link, dynamically adapts to the mutation characteristics caused by the environment of the RFID reading intelligent terminal, and realizes the effective suppression of phase jumps in RFID reading data and the precise correction of global errors.

[0053] In this embodiment, the data collection and analysis method of the RFID reading intelligent terminal may further include the following steps: After the RFID reading intelligent terminal is phase-corrected, the radio frequency signal of the RFID reading intelligent terminal is collected in real time, and the radio frequency signal is decomposed based on variational mode to obtain a low-frequency residual component collected in real time; The residual component collected in real time is aligned with the reference template frame by frame, and the cross-correlation entropy value between the two in the time-frequency domain is calculated; If the cross-correlation entropy value is lower than a preset threshold, it is determined that multipath interference suppression is effective; if the cross-correlation entropy value exceeds the threshold, it is marked as an insufficiently suppressed area and its spatiotemporal distribution characteristics are recorded; Generate parameter adjustment vectors based on the spatiotemporal distribution characteristics of the insufficiently suppressed area: for high-frequency phase jump areas, increase the curvature adjustment coefficient of the compensation function proportionally to enhance the negative gradient compensation strength; for low-frequency residual accumulation areas, increase the weight coefficient of the linear offset term to strengthen global error suppression; The multipath effect is continued to be suppressed in the above manner.

[0054] It should be noted that the low-frequency residual component collected in real time is extracted based on variational mode decomposition, which provides a high-precision signal basis for the quantitative evaluation of the multipath interference suppression effect; by calculating the time-frequency domain cross-correlation entropy value between the residual component and the reference template, it is possible to accurately identify the insufficient suppression area and record its spatiotemporal distribution characteristics, providing a scientific basis for parameter adjustment; according to the different characteristics of the high-frequency phase jump area and the low-frequency residual accumulation area, the curvature adjustment coefficient and the weight coefficient of the linear offset term of the compensation function are dynamically adjusted to achieve accurate suppression of multipath interference; finally, by continuously optimizing the compensation function parameters, a closed-loop feedback mechanism is formed to improve the real-time and robustness of multipath interference suppression, providing a reliable technical guarantee for the stable operation of RFID reading intelligent terminals in complex environments.

[0055] like Figure 3 As shown, the second aspect of the present invention discloses a data collection and analysis system 6 for an RFID reading intelligent terminal, wherein the data collection and analysis system for an RFID reading intelligent terminal comprises a memory 41 and a processor 52, wherein the memory 41 stores a data collection and analysis method program for an RFID reading intelligent terminal, and when the data collection and analysis method program for the RFID reading intelligent terminal is executed by the processor 52, any step of the data collection and analysis method for the RFID reading intelligent terminal is implemented.

[0056] The third aspect of the present invention discloses a computer-readable storage medium, which includes a data collection and analysis method program for an RFID reading range intelligent terminal. When the data collection and analysis method program for the RFID reading range intelligent terminal is executed by a processor, any step of the data collection and analysis method for the RFID reading range intelligent terminal is implemented.

[0057] The above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A data collection and analysis method for an RFID reading range intelligent terminal, characterized in that: The following steps are involved: Acquire the radio frequency signal of the RFID reading intelligent terminal, perform variational modal decomposition processing on the radio frequency signal, and obtain the high-order IMF component including the carrier fundamental frequency and the low-frequency residual component reflecting the multipath effect; Perform segmented windowing and Hilbert transform analysis on each high-order IMF component to obtain the phase mutation signal segment of the RF signal; Performing first-order difference calculation on the instantaneous phase sequence of the phase mutation signal segment to obtain the phase gradient sequence of the phase mutation signal segment, and constructing a phase gradient change curve according to the phase gradient sequence combined with weighted sliding average; Performing characteristic analysis on the phase gradient change curve to obtain mutation characteristic parameters of the phase mutation signal segment, wherein the mutation characteristic parameters include a mutation starting point, an effective peak point, and a mutation duration; The multipath interference suppression effect level of the RFID reading intelligent terminal is evaluated according to the mutation characteristic parameters of the phase mutation signal segment.

2. The data collection and analysis method of the RFID reading intelligent terminal according to claim 1 is characterized in that: The radio frequency signal of the RFID reading intelligent terminal is obtained, and the radio frequency signal is subjected to variational mode decomposition processing to obtain a high-order IMF component including the carrier fundamental frequency and a low-frequency residual component reflecting the multipath effect, specifically: Setting a mode number K and a penalty factor α, initializing a variational mode decomposition framework based on the mode number K and the penalty factor α, and importing the radio frequency signal into the variational mode decomposition framework for iterative decomposition; In each iterative decomposition process, instantaneous frequency smoothness constraints and energy density gradient constraints are simultaneously applied to each modal component; After each iterative decomposition, the center frequency interval between adjacent modes is detected. When the center frequency interval between any two adjacent modes is less than the preset minimum allowable interval, the bandwidth parameter linkage adjustment mechanism is activated to give priority to compressing the bandwidth upper limit of the high-frequency mode and recalculate the energy constraint condition. An adaptive stop threshold is set based on the energy decay rate of the decomposed residual. When the residual energy changes in three consecutive iterations are all lower than the threshold, the decomposition process is judged to be over and each modal component is output. According to the overlap between the center frequency of each modal component and the theoretical interval of the carrier fundamental frequency and the satisfaction of the dual constraints, the pseudo-modes caused by multipath interference are eliminated, and the effective carrier components that meet the instantaneous frequency stability and spectrum energy focusing are retained; Perform Hilbert marginal spectrum analysis on each effective carrier component to obtain the instantaneous frequency peak and fluctuation variance of each effective carrier component; If the instantaneous frequency peak of a certain effective carrier component is within the theoretical fluctuation range of the carrier fundamental frequency and its fluctuation variance is less than half of that of the adjacent component, the effective carrier component is determined to be a high-order IMF component; Otherwise, the effective carrier component is determined as a low-frequency residual component.

3. The data collection and analysis method of the RFID reading intelligent terminal according to claim 2 is characterized in that: Perform Hilbert marginal spectrum analysis on each effective carrier component to obtain the instantaneous frequency peak and fluctuation variance of each effective carrier component, specifically: The effective carrier component is preprocessed by segmented windowing, and an adaptive window length adjustment strategy is adopted to dynamically set the window function length according to the center frequency of the effective carrier component so that the signal in the window covers at least three base frequency cycles to suppress spectrum leakage; Performing complex analysis processing on the windowed component signal segment, generating a corresponding analytical signal through Hilbert transform, extracting the real part and the imaginary part of the analytical signal, and cross-arranging and combining the real part and the imaginary part of the analytical signal to form a complex sequence; The instantaneous frequency of the complex sequence is calculated point by point based on a phase difference algorithm, the instantaneous frequency of the complex sequence is fitted into a continuous phase sequence curve according to a cubic spline interpolation method, and the second-order derivative of the phase sequence curve is obtained as an instantaneous frequency sequence; Performing probability density estimation on the instantaneous frequency sequence to obtain a probability density function of the instantaneous frequency, mapping each frequency value and its corresponding probability density in the probability density function of the instantaneous frequency, and constructing a probability density curve of the instantaneous frequency by drawing points and connecting lines with the frequency value as the horizontal axis and the probability density as the vertical axis; The peak position of the probability density curve is taken as the instantaneous frequency peak, and the distribution variance of the probability density curve is taken as the fluctuation variance.

4. The data collection and analysis method of the RFID reading intelligent terminal according to claim 1 is characterized in that: Each high-order IMF component is subjected to segmented windowing and Hilbert transform analysis to obtain the phase mutation signal segment of the RF signal, specifically: Implement time-varying adaptive segmented windowing for each high-order IMF component, and dynamically adjust the window function length according to the instantaneous frequency peak of the high-order IMF component to ensure that the signal in the window contains at least two complete phase fluctuation cycles to balance the time-frequency resolution; Each windowed high-order IMF component signal segment is converted into a complex analytical signal by Hilbert transform, and then the phase angle of the complex analytical signal is calculated point by point by inverse tangent operation to obtain the instantaneous phase sequence of each high-order IMF component signal segment; The phase difference values ​​of adjacent instantaneous phase sequences are calculated frame by frame, the calculated phase difference values ​​of adjacent instantaneous phase sequences are weighted, and each weighted result is used as the phase fluctuation intensity index of the corresponding high-order IMF component signal segment; The phase fluctuation intensity index of each high-order IMF component signal segment is compared with a preset index threshold, and the high-order IMF component signal segment whose phase fluctuation intensity index is greater than the preset index threshold is marked as a phase mutation signal segment.

5. The data collection and analysis method of the RFID reading intelligent terminal according to claim 1 is characterized in that: The instantaneous phase sequence of the phase mutation signal segment is calculated by first-order difference to obtain the phase gradient sequence of the phase mutation signal segment, and a phase gradient change curve is constructed according to the phase gradient sequence combined with weighted sliding average, specifically: Based on the sampling clock of the phase mutation signal segment, the phase mutation signal segment is processed in an equal time manner to generate an equal interval phase sequence based on timestamp alignment; Performing forward difference and backward difference on each equally spaced phase sequence at the same time, calculating the arithmetic mean of the forward difference and the backward difference as the initial value of the instantaneous phase gradient; and symmetrically weighting the initial value of the instantaneous phase gradient to obtain a weighted phase gradient sequence; The weighted phase gradient sequence is subjected to three-level progressive sliding average processing: the first level performs linear weighted accumulation within the window and outputs a rough smoothed gradient; the second level detects residual pulse noise based on the rough smoothing result and uses median filtering to replace abnormal points; The third stage applies exponential decay weighting to the filtered data to enhance the contribution of recent data; The phase gradient values ​​after multi-level processing are remapped according to the timestamp to construct a complete phase gradient change curve; and the window length, weight and filter state parameters at each time point are synchronously recorded to form a metadata chain for gradient calculation.

6. The data collection and analysis method of the RFID reading intelligent terminal according to claim 1 is characterized in that: Performing feature analysis on the phase gradient change curve to obtain the mutation feature parameters of the phase mutation signal segment is specifically: Calculating a global gradient mean of the phase gradient change curve, and using the global gradient mean as a baseline mean for mutation detection; Performing a forward search on the phase gradient change curve, obtaining a time point in the phase gradient change curve when the gradient exceeds the baseline mean as a preliminary judgment position of the starting point; Perform a first-order derivative sign consistency check on the neighborhood data of the initial position of the starting point. If the derivatives of several consecutive sampling points are positive and the gradient increment increases monotonically, it is confirmed as the mutation starting point. Otherwise, the search range is expanded until the condition is met. In the mutation section, the parabolic interpolation method is used to optimize the sub-sampling accuracy of discrete gradient extreme points, locate the peak gradient point, and calculate the gradient variance between the peak gradient point and the sampling points before and after it. If the gradient variance is less than the preset tolerance and the gradient slope along the descending edge is greater than 70% of the slope along the ascending edge, it is determined to be a valid peak point, otherwise resampling correction is triggered; The time it takes for the gradient to drop from the effective peak point to the baseline mean is traced back as the end point of the mutation, and the absolute time difference from the starting point to the end point is calculated as the duration of the mutation.

7. The data collection and analysis method of an RFID reading distance intelligent terminal according to claim 1 is characterized in that: The multipath interference suppression effect level of the RFID reading intelligent terminal is evaluated according to the mutation characteristic parameters of the phase mutation signal segment, specifically: Based on the theoretical model of RFID signal propagation in an ideal interference-free environment, combined with the carrier wavelength and the spatial geometric parameters of the range reader, a benchmark model of phase mutation parameters is established. Importing the mutation characteristic parameters of the phase mutation signal segment into the benchmark model: comparing the density of the mutation starting point with the benchmark model to generate a density deviation coefficient; calculating the percentage excess of the peak gradient relative to the theoretical upper limit; and calculating the proportion of abnormal mutations whose duration exceeds the theoretical range; The characteristic parameter weights are determined by the entropy weight method, and the density deviation coefficient, percentage over-limit and abnormal mutation ratio are standardized, and then weighted summation is performed to obtain the multipath interference suppression index. The multipath interference suppression index is input into a pre-trained three-branch convolutional neural network, and the multipath interference suppression effect level is output through a feature fusion layer.

8. A data collection and analysis system for an RFID reading range intelligent terminal, characterized in that: The data collection and analysis system of the RFID reading range intelligent terminal includes a memory and a processor, wherein the memory stores a data collection and analysis method program of the RFID reading range intelligent terminal. When the data collection and analysis method program of the RFID reading range intelligent terminal is executed by the processor, the steps of the data collection and analysis method of the RFID reading range intelligent terminal as claimed in any one of claims 1 to 7 are implemented.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a data collection and analysis method program for an RFID reading range intelligent terminal. When the data collection and analysis method program for an RFID reading range intelligent terminal is executed by a processor, the steps of the data collection and analysis method for an RFID reading range intelligent terminal as described in any one of claims 1 to 7 are implemented.

Images