Millimeter wave detection multipath interference suppression method and system

Abstract

The invention discloses a millimeter wave detection multipath interference suppression method and system, and particularly relates to the technical field of millimeter wave detection, and the method comprises the following steps: collecting a millimeter wave receiving signal, and executing structure analysis to extract a direction distribution feature and an inter-frame similar feature; respectively constructing a path distribution index and an evolution trend index based on the path distribution index and the evolution trend index; forming an input signal combination by two indexes, and driving rule triggering and activating state calling logic; a rule subset is dynamically generated, rule set reconstruction is executed, then the rule subset is applied to an interference judgment process, a path response decision is completed, and closed-loop suppression control is formed; direction distribution features and inter-frame similar features are extracted to construct double indexes, the double indexes serve as path state quantity to drive rule activation, dynamic reconstruction and updating of a rule set are achieved in combination with state calling logic, the updated rule set is applied to an interference judgment process, a closed-loop mechanism of path interference suppression is constructed, and the path interference suppression efficiency is improved. And cooperative control of spatial separation and time stability is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of millimeter wave detection, and more particularly to a millimeter wave detection multipath interference suppression method and system. BACKGROUND

[0002] Millimeter wave detection has been widely used in vehicle radar, security and awareness, unmanned system navigation and other fields due to its advantages of high bandwidth, short wavelength and strong penetration. In actual detection process, due to the existence of multiple reflection surfaces in the environment, multiple propagation paths are easily generated at the receiving end, some of which are caused by sidelobe reflection or diffraction paths in non-main propagation direction, forming a typical multipath interference phenomenon. Especially in non-line-of-sight environment, complex scene or dynamic interference background, the interference paths not only present the characteristics of dense spatial distribution and energy transient fluctuation, but also often accompany unstable timing behavior, which poses a significant challenge to the focused recognition of target direction.

[0003] Existing multipath interference suppression methods mostly rely on fixed thresholds, static filters or pre-trained models for judgment, which are difficult to adapt to continuous changes in channel environment. Some schemes try to introduce residual detection or adaptive filtering, but still have problems such as rule response lag, high misjudgment rate and inability to capture path evolution trend.

[0004] In view of the above shortcomings, the present application provides a millimeter wave detection multipath interference suppression method and system, which proposes a dynamic interference control framework combining "double exponential driving + rule evolution" in view of the path dispersion and timing stability characteristics hidden in millimeter wave receiving signals. By cross-sampling within a fixed time window to construct direction distribution features and inter-frame similarity features, path distribution exponent and evolution trend exponent are generated respectively, which are used as current state representation signals to drive the dynamic activation, reconstruction and elimination of rule system, thereby forming an adaptive rule library management mechanism, and finally acting on the response decision-making process at the path level, realizing stable focusing on the main path and dynamic suppression of interference paths. SUMMARY

[0005] To achieve the above purpose, the present application provides the following technical scheme: A millimeter wave detection multipath interference suppression method, comprising the following steps: Collecting millimeter wave receiving signals and performing initial structure analysis on the collected data, extracting direction distribution features reflecting path dispersion degree and inter-frame similarity features reflecting time continuity by cross-sampling of fixed time domain window and spatial beam; Constructing path distribution exponent according to direction distribution features and evolution trend exponent according to inter-frame similarity features, path distribution exponent is used to measure the spatial separation degree of propagation path, and evolution trend exponent is used to measure the stability degree of each path with time; The path distribution index and the evolution trend index are introduced into the rule-driven process as an input signal combination to trigger a rule trigger factor, and to activate state calling logic in a preset meta-rule structure according to the input signal combination; Under the guidance of the state calling logic, a rule subset meeting the current state constraints is generated from the rule set, and the boundary conditions of the retained rules, the replaced rules and the eliminated rules are determined, the automatic reconstruction of the rule set is performed, and the reconstructed rule set is input into the interference judgment process for execution to complete the time sequence response closed loop of interference suppression.

[0006] In a preferred embodiment, the step of collecting the millimeter wave receiving signal further comprises: The continuous direction scanning in the angle range is performed according to the fixed beam width, and the scanning results are synchronously buffered at fixed time intervals; then, the instantaneous power value of the scanning signal at each angle position is calculated, and the power value is compared with the average value at the same angle in the historical frame; if the current power value is greater than the average value multiplied by a set gain coefficient, the angle position is marked as a high-energy path point to generate a path initial value set; The direction distribution feature is obtained by dividing the number of high-energy path points corresponding to each angle in the path initial value set by the total number of angles, representing the high-energy path density per unit angle; the inter-frame similarity feature is obtained by calculating the cosine similarity of the matching degree of the high-energy path points at the same angle in the adjacent two frames, and the matching degree is calculated based on the path segment power vector to obtain the cosine value of the included angle, and then the matching degrees at all angles are weighted and averaged to obtain the inter-frame similarity feature value.

[0007] In a preferred embodiment, the path distribution index is constructed in the following manner: Within a fixed time window, based on the direction distribution feature, first, the number of high-energy path points in each angle direction is arranged in order of angle number to form a one-dimensional angle energy vector; then, discrete cosine transformation is performed on the angle energy vector to expand it into a group of frequency component coefficients, denoted as the first component to the last component; the lowest frequency component coefficient and the highest frequency component coefficient are taken out from the group of components, and the difference value is calculated and denoted as the first amplitude difference; the arithmetic mean value of all component coefficients is denoted as the second average value. Divide the first amplitude difference by the second average value to obtain a normalized discrete value as a spatial energy diffusion initial value. Input the initial value into a structure balancing function, specifically, the ratio of the maximum value to the second largest value in the angle energy vector is taken as a compression factor, multiplied by the normalized discrete value to form the final path distribution index.

[0008] In a preferred embodiment, the evolution trend index is constructed in the following manner: In a fixed time window, based on inter-frame similarity features, firstly, a cosine similarity calculation is performed on the path segment power vectors between each frame and the previous frame in the window, forming a similarity sequence. The frames with a similarity greater than 0.7 in the sequence are connected by directed edges, with the frames as nodes, to construct a path continuity stability graph; in the graph, a directed subgraph with the longest continuous frame connection is extracted, and the ratio of the number of frames contained in the subgraph to the total number of frames in the time window is defined as the stability ratio; the stability ratio is multiplied by the reciprocal of the maximum jump value of adjacent frames in the power change sequence to form a preliminary trend value; the preliminary trend value is input into a smoothing modulation function, and the smoothing modulation function is calculated as follows: taking the preliminary trend value as the independent variable, calculating the difference between the preliminary trend value and the sliding mean, and constructing a reciprocal modulation factor with the square of the difference plus a constant as the denominator, multiplying the preliminary trend value by the reciprocal modulation factor to obtain an evolution trend index.

[0009] In a preferred embodiment, the specific way of introducing the path distribution index and the evolution trend index as input signals into the rule-driven process includes: Firstly, a two-dimensional input signal vector is constructed, with the path distribution index and the evolution trend index as the first dimension and the second dimension signal quantities respectively, forming an input combination in each processing period; secondly, the input combination is mapped to a pre-set index grid, and the grid is formed by dividing the value range of the path distribution index and the evolution trend index, and each grid cell corresponds to a set of rule trigger factors; the matching rule trigger factors are extracted in the grid cell where the input combination is located, and the state call logic number corresponding to the trigger factors is specified in the internal mapping table to activate the logic.

[0010] In a preferred embodiment, automatic reconstruction specifically includes the following operations: Firstly, based on the state call logic number, in each processing period, the input signal fitting interval defined by all currently inactive rules is calculated, i.e. the upper and lower bounds of the path distribution index and the evolution trend index allowed by the state call logic number, forming a two-dimensional rectangular region; in the time window of the current processing period, the actual input signal combination is checked frame by frame to see if it falls within the rectangular region, and the number of frames that meet the condition is counted; if the frame value is not less than half of the total number of frames in the time window, the rule is determined as a candidate rule with input signal fitting degree up to standard, and is added to the rule set; For each rule activated in the current rule set, the barycentric angle sequence of the path response direction in each of its first five processing periods is extracted, combined with the input signal of the corresponding period to calculate the difference direction, i.e. whether the path distribution index and the evolution trend index are the same as the difference sign of the last period; if the sign direction is the same at least three times in the five periods, i.e. the difference is positive or negative, it is considered that the response direction is consistent with the change direction of the input signal; such rules will be combined into a combined rule group, and the combined logic is to merge the input signal adaptation intervals of each rule into a new input definition range in the minimum coverage mode; For all rules that are not activated by any frame in the first five periods and whose Euclidean distance with the current input signal combination exceeds the preset distance threshold, mark them as eliminated objects; Finally, the newly added rules, the combined rule group and the rules that are not eliminated are written into the rule set of the next period together, completing the automatic reconstruction process of the rule set.

[0011] In a preferred embodiment, the execution of the interference judgment process is based on the reconstructed rule set and the path state data of each path segment, which determines the path response instruction for each path segment. The path state data consists of the direction angle, power intensity, existing time frame number and inter-frame energy fluctuation amplitude of each path segment in the current period, and is compared with the path response conditions defined in each rule of the rule set; In the comparison process, if the path state data completely satisfies the activation condition of a rule, the path response instruction corresponding to the rule takes effect; if multiple rules satisfy the condition at the same time, the corresponding preset priority in the rule set determines the instruction, and the mechanism of priority response covering the later response is adopted when the instructions conflict; The path response instruction includes three types: path suppression instruction, path reservation instruction and path replacement instruction.

[0012] In a preferred embodiment, a millimeter wave detection multipath interference suppression system specifically includes: A signal acquisition and analysis unit is used to collect millimeter wave receiving signals and perform primary structure analysis in a fixed time domain window with a spatial beam cross-sampling method to extract direction distribution features representing the degree of path dispersion and inter-frame similarity features representing time continuity; A state construction unit is used to construct a path distribution index based on the direction distribution features and an evolution trend index based on the inter-frame similarity features. The path distribution index is used to measure the spatial separation degree of the propagation path, and the evolution trend index is used to measure the stability degree of each path over time; A driving trigger unit is used to combine the path distribution index and the evolution trend index to form an input signal, trigger a rule trigger factor, and activate the state calling logic in the preset meta-rule structure accordingly; A rule reconfiguration unit is configured to generate a rule subset satisfying current state constraints from a rule set under the guidance of state calling logic, and determine the boundary conditions of retained rules, replaced rules and eliminated rules to perform an automatic reconfiguration process of the rule set; An interference execution unit is configured to receive the reconfigured rule set, match the rule set with current path state data in an interference determination process, issue path response instructions according to signal-driven logic to implement path suppression, retention or replacement, and complete a time sequence response closed loop of interference suppression.

[0013] The technical effects and advantages of the present application are as follows: The present application can extract key information reflecting the dispersion characteristics and time continuity of the propagation path by structurally analyzing the collected millimeter wave received signal and combining fixed time domain window and spatial beam cross sampling. In the initial stage of path identification, the system uses time and space two-dimensional feature mining methods to obtain direction distribution characteristics and inter-frame similarity characteristics, respectively, to provide basic parameters for subsequent state construction. This process has clear separation in data dimension, so that the path set has a quantifiable and distinguishable structure at the initial stage of construction, laying a foundation for overall path behavior analysis and rule guidance mechanism.

[0014] The present application uses direction distribution characteristics to construct path distribution index and uses inter-frame similarity characteristics to construct evolution trend index to measure the spatial separation degree and time stability degree of the propagation path, respectively, and combines the two as a driving signal to participate in the dynamic activation of the rule system. The structural state of the path set represented by the two indexes captures the path evolution trend in the environment from the spatial dimension and the time dimension, respectively, to realize the coupling between state driving and behavior triggering. This structure allows the system to realize a rule scheduling mode consistent with the current path environment without directly relying on model reasoning, thereby improving the synchronization between the rule system and the external environment.

[0015] The present application constructs a mechanism for guiding rule subset selection and updating by state calling logic, reconstructs the rule set according to the path distribution index and evolution trend index under the current state, and completes the closed loop feedback of interference suppression. In the dynamic rule control structure, the state calling logic is activated to extract a subset satisfying the current state constraints from the rule set, and the boundary conditions of retention, replacement and elimination are defined to realize the continuous adaptive adjustment of the rule set. Finally, the updated rule set is applied to the interference determination process to form a complete closed loop control logic from data evaluation, state generation, rule action to result feedback, ensuring that the interference response mechanism has periodic correction and real-time adjustment capability. BRIEF DESCRIPTION OF DRAWINGS

[0016] For the convenience of those skilled in the art, the present application will be further described below with reference to the accompanying drawings; Figure 1 A schematic diagram of a millimeter wave detection multipath interference suppression method according to the present application.

[0017] Figure 2 A schematic diagram of a millimeter wave detection multipath interference suppression system according to the present application. DETAILED DESCRIPTION

[0018] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0019] Reference Signs List Figure 1 - Figure 2 The following embodiments are obtained: Embodiment 1: A millimeter wave detection multipath interference suppression method, comprising the following steps: Millimeter wave receiving signals are collected, and preliminary structural analysis is performed on the collected data. Direction distribution characteristics reflecting the degree of path dispersion and inter-frame similarity characteristics reflecting time continuity are extracted by fixed time domain window and spatial beam cross sampling. This step is the basic input stage of the entire method, which aims to sample signals in the millimeter wave frequency band from the target environment and build a time-space basis for data structure analysis. The fixed time domain window ensures the time consistency of the sampled data, and the spatial beam cross sampling establishes a multi-dimensional path data map through the receiving response of different direction angles. The direction distribution characteristics are used to describe the dense or separated state of the path in the angle, reflecting the distribution complexity of the path in space; the inter-frame similarity characteristics compare the relative consistency of the path in different time frames, depicting the continuity and stability of the path with time. The extraction process of the two characteristics establishes a comprehensive representation of the propagation path state in the time and space dimensions, which serves as the basis for subsequent modeling and driving.

[0020] The path distribution index is used to measure the spatial separation degree of the propagation path, and the evolution trend index is used to measure the stability degree of each path over time. This step converts the two types of original feature data extracted in the previous step into standardized numerical indicators, which correspond to the state measurement of the propagation path set in the angle domain and the time domain, respectively. The path distribution index is an index constructed by statistical and mathematical modeling of the direction distribution feature, which is used to measure whether the path is highly concentrated or significantly separated in the spatial direction within the current time window, thereby reflecting the complexity of multi-path propagation in the channel. The evolution trend index quantifies the consistency of the path behavior in the continuous time period according to the inter-frame similarity feature, which is used to determine whether the path is stable or quickly fluctuates and disappears. These two indexes express the structural state of the overall propagation environment in a unified numerical form, which is the direct driving signal of the subsequent rule triggering mechanism.

[0021] The path distribution index and the evolution trend index are combined as input signals and introduced into the rule-driven process to trigger the rule triggering factor, and the state call logic in the preset meta-rule structure is activated according to the input signal combination. This step combines the two indexes into a two-dimensional input signal, which is used as the state representation entry of the intelligent decision mechanism. The rule-driven process is composed of pre-designed meta-rule structures, and each meta-rule corresponds to a certain interval of the input signal combination through the triggering factor. When the input signal combination enters this interval, the corresponding state call logic is activated. The state call logic defines the allowed or prohibited rule behavior in the current state, including which rules can participate in the calculation, which rules need to be suspended, and how to perform logical migration. Through this mechanism, the system does not directly make a final interference judgment based on the index, but rather drives the operation of the rule set in a way that realizes the dynamic management of the rule set ecology.

[0022] Under the state calling logic guidance, a rule subset satisfying the current state constraints is generated from the rule set, and the boundary conditions of the retained rules, the replaced rules and the eliminated rules are determined, the automatic reconstruction of the rule set is performed, the reconstructed rule set is input to the interference judgment process for execution, and the timing response closed loop of the interference suppression is completed. This step realizes the self-organization and evolution of the rule set, and ensures that the interference suppression mechanism has adaptability and updating ability. The state calling logic first limits the range of rules allowed to be active, and on this basis, the rule set is screened to form a rule subset that is legal and effective under the current state. By analyzing the rule historical response data and the input signal matching situation, it is determined which rules should be retained, which should be replaced with other rules, and which have been eliminated, thereby forming clear boundary conditions for structure updating. After the automatic reconstruction process is performed, a new set of rules with optimized structure is obtained, and it is passed to the interference judgment process. Finally, the path retention, replacement or suppression is realized through the path response instructions based on the rules, so that the system continuously updates the response strategy in multiple processing cycles to form a closed loop time continuous interference suppression logic.

[0023] The step of collecting the millimeter wave received signal further comprises: performing continuous direction scanning in an angle range according to a fixed beam width, and performing synchronous caching of the scanning results at a fixed time interval; in this step, the fixed beam width refers to an antenna array or beam control device in the receiving device used to form a directional received signal, which is set to scan within a certain angle coverage range, and the angle step of each scan is consistent. For example, the beam width can be set to 5 degrees, and the scanning range is 0° to 180°, then 36 scanning directions are generated in this range. Continuous direction scanning refers to sequentially performing energy reception and forming an angle response sample for each direction according to the set angle sequence; the fixed time interval refers to the time interval of sampling each frame of signal, for example, a complete scan is completed every 10 milliseconds; synchronous caching refers to uniformly saving the receiving results of multiple directions in the same time frame to form a frame-level scanning snapshot for subsequent analysis. Subsequently, the instantaneous power value of the scanned signal at each angle position is calculated, and the power value is compared with the average value of the historical frame at the same angle, if the current power value is greater than the average value multiplied by a set gain coefficient, the angle position is marked as a high-energy path point, and a path initial value set is generated; the instantaneous power value refers to the received signal power measured at a specific angle in the current frame, which can usually be obtained by signal amplitude square processing; the average value of the historical frame at the same angle is the sliding average of the instantaneous power value of the current angle in the past several frames, reflecting the background power level; the gain coefficient is a preset adjustment parameter for determining whether the current signal is a significant energy event, and the typical value range is, for example, 1.5-3.0; when the instantaneous power value of the current frame at a certain angle exceeds the historical average value multiplied by the gain coefficient, it is considered that there is significant signal activity in the angle direction at the current time, and it is determined as a high-energy path point; the angle number that meets the above conditions is recorded to form a path initial value set, that is, a direction set of all suspected existing effective reflection or direct path in the current frame.

[0024] The direction distribution feature is obtained by dividing the number of high-energy path points corresponding to each angle in the path initial value set by the total number of angles, representing the high-energy path density per unit angle; in each time window, the system performs statistics on the path initial value sets recorded in multiple continuous frames; for each angle position, the number of frames in which it is marked as a high-energy path point in the time window is calculated, which is the number of high-energy path points for the angle; the statistical value of each angle is divided by the total number of scanning angles (for example, the aforementioned 36 directions), and the path density distribution per unit angle is obtained; the unit angle density is summarized to form the direction distribution feature describing the spatial path distribution characteristics, which is used for subsequent construction of the path distribution index.

[0025] The inter-frame similarity feature is obtained by calculating the cosine similarity of the matching degree of the high-energy path points in the same angle in two adjacent frames. The matching degree is calculated based on the angle cosine value of the path segment power vector, and the inter-frame similarity feature value is obtained by weighted average of the matching degrees in all angles. In each two adjacent time frames, the system forms a path segment power vector for the instantaneous power value in each angle direction, i.e., a time pair vector is formed at each angle position, a two-dimensional vector is formed by the two power values at each angle position, and the angle cosine value between the two vectors, i.e., the cosine similarity, is calculated. The set of cosine values in all directions represents the path similarity distribution between the frame pair.

[0026] The path distribution index is constructed as follows: in a fixed time window, based on the direction distribution feature, first, the number of high-energy path points in each angle direction is arranged in a one-dimensional angle energy vector in order of angle number; the fixed time window refers to a time period containing a number of consecutive processing frames, such as ten frames or twenty frames, used to form a period of continuous statistical analysis; the direction distribution feature has been defined in the foregoing, i.e., the statistical density of high-energy path points in a unit angle, used to reflect the distribution of paths in the spatial angle domain; in the time window, for each angle number, the number of times it is identified as a high-energy path point in the window is counted; in ascending order of angle number, the statistical values corresponding to each angle are arranged in order to form an angle energy vector, which reflects the energy distribution changes in the entire spatial scanning angle range. Then, a discrete cosine transform is performed on the angle energy vector to expand it into a set of frequency component coefficients, denoted as the first component to the last component; the discrete cosine transform is a method for converting discrete data in time or space domain into frequency domain representation, commonly used to extract signal change trends and structural features; after the discrete cosine transform is performed on the angle energy vector, a number of frequency components are obtained, each component representing the change intensity of the original vector at a certain frequency; the first component represents the lowest frequency component, which usually represents the overall energy average trend; the last component represents the highest frequency component, which reflects the rapid change feature. The difference between the lowest frequency component coefficient and the highest frequency component coefficient is calculated and denoted as the first amplitude difference; the values of the lowest frequency component and the highest frequency component in the transformation result are selected to represent the behavior of the spatial energy at the two extreme scales of overall distribution and local fluctuation, respectively; by calculating the difference between the two, a representative quantity of the change amplitude of the path distribution in the direction domain is obtained, which is called the first amplitude difference. The arithmetic mean of all component coefficients is denoted as the second average value; the arithmetic mean of all frequency component coefficients obtained by the discrete cosine transform is calculated, i.e., the overall intensity level of all frequency behaviors, which is called the second average value.

[0027] The first amplitude difference is divided by the second average value to obtain a normalized discrete value as a spatial energy diffusion initial value; the first amplitude difference is divided by the second average value to form a normalized ratio, which eliminates the influence of different path quantities and absolute energy values; the ratio is used to measure the diffusion degree of the path in the direction domain, and the larger the value, the higher the direction change amplitude and the more obvious the spatial discreteness; the value is defined as the spatial energy diffusion initial value, which is the basis for subsequent calculation of the path distribution index. The initial value is input into the structure balance function, specifically, the ratio of the maximum value to the second maximum value in the angle energy vector is taken as a compression factor, which is multiplied by the normalized discrete value to form the final path distribution index. The structure balance function is used to adjust the influence degree of the path in the local spatial set; first, find the number of high-energy path points with the highest frequency in the angle energy vector, that is, the maximum value and the second maximum value; calculate the ratio between the two as a compression factor, which is used to represent whether there is a concentration bias in the path distribution; if the ratio is close to one, it indicates that the path is relatively uniform in multiple directions; if the ratio is very large, it indicates that the path is mainly concentrated in a small number of directions; multiply the compression factor by the spatial energy diffusion initial value described above to obtain the final path distribution index, which is usually limited to a range of zero to one, as a unified index to measure the degree of spatial dispersion.

[0028] The evolutionary trend index is constructed in the following way: in a fixed time window, based on the inter-frame similarity feature, first, the cosine similarity calculation is performed on the path segment power vectors between each frame and the previous frame in the window to form a similarity sequence. The fixed time window refers to an interval composed of a plurality of continuous time frames, which is used to perform trend statistics and behavior judgment; the inter-frame similarity feature has been defined in the foregoing, which is used to measure whether the path structure between adjacent time frames remains stable; the path segment power vector refers to a multi-dimensional vector composed of the received power value corresponding to each direction in each time frame; the path segment power vectors of each two adjacent time frames in the window are extracted, and the cosine similarity calculation is performed on the two vectors to measure the continuation degree of the path form in the time dimension; the cosine similarity values between all frame pairs are arranged in time sequence to form a numerical sequence, i.e., the similarity sequence. The frame pairs with a similarity greater than 0.7 in the sequence are connected by directed edges, the frames are regarded as nodes, and a path continuous stability graph is constructed; the frame pairs with a similarity greater than a set threshold in the similarity sequence are extracted, the threshold is set to 0.7, which means that the path change is less than 30 degrees; each frame is regarded as a node in the graph, and when the similarity between the current frame and the previous frame exceeds the threshold, a directed edge from the previous frame to the current frame is constructed. Through the edges of all frame pairs meeting the condition, a graph structure representing the stability of the path continuation over time is formed, which is called a path continuous stability graph. In the graph, a directed subgraph with the longest continuous frame connection is extracted, and the ratio of the number of frames contained in the subgraph to the total number of frames in the time window is defined as the stability ratio; in the path continuous stability graph, the directed edge sequence with the longest uninterrupted connection is regarded as a stable propagation path; the number of frames contained in the longest directed subgraph represents the length of the continuous structure stable time period in the system; the ratio calculation of this frame number to the total number of frames in the time window obtains the stability ratio, which reflects the degree of path continuity over time.

[0029] The stable proportion is multiplied by the reciprocal of the maximum jump value of adjacent frames in the power change sequence to form a preliminary trend value; the power change sequence refers to a sequence formed by the difference between the received power values in each angle direction between two adjacent frames; in this sequence, the maximum jump amplitude, i.e., the maximum power change value, is found; the reciprocal of the maximum value, i.e., one divided by the jump value, is taken as an adjustment factor; the stable proportion is multiplied by the adjustment factor to obtain the preliminary trend value, which is used to jointly reflect the path time continuity and energy stability. The preliminary trend value is input into a smoothing modulation function, and the smoothing modulation function is calculated in the following manner: the difference between the preliminary trend value and the sliding mean value is calculated, a constant is added to the square of the difference value to construct a reciprocal modulation factor, and the preliminary trend value is multiplied by the reciprocal modulation factor to obtain an evolution trend index. The smoothing modulation function is used to suppress the interference of incidental fluctuations on the trend value; first, the sliding mean value of the preliminary trend values in multiple continuous time windows is calculated to form a reference benchmark of the trend history; then, the difference between the current preliminary trend value and the sliding mean value is obtained; the square of the difference value is added to a fixed positive constant to be used as the denominator; a fixed positive number is used as the numerator to construct a reciprocal modulation factor; finally, the modulation factor is multiplied by the current preliminary trend value to obtain the evolution trend index; the index is normalized and limited between zero and one to represent the time stability degree of the entire path set in the current time window.

[0030] The specific way of introducing the path distribution index and the evolution trend index as input signals into the rule-driven process includes: firstly, constructing a two-dimensional input signal vector, the path distribution index and the evolution trend index being respectively the first dimension and the second dimension signal amount, forming an input combination in each processing period; the path distribution index is used to reflect the distribution discreteness of the propagation path in the spatial angle in the current time window; the evolution trend index is used to reflect the structural stability degree of the path set in the time dimension; the processing period refers to the time period for the system to complete once rule set determination and interference suppression response, for example, once every one hundred frames; in each processing period, the path distribution index and the evolution trend index calculated in the current period are extracted, and the two are used as a group of signal values; the signal amount pair constitutes an input combination, representing the structural state of the current propagation environment; each input combination is regarded as a pair of two-dimensional numerical input, as the entrance parameter for driving the rule system to work. Secondly, the input combination is mapped to a preset index grid, and the grid is formed by dividing the value range of the path distribution index and the evolution trend index, and each grid cell corresponds to a set of rule trigger factor set; the index grid is to divide the value range of the path distribution index and the evolution trend index into several fixed intervals to form a two-dimensional grid structure; for example, the value range of the path distribution index is divided from zero to one into ten equal interval sections, and the evolution trend index is also divided in the same way; the two-dimensional grid forms one hundred grid cells, and each grid cell represents a specific state interval; each grid cell is bound to a set of rule trigger factor set in advance, that is, a list of potential rules that can be activated in the state interval; the binding relationship is usually established through a rule configuration file, or generated by a rule learning mechanism during system initialization.

[0031] The matching rule trigger factor is extracted in the grid cell where the input combination is located, and the state call logic number to be activated is specified in the internal mapping table of the corresponding trigger factor. The input combination in the current processing period is compared with the two-dimensional index grid to locate the specific grid cell to which it belongs; after entering the cell, the rule trigger factor set bound thereto is extracted as the response entrance that can be triggered in the current state; each rule trigger factor contains a mapping table, which defines the state call logic number pointed to by the trigger factor when activated; the state call logic number is a control instruction index used to activate a certain rule action mode inside the system, which is used to instruct the system how to select the rule set, reconstruct the rule structure, and allocate the priority in the current state; For example, if the current path distribution index is 0.65 and the evolution trend index is 0.38, it corresponds to the grid cell in the seventh row and the fourth column; the cell corresponds to a rule trigger factor whose mapping table points to the state logic with a state call logic number of three.

[0032] The automatic reconstruction specifically includes the following operations: first, based on the state calling logic number, in each processing period, calculate the defined input signal fitting interval of all current inactive rules, that is, the upper and lower limits of the respective allowed values of the path distribution index and the evolution trend index corresponding to the state calling logic number, to form a two-dimensional rectangular region; each rule is preset with a set of input signal fitting intervals according to its corresponding state calling logic number when it is generated, that is, the minimum and maximum values that can be accepted in the path distribution index and the evolution trend index two dimensions, to form a rectangular boundary range, which is used to judge whether the current input matches the rule. For example, the input signal fitting interval of a rule is defined as the path distribution index between 0.5 and 0.8, and the evolution trend index between 0.3 and 0.6, and the interval is a two-dimensional rectangular region.

[0033] In the time window of the current processing period, check whether the actual input signal combination falls into the rectangular region frame by frame, count the number of frames that meet the conditions, and if the frame value is not less than half of the total number of frames in the time window, the rule is determined as a candidate rule with input signal fitting degree meeting the standard, and is added to the rule set; the time window refers to the set of consecutive frames contained in the current processing period, for example, set to ten frames, indicating that the input signal combination is received ten times in the period; for the input signal combination of each frame, the path distribution index and evolution trend index values are taken out respectively to judge whether they fall into the two-dimensional rectangular interval corresponding to the rule; if the number of frames that meet the conditions is greater than or equal to half of the number of frames in the time window, for example, more than five frames, it is considered that the rule has sufficient input matching ability in the current period; such rules are determined as candidate rules with input signal fitting degree meeting the standard, and are added to the rule set in this round for participating in subsequent rule screening and interference response.

[0034] For each rule in the current rule set that has been activated, in the first five processing periods, the sequence of the center angle of the path response direction in each period is extracted, and the difference direction is calculated with the input signal combination corresponding to the period, that is, whether the difference values of the path distribution index and the evolution trend index are the same as the last period; the center angle of the path response direction refers to the direction angle position with the most concentrated energy distribution in the response path of the rule in each period, which is used to represent the main directionality of the path; in the next five processing periods, the sequence of the path center angle generated by each rule in each period is extracted, and the difference values of the path distribution index and the evolution trend index corresponding to the period are calculated respectively; the path distribution index of the current period is subtracted from the value of the last period to obtain the numerical direction, and then it is judged whether the sign is positive or negative, and the evolution trend index is the same; if both indices show the same sign difference in the same period, for example, both positive or both negative, it is considered that the difference direction is consistent.

[0035] If the sign directions are the same at least three times in five cycles, that is, the difference values are all positive or all negative, it is considered that the response direction of the rule is consistent with the change direction of the input signal. The number of times that the difference value directions are consistent is counted in five cycles. If the path distribution index and the evolution trend index are positive or negative at the same time for three times or more, it is indicated that the response direction of the rule is consistent with the change trend of the input signal. Such a rule is identified as having an evolutionary coupling relationship with the input behavior. Such rules will be combined and merged into a combined rule group. The combination logic is to merge the input signal adaptation intervals of the rules into a new input definition range in a minimum coverage manner. A combined rule group is formed by the above rules that meet the direction consistency condition. The minimum coverage manner means that the minimum lower bound and the maximum upper bound of the path distribution index and the minimum lower bound and the maximum upper bound of the evolution trend index are extracted for the input signal adaptation intervals of all rules, respectively, to form a new rectangular range. This range includes all the input coverage intervals of the original rules, but removes the overlapping definitions, simplifying the rule redundancy. The newly generated combined rule group will use this coverage range as the new input definition, replacing the original multiple redundant rules, reducing system complexity and improving response efficiency.

[0036] All rules that are not activated by any frame in the first five cycles and whose input signal adaptation interval has a Euclidean distance from the current input signal combination exceeding a preset distance threshold are marked as eliminated objects. If a rule has never been activated in the last five processing cycles, that is, it has never met the input adaptation interval requirement in the input matching stage, it indicates that its response ability is low. In addition, if the average Euclidean distance between its input signal adaptation interval and the actual input signal combination in the current cycle is greater than the threshold set by the system, for example, greater than zero point three, it means that the preset state of the rule is too different from the current system state. The rule that meets the above two conditions is considered as a redundant or invalid rule, and is marked for elimination and removed from the subsequent rule set. Finally, the new rules, the combined rule groups and the rules that are not eliminated are written into the rule set of the next cycle together, completing the automatic reconstruction process of the rule set.

[0037] After completing the candidate selection, combination merging and invalid elimination of the rule set in the current cycle, a new rule set is formed. The set includes: new rules that meet the input adaptation degree, combined rule groups that are merged according to the direction response consistency, and rules that are still active or state-related and are not eliminated in the current cycle. The new rule set will be the basis for the execution of the interference judgment process in the next cycle, realizing the self-updating and evolution mechanism of the system rule library, and thus completing the automatic reconstruction process of the rule set in the processing cycle.

[0038] Assume that the current system is in the sixth processing cycle, the current time window of the system contains ten frames of data, and fifteen rules have been accumulated in the past five processing cycles, numbered from one to fifteen. According to the input signal combination of the path distribution index and the evolution trend index collected in the current processing cycle, the automatic reconstruction operation of the rule set is performed, and the process is as follows: I. Candidate screening: In the current cycle, the system scans the rules that are not activated, i.e., rules numbered nine to fifteen, a total of seven rules. The system reads the state call logic number bound to each rule to obtain the input signal fitting interval defined by the rule. Taking rule number twelve as an example, its path distribution index interval is zero point four to zero point seven, and its evolution trend index interval is zero point three five to zero point six five. The system traverses the current ten frames of input signal combination and finds that the path distribution index and evolution trend index values of six frames fall into the two-dimensional rectangular fitting interval of rule number twelve at the same time, more than half of the total number of frames, so rule number twelve is marked as a candidate rule that meets the input signal fitting degree requirement and is ready to be added to the rule set. After screening by the same process, a total of three rules, rules ten, twelve and fourteen, meet the conditions and are added as new candidate rules.

[0039] II. Combination merging: The system continues to process the currently activated rules, i.e., rules one to eight, a total of eight rules. For these eight rules, the system extracts the path response direction barycenter angle sequence in the past five processing cycles and the path distribution index and evolution trend index of each cycle. Taking rule number three as an example, in the past five cycles, the difference value direction of its path distribution index and the previous cycle is positive, negative, positive, positive, and negative, and the difference value direction of its evolution trend index is positive, negative, positive, positive, and negative. Statistics show that there are three times when the sign directions of the two are consistent, meeting the combination rule condition. The system further finds that rules three, five and six all meet this condition in the past five cycles, so they are combined into a combined rule group. The input signal fitting interval of the combined rule group is: the path distribution index is combined from the minimum lower bound and the maximum upper bound of the intervals of the three to form zero point three eight to zero point six nine, and the evolution trend index is combined from the intervals of the three to zero point two five to zero point five eight. After combination, the original three rules are invalid, and the combined rule group replaces the new rule set.

[0040] III. Invalid elimination: The system checks rules numbered seven, nine and thirteen that have not been activated in the past five cycles. Further calculation of the Euclidean distance between the input signal fitting interval of these rules and the average input signal combination of the current cycle shows that the distance is greater than the set threshold of zero point three five. Therefore, rules seven, nine and thirteen are marked as redundant rules by the system and removed from the rule set.

[0041] Four, new rule set generation: after the above three steps, the system constructs a new rule set for the next period, including: new candidate rules: No. 10, No. 12, No. 14; Combined rule group: merged rule composed of No. 3, No. 5, No. 6; Active rules retained: No. 1, No. 2, No. 4, No. 8; Eliminated rules: No. 7, No. 9, No. 13. The final new rule set contains eight effective rules, and enters the rule-driven and interference judgment process of the seventh processing period.

[0042] The execution of the interference judgment process is based on the reconstructed rule set and the path state data of the current path segment, and determines the path response instruction issued to each path segment. The interference judgment process is an operation process executed after the rule set is updated, aiming to determine whether to apply interference control action to a certain path according to the actual state of each propagation path in the current frame, combined with the conditions defined in the rule set. The reconstructed rule set is the new rule set generated by candidate screening, combination merging and invalid elimination in the last processing step, which is the only rule basis for matching judgment in the current period. Path segment refers to a millimeter wave propagation path that can be identified in the spatial direction and time continuity dimension; the system judges each path segment one by one to determine whether to apply a response.

[0043] The path state data is composed of the direction angle, power intensity, existing time frame number and inter-frame energy fluctuation amplitude of each path segment in the current period, and is compared with the path response conditions defined in each rule in the rule set. Path state data is the feature information about path segments collected and structured by the system in the current processing period, including the following four parts: one is the direction angle, which is the angle position of the main energy of the path segment in the current period, used to determine whether it belongs to the target propagation direction or the offset direction; two is the power intensity, which is the weighted average or maximum value of the received power value of all frames in the current period, which measures whether the path energy exceeds the interference threshold; three is the existing time frame number, which is the number of frames that the path segment appears continuously in the current period time window, used to determine whether the path has the characteristic of persistence; four is the inter-frame energy fluctuation amplitude, which is the standard deviation or maximum change of the path power between consecutive frames, used to identify whether the path has the characteristic of rapid fluctuation.

[0044] The system matches the above data of each path segment with the predefined path response conditions in each rule in the rule set one by one, and performs condition comparison. In the comparison process, if the path state data completely meets the activation condition of a rule, the path response instruction corresponding to the rule takes effect; each rule defines a set of path response conditions, including the tolerance range of the direction angle, the lower limit or upper limit of the power intensity, the minimum frame threshold of the duration, the determination limit of the fluctuation amplitude, etc.; when the state data of a path falls within the activation range defined by the rule in all the above fields, it means that the path meets the execution condition of the rule; in this case, the rule is determined to be activated by the system, and the path response instruction bound to it takes effect immediately as the control action for the path segment.

[0045] If multiple rules meet the conditions at the same time, the corresponding preset priority in the rule set determines that the mechanism of priority response covering the later response is adopted in the case of instruction conflict; in actual operation, it may occur that the same path segment meets the activation conditions of multiple rules at the same time; to avoid response conflict, the system presets an execution priority for each rule in the rule set, which is usually a numerical weight or a rule level; if multiple rules are activated at the same time, the response instruction corresponding to the rule with the highest priority will be executed; if the priorities of multiple activated rules are the same, the mechanism of priority response covering the later response is adopted, that is, the priority response covering the later response mechanism is adopted to ensure response consistency and predictability.

[0046] The path response instruction includes three types: path suppression instruction, path reservation instruction and path replacement instruction. The path suppression instruction is used for path segments identified as interference sources or sidelobe reflections, and performs direction suppression, gain down-regulation or data shielding operation, with the purpose of reducing the interference on the system judgment result; the path reservation instruction is used for the case of identifying the target propagation path or the trusted path segment, and keeping its pass, data and weight unchanged in the subsequent signal processing process; the path replacement instruction is applicable to the case where there are multiple candidate paths but only the optimal path needs to be reserved, for example, replacing the main path with the secondary path when the energy is similar, and specific operations are performed according to the rule set.

[0047] Implementation example: in the current processing period, the system identifies three path segments: the direction angle of the first path segment is thirty degrees, the power intensity is medium-high, the existence time is nine frames, and the inter-frame energy fluctuation amplitude is small; the direction angle of the second path segment is seventy degrees, the power intensity is high, the existence time is only two frames, and the fluctuation amplitude is large; the direction angle of the third path segment is close to the first one, but the power intensity is low, the existence time is seven frames, and the fluctuation amplitude is moderate.

[0048] The rule set has the following rules: the first rule sets the direction between twenty and forty degrees, the power intensity exceeds the median, the existence time is not less than eight frames, and the fluctuation amplitude is low, and the binding path reservation instruction is reserved; the second rule sets the direction between sixty and eighty degrees, the power is higher than the threshold, the existence time is less than three frames, and the fluctuation amplitude is high, and the binding path inhibition instruction is inhibited; the third rule is almost the same as the first rule, but the priority is lower, and the binding path replacement instruction is replaced. After the system performs the comparison: the first path segment completely satisfies the first rule, and the path reservation instruction is activated; the second path segment satisfies the second rule, and the path inhibition instruction is activated; the third path segment satisfies the first and third rules at the same time, but the first rule has higher priority, so the path reservation instruction is executed, replacing the path replacement instruction. Finally, the system applies the reservation response to the path segments one and three, and applies the inhibition response to the path segment two, completing the interference judgment process of the current period.

[0049] Embodiment 2: A millimeter wave detection multipath interference suppression system, specifically comprising: The signal acquisition and analysis unit is used for collecting millimeter wave receiving signals, and performing primary structure analysis in a fixed time domain window in a spatial beam cross-sampling manner to extract direction distribution features representing path dispersion degree and inter-frame similarity features representing time continuity; The state construction unit is used for constructing a path distribution index according to the direction distribution features and constructing an evolution trend index according to the inter-frame similarity features, the path distribution index is used for measuring the spatial separation degree of the propagation path, and the evolution trend index is used for measuring the stability degree of each path over time; The driving trigger unit is used for combining the path distribution index and the evolution trend index to form an input signal, triggering a rule trigger factor, and activating state calling logic in a preset meta-rule structure according to the rule trigger factor; The rule reconstruction unit is used for generating a rule subset that satisfies the current state constraint from the rule set under the guidance of the state calling logic, and determining the boundary conditions of the reservation rule, the replacement rule and the elimination rule according to the rule subset, and executing the automatic reconstruction process of the rule set; The interference execution unit is used for receiving the reconstructed rule set, matching the current path state data in the interference judgment process, issuing a path response instruction according to the signal driving logic, realizing path inhibition, reservation or replacement, and completing the time sequence response closed loop of interference suppression. The five functional units are sequentially connected according to the method steps to form an interference suppression processing link.

[0050] The above formulas are dimensionless values, and the formulas are obtained by software simulation of a large amount of data to obtain a formula closest to the actual situation. The preset parameters in the formula are set by a person skilled in the art according to the actual situation.

[0051] It should be understood that the magnitude of the sequence of the above processes does not mean the order of execution, the execution order of the processes should be determined according to its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0052] Those skilled in the art can realize that the units and algorithm steps of the examples described in combination with the embodiments disclosed herein can be realized in electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software depends on the specific application and design constraints of the technical solution. The skilled person can use different methods to realize the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0053] Those skilled in the art can clearly understand that, for the convenience and brevity of the description, the specific working processes of the above-described system, device and unit can refer to the corresponding processes in the foregoing method embodiments, which will not be described here.

[0054] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited to this. Any person skilled in the art can easily think of changes or replacements within the technical scope disclosed by the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for suppressing multipath interference in millimeter-wave detection, characterized in that, Includes the following steps: Millimeter-wave received signals are acquired, and initial structural analysis is performed on the acquired data. By cross-sampling with a fixed time-domain window and spatial beam, directional distribution features reflecting the path dispersion degree and inter-frame similarity features reflecting time continuity are extracted. A path distribution index is constructed based on directional distribution characteristics, and an evolutionary trend index is constructed based on inter-frame similarity characteristics. The path distribution index is used to measure the spatial separation of propagation paths, and the evolutionary trend index is used to measure the stability of each path over time. The path distribution index and evolution trend index are combined as input signals and introduced into the rule-driven process to trigger the rule triggering factor, and the state call logic in the preset meta-rule structure is activated according to the combination of input signals. Guided by the state invocation logic, a subset of rules that satisfy the current state constraints is generated from the rule set, and the boundary conditions for retaining, replacing, and eliminating rules are determined. Automatic reconstruction of the rule set is then performed, and the reconstructed rule set is input into the interference judgment process for execution, thus completing the time-series response closed loop for interference suppression.

2. The millimeter-wave detection multipath interference suppression method according to claim 1, characterized in that, The steps for acquiring millimeter-wave received signals further include: A continuous directional scan is performed within an angle range with a fixed beamwidth, and the scan results are synchronously buffered at fixed time intervals. Then, the instantaneous power value of the scan signal at each angular position is calculated and compared with the average value of historical frames at the same angle. If the current power value is greater than the average value multiplied by a set gain coefficient, the angular position is marked as a high-energy path point, and a set of initial path values ​​is generated.

3. The millimeter-wave detection multipath interference suppression method according to claim 2, characterized in that, The path distribution index is constructed as follows: Within a fixed time window, based on the directional distribution characteristics, the number of high-energy path points under each angular direction is first arranged in the order of angular numbering to form a one-dimensional angular energy vector; then, a discrete cosine transform is performed on the angular energy vector to expand it into a set of frequency component coefficients, denoted as the first component to the last component; the lowest frequency component coefficient and the highest frequency component coefficient are taken from this set of components, and their difference is calculated and denoted as the first amplitude difference. The arithmetic mean of all component coefficients is then recorded as the second mean. Divide the first amplitude difference by the second average value to obtain the normalized discrete value, which is used as the initial value for spatial energy diffusion. The initial value is input into the structural equilibrium function, specifically by using the ratio of the maximum to the second largest value in the angular energy vector as a compression factor, multiplied by the normalized discrete value to form the final path distribution index.

4. The millimeter-wave detection multipath interference suppression method according to claim 3, characterized in that, The evolutionary trend index is constructed as follows: Within a fixed time window, based on inter-frame similarity features, cosine similarity calculation is first performed on the path segment power vector between each frame and the previous frame in the window to form a similarity sequence. Establish directed edge connections between frame pairs greater than 0.7 in the sequence, and construct a path-continuous stable graph with frames as nodes. In this graph, extract the directed subgraph with the longest continuous frame connection, and define the stability ratio as the ratio of the number of frames in the subgraph to the total number of frames in the time window. The stable ratio is multiplied by the reciprocal of the maximum jump value of adjacent frames in the power change sequence to form a preliminary trend value. This preliminary trend value is input into the smoothing modulation function. The smoothing modulation function is calculated as follows: the preliminary trend value is used as the independent variable, the difference between it and the moving average is calculated, and the square of the difference plus a constant is used as the denominator to construct an inverse proportional modulation factor, which is then multiplied by the preliminary trend value to obtain the evolution trend index.

5. The millimeter-wave detection multipath interference suppression method according to claim 4, characterized in that, Specific methods for incorporating path distribution index and evolution trend index as input signals into rule-driven processes include: First, a two-dimensional input signal vector is constructed, with the path distribution index and evolution trend index serving as the first and second dimensions of the signal, respectively, forming an input combination in each processing cycle. Second, this input combination is mapped to a preset index grid, which is formed by dividing the numerical ranges of the path distribution index and evolution trend index. Each grid cell corresponds to a set of rule triggering factors. The matching rule triggering factors are extracted from the grid cell where the input combination is located, and the corresponding triggering factor specifies the state call logic number to be activated in its internal mapping table.

6. The millimeter-wave detection multipath interference suppression method according to claim 5, characterized in that, Automatic refactoring specifically includes the following operations: First, based on the state call logic number, within each processing cycle, the input signal adaptation interval defined by all currently inactive rules is calculated, that is, the upper and lower bounds of the allowed values ​​of the path distribution index and evolution trend index corresponding to the state call logic number, forming a two-dimensional rectangular area; within the time window of the current processing cycle, it is checked frame by frame whether the actual input signal combination falls into the rectangular area at the same time, and the number of frames that meet the conditions is counted. If the value of the frame is not less than half of the total number of frames in the time window, the rule is determined to be a candidate rule with qualified input signal adaptation and added to the rule set; For each activated rule in the current rule set, extract the centroid angle sequence of the path response direction in each of the previous five processing cycles, and combine it with the input signal of the corresponding cycle to calculate the difference direction, i.e., whether the difference sign of the path distribution index and evolution trend index is the same as that of the previous cycle; if the sign direction is the same at least three times in the five cycles, i.e., the difference is all positive or all negative, then it is considered that its response direction is consistent with the change direction of the input signal; such rules will be combined and merged into a combined rule group, and the combination logic is to merge their respective input signal adaptation intervals into a new input definition range in the least coverage manner; For all rules that have not been activated by any frame in the first five cycles and whose Euclidean distance between their input signal adaptation range and the current input signal combination exceeds a preset distance threshold, mark them as elimination objects. Finally, the newly added rules, combined rule groups, and rules that have not been eliminated are written into the rule set for the next cycle, completing an automatic reconstruction process of the rule set.

7. The millimeter-wave detection multipath interference suppression method according to claim 6, characterized in that, The execution of the interference determination process is based on the reconstructed rule set and the current path status data of each path segment. It determines the path response command to be issued to each path segment. The path status data consists of the direction angle, power intensity, existence time in frames and inter-frame energy fluctuation amplitude of each path segment in the current period, and is compared with the path response conditions defined by each rule in the rule set. During the comparison process, if the path status data fully meets the activation conditions of a certain rule, the path response command corresponding to that rule will take effect; if multiple rules meet the conditions at the same time, the corresponding preset priority in the rule set will determine the response, and in case of command conflict, the mechanism of prior response overriding the later response will be adopted. Path response instructions include three types: path suppression instructions, path preservation instructions, and path replacement instructions.

8. A millimeter-wave detection multipath interference suppression system, based on the millimeter-wave detection multipath interference suppression method according to any one of claims 1-7, characterized in that, Specifically, it includes: The acquisition and analysis unit is used to acquire millimeter-wave received signals and perform initial structural analysis in a spatial beam cross-sampling manner under a fixed time-domain window to extract directional distribution features representing the degree of path dispersion and inter-frame similarity features representing temporal continuity. The state construction unit is used to construct the path distribution index based on the directional distribution characteristics and the evolution trend index based on the inter-frame similarity characteristics. The path distribution index is used to measure the spatial separation of the propagation path, and the evolution trend index is used to measure the stability of each path over time. The driving triggering unit is used to combine the path distribution index and the evolution trend index to form an input signal, trigger the rule triggering factor, and activate the state call logic in the preset meta-rule structure accordingly. The rule reconstruction unit is used to generate a subset of rules that satisfy the current state constraints from the rule set under the guidance of the state invocation logic, and to determine the boundary conditions of retaining rules, replacing rules and eliminating rules accordingly, and to execute the automatic reconstruction process of the rule set. The interference execution unit is used to receive the reconstructed rule set, match it with the current path state data in the interference judgment process, issue path response instructions according to the signal-driven logic, realize path suppression, retention or replacement, and complete the timing response closed loop of interference suppression.

Images