5g network speed detection system based on massive antenna technology

Abstract

The present application belongs to the technical field of wireless communication network detection oriented to large-scale antenna beam state, and specifically relates to a 5G network speed detection system based on large-scale antenna technology. The system comprises a detection data synchronous acquisition module, a beam state decoupling module and a speed correction output module. Throughput rate sampling sequences, beam number change sequences, beam reference signal quality, space division multiplexing layers, scheduling resource occupation proportion and wireless link retransmission information are collected within a detection time window. Stable beam segments, switching transition segments and beam mismatch segments are divided according to the beam number change and the reference signal quality, and the throughput rate sampling values are weighted, screened and corrected in combination with the segment type, scheduling continuity and retransmission density, so as to output the beam state decoupled network speed detection result and the abnormal reason label. The present application can reduce the interference of beam switching and beam mismatch on the detection values.

Description

Technical Field

[0001] This invention belongs to the field of wireless communication network detection technology for massive MIMO antenna beam status, specifically a 5G network speed detection system based on massive MIMO antenna technology. Background Technology

[0002] In 5G networks, massive MIMO antennas typically form directional beams through multi-element collaboration, and complete wireless access and data transmission through beamforming, beam scanning, beam switching, and spatial multiplexing. Existing network speed testing systems generally establish upload or download speed test streams between the terminal and the testing server, and statistically analyze the number of bytes transmitted, transmission latency, packet loss, and retransmission within a set testing time window to calculate the uplink or downlink rate. For the wireless side status, conventional solutions usually only record basic parameters such as serving cell, signal strength, reference signal quality, and channel quality indication, and use these parameters as supplementary descriptions of the testing results, without correlating changes in the beam number, beam quality fluctuations, and spatial multiplexing layer number of massive MIMO antennas with the throughput sampling process at the same temporal granularity.

[0003] In relatively common implementations, network speed detection systems acquire throughput samples at fixed sampling periods and then average, peak, or remove outliers after the detection is complete. Some solutions combine factors such as wireless link retransmission counts, scheduling resource occupancy ratios, or signal quality changes to filter or weight low-rate samples to reduce the impact of random jitter on the detection results. However, these processes typically rely on the end-to-end rate sequence itself or a single wireless quality metric for correction, failing to distinguish between different states such as stable beam transmission, beam switching transitions, and beam mismatch. Since beam state changes in massive MIMO scenarios are short-term and spatially correlated, relying solely on sudden rate drops or signal quality changes is insufficient to accurately identify the source of rate anomalies.

[0004] The core problem with existing technologies lies in the lack of decoupling between network speed detection results and the operational status of massive MIMO beams. When beam number changes, beam reference signal quality steps, scheduling resource continuity interruptions, or spatial multiplexing layer abrupt changes occur within the detection time window, the throughput sampling sequence will exhibit short-term drops, zero-rate points, or concentrated retransmission segments. Conventional detection methods tend to incorporate these fragmented rate changes caused by beam switching, handover guard intervals, scheduling gaps, or beam mismatches into the overall network speed calculation, causing the detected values ​​to deviate from the transmission capacity under stable wireless bearer conditions. This results in high dispersion of detection results for the same location and the same serving cell under similar wireless conditions, and it is difficult to determine whether the rate drop is due to a decrease in actual network bearer capacity or detection distortion caused by beam state disturbances. Summary of the Invention

[0005] The purpose of this invention is to provide a 5G network speed detection system based on massive MIMO technology, which can solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A 5G network speed detection system based on massive MIMO technology includes a detection data synchronous acquisition module, a beam state decoupling module, and a rate correction output module. The detection data synchronous acquisition module is used to synchronously acquire throughput sampling sequence, beam number change sequence, beam reference signal quality, spatial multiplexing layer number, scheduling resource occupancy ratio and wireless link retransmission information within the network speed detection time window; The beam state decoupling module is used to divide the network speed detection time window into a stable beam segment, a switching transition segment, and a beam mismatch segment according to the beam number change sequence and the beam reference signal quality. The rate correction output module is used to perform weighted correction on the throughput sampling sequence according to the segment type, generate 5G network speed detection results decoupled by beam state, and form a 5G network speed detection system based on massive MIMO technology.

[0007] Preferably, the detection data synchronization acquisition module includes a detection task identifier generation submodule, a wireless side parameter acquisition submodule, and a velocity flow parameter acquisition submodule; The detection task identifier generation submodule assigns a unified detection task identifier to the throughput sampling sequence and large-scale antenna beam status data within the same network speed detection time window. The wireless side parameter acquisition submodule records the serving cell, beam number, beam reference signal quality, modulation and coding level, spatial multiplexing layer number, and scheduling resource usage ratio according to the unified detection task identifier; The velocity flow parameter acquisition submodule records the upload or download direction, sampling time, sampling byte size, and number of retransmissions at the transport layer according to the unified detection task identifier.

[0008] Preferably, the beam state decoupling module includes a beam change detection submodule and a segment boundary determination submodule; The beam change detection submodule generates a beam event sequence based on changes in beam number, step change in beam reference signal quality, interruption of scheduling resource continuity, and sudden change in the number of spatial multiplexing layers at adjacent sampling times. The segment boundary determination submodule aligns the rate slope changes of the beam event sequence with those of the throughput sampling sequence in time, and marks the sampling intervals between adjacent beam events as stable beam segments, switching transition segments, or beam mismatch segments.

[0009] Preferably, the rate correction output module includes a sampling confidence calculation submodule and an abnormal sampling screening submodule; The sampling reliability calculation submodule generates a reliability weight for each throughput sampling point based on the segment type, beam reference signal quality stability, scheduling resource occupation continuity, wireless link retransmission density, and spatial multiplexing layer retention time. The abnormal sampling and screening submodule filters or downweights rate drop points within the switching protection interval, zero rate points within the scheduling window, and distortion points in retransmission concentrated segments according to the confidence weight.

[0010] Preferably, the rate correction output module further includes a normalization correction submodule; The normalization correction submodule establishes a set of rate correction factors based on the beam reference signal quality reference interval, the spatial multiplexing layer reference state, the scheduling resource occupancy ratio reference interval, and the wireless link retransmission reference interval. It then maps the set of rate correction factors to the throughput sampling sequence processed according to the confidence weight segment by segment to obtain the main rate value corresponding to the stable beam segment, the compensation rate value corresponding to the handover transition segment, and the low confidence rate value corresponding to the beam mismatch segment.

[0011] Preferably, the detection data synchronization acquisition module further includes an adaptive sampling granularity configuration submodule; The adaptive sampling granularity configuration submodule generates the basic sampling granularity and event-enhanced sampling granularity for the current network speed detection time window based on the beam event density, scheduling cycle length, beam reference signal quality fluctuation frequency, and throughput sampling dispersion within the historical network speed detection time window. When the beam number change sequence occurs, or when the beam reference signal quality undergoes a step change, the wireless side parameter acquisition submodule supplements the beam status data and velocity flow parameters according to the event enhancement sampling granularity.

[0012] Preferably, the beam state decoupling module further includes a state transition constraint submodule; The state transition constraint submodule constructs a segment state transition matrix based on the beam number retention ratio within the continuous sampling interval, the beam reference signal quality drop-off magnitude, the spatial multiplexing layer stabilization time, and the number of throughput slope reversals. The segment state transition matrix is ​​set with transition constraints for entering a switching transition segment from a stable beam segment, entering a beam mismatch segment from a switching transition segment, and returning to a stable beam segment from a beam mismatch segment.

[0013] Preferably, the rate correction output module further includes a result labeling submodule; The result labeling submodule generates a confidence interval for the detection result, a beam stability identifier, and an anomaly cause label based on the distribution of the confidence weight, the deviation between the main rate value and the compensation rate value, the proportion of the low confidence rate value within the network speed detection time window, the wireless link retransmission density, and the continuity of scheduling resource occupation. The abnormal cause labels include beam switching interference labels, beam mismatch labels, scheduling gap labels, and retransmission concentration labels.

[0014] Preferably, it also includes a correction parameter calibration module; The correction parameter calibration module establishes a historical detection sample library according to the combination of serving cell, frequency band, detection direction, beam number combination and spatial multiplexing layer combination, and extracts the stable beam segment rate distribution, handover transition segment duration distribution, beam mismatch segment retransmission distribution and abnormal cause label distribution of similar detection tasks from the historical detection sample library. The correction parameter calibration module updates the confidence weight, the rate correction factor set, and the judgment parameters of the anomaly cause label based on the distribution results of the similar detection tasks.

[0015] Preferably, it also includes a consistency verification module; The detection consistency verification module marks the current network speed detection time window as a consistent window, a window that needs to be retested, or a window that needs to be encrypted, based on the similarity of beam number change sequences between adjacent network speed detection time windows, the throughput difference at the basic sampling granularity, the beam reference signal quality difference at the event-enhanced sampling granularity, and the wireless link retransmission difference. For the sampling window that needs to be encrypted, the detection consistency verification module sends a configuration instruction to the adaptive sampling granularity configuration submodule to shorten the basic sampling granularity and extend the duration of event enhancement sampling.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. Within the network speed detection time window, throughput sampling sequences, beam number change sequences, beam reference signal quality, spatial multiplexing layer number, scheduling resource occupancy ratio, and wireless link retransmission information are simultaneously collected. Wireless side parameters are temporally correlated with speed measurement stream parameters using a unified detection task identifier, ensuring that network speed detection data and massive MIMO beam status are analyzed on the same basis. Beam event sequences are generated based on beam number changes, beam reference signal quality steps, scheduling resource continuity interruptions, and spatial multiplexing layer abrupt changes. These sequences, combined with rate slope changes in the throughput sampling sequence, determine stable beam segments, handover transition segments, and beam mismatch segments, enabling differentiated processing of rate samples corresponding to different beam states within the detection time window. By assigning confidence weights to different segments and filtering or downweighting rate drop points within handover guard intervals, zero-rate points within scheduling windows, and distortion points in retransmission concentration segments, the impact of beam status disturbances on the overall network speed detection value is reduced, making the output detection results closer to the transmission state under stable beam carrying conditions.

[0017] 2. A set of rate correction factors is established based on the beam reference signal quality reference interval, the spatial multiplexing layer reference state, the scheduling resource occupancy ratio reference interval, and the wireless link retransmission reference interval. This yields the main rate value corresponding to stable beam segments, the compensation rate value corresponding to handover transition segments, and the low-confidence rate value corresponding to beam mismatch segments, providing segment-based source information for network speed detection results. By adaptively configuring the sampling granularity, beam state data and velocity flow parameters are supplemented when beam number changes or beam reference signal quality steps occur, improving data integrity under short-term beam events. Through state transition constraints, result labeling, historical detection sample calibration, and detection consistency verification, a confidence interval for the detection results, beam stability identifiers, and anomaly cause labels are generated. The confidence weight, rate correction factor, and anomaly cause determination parameters are updated based on the historical distribution of similar detection tasks. This processing reduces the result dispersion caused by differences in sampling granularity, beam event density, and retransmission between adjacent detection time windows, making the detection results suitable for rate analysis under massive MIMO antenna beam states in 5G mobile communication networks. Attached Figure Description

[0018] Figure 1 This is a flowchart of the overall processing flow of a 5G network speed detection system based on massive MIMO technology. Figure 2 Flowchart for generating the synchronous acquisition and joint recording queue of detection data; Figure 3 Flowchart for beam event recognition and segment state decoupling; Figure 4 Flowchart for sampling confidence calculation, rate correction, and result labeling; Figure 5This is a flowchart for adaptive sampling, historical parameter calibration, and consistency verification of detection. Detailed Implementation

[0019] As attached Figure 1 In one embodiment, a fifth-generation mobile communication network speed detection system based on massive MIMO technology includes a detection data synchronization acquisition module, a beam state decoupling module, and a rate correction output module. The detection data synchronization acquisition module simultaneously acquires the throughput sampling sequence, beam number change sequence, beam reference signal quality, spatial multiplexing layer number, scheduling resource occupancy ratio, and radio link retransmission information within a single network speed detection time window. The throughput sampling sequence is calculated from the number of bytes transmitted by the speed measurement stream between adjacent sampling times. The beam number change sequence consists of the beam identifiers allocated by the serving cell to the terminal under test or the detection task within the time window. The beam reference signal quality describes the available radio quality of the corresponding beam at the sampling time. The spatial multiplexing layer number describes the data transmission layer state that the massive MIMO can allocate at that sampling time. The scheduling resource occupancy ratio describes the radio resource share obtained by the detection task. The radio link retransmission information characterizes the retransmission density caused by changes in link quality during transmission.

[0020] In this embodiment, the detection data synchronization acquisition module incorporates all acquisition objects into the same detection time reference. Velocity flow data is typically based on transmission statistics time, while radio-side data is typically based on serving cell scheduling time or measurement report time. Directly splicing these two types of data can cause misalignment between beam events and rate changes. To avoid this misalignment, the detection data synchronization acquisition module generates a unified detection task identifier for each detection, records the time of the throughput sampling point as the master detection time, and records the acquisition time of the radio-side parameters as candidate association times. For each master detection time, the record with the smallest time distance from the candidate association times is selected as the association object. When a candidate association time cannot cover the master detection time, the beam number, spatial multiplexing layer number, and scheduling resource occupancy ratio are supplemented using the hold-up state of adjacent radio-side records. Simultaneously, the supplemented record is marked as a presumed record for reduced weight in subsequent confidence calculations. The above processing does not change the original radio-side data; it only forms an aligned copy that can be used for decoupling analysis.

[0021] The alignment process can be determined using the following formula to determine the time distance between the throughput sampling point and the wireless side parameter record:

[0022] in, Indicates the first The minimum time distance between each throughput sampling point and the wireless side parameter record. Indicates the first Sampling time of each throughput sampling point Represents the set of wireless side parameter records The first in Each recording time point. For example, if the sampling time for a certain throughput sampling point is 120 milliseconds after the start of detection, and the wireless side parameter recording times are 110 milliseconds, 126 milliseconds, and 145 milliseconds respectively, then the three time intervals are 10 milliseconds, 6 milliseconds, and 25 milliseconds respectively. Take 6 milliseconds, and associate this sampling point with the beam number, beam reference signal quality, spatial multiplexing layer number, and scheduling resource usage ratio corresponding to 126 milliseconds.

[0023] The beam state decoupling module receives the joint sequence after time alignment and divides the network speed detection time window into stable beam segments, handover transition segments, and beam mismatch segments based on the beam number change sequence and beam reference signal quality. Stable beam segments correspond to sampling intervals where the beam number remains constant, reference signal quality fluctuations are small, scheduling resource allocation is continuous, and throughput changes are gradual. Handover transition segments correspond to transitional sampling intervals after beam number changes, reference signal quality undergoes a step change, or scheduling resource continuity is interrupted. Beam mismatch segments correspond to sampling intervals where the beam number remains constant but reference signal quality continuously declines, retransmission information appears frequently, and the throughput slope reverses abnormally. The beam state decoupling module does not directly determine the segment type based on a single indicator but incorporates beam number, beam reference signal quality, spatial multiplexing layer number, scheduling resource occupancy ratio, and throughput changes into the determination, ensuring that segment division matches the actual operating state of the massive MIMO antenna.

[0024] In this embodiment, the beam state decoupling module generates beam event strength based on sampling points. The calculation of beam event strength employs multiple normalization factors to prevent any single radio-side indicator from abnormally dominating segment partitioning. Whether the beam number changes is expressed using an indicator function; changes in beam reference signal quality are expressed using the difference between adjacent sampling points; changes in the number of spatial multiplexing layers are expressed using the layer difference; and the continuity of scheduling resources is expressed using whether a window of resource occupancy occurs. The beam event strength can be expressed as: in, Indicates the first The beam event intensity corresponding to each sampling point , , and These represent the weighting coefficients for changes in beam number, beam reference signal quality, number of spatial multiplexing layers, and scheduling window, respectively. Indicates the first Beam number of each sampling point Indicates the first Beam reference signal quality at each sampling point Indicates the first Number of spatial multiplexing layers per sampling point Indicates the first The proportion of scheduling resources used by each sampling point The scale factor represents the quality difference of the beam reference signal within the current detection time window. This represents the scale factor indicating the difference in the number of spatial multiplexing layers within the current detection time window. For example, if the beam number of adjacent sampling points changes from 3 to 5, and the change in beam reference signal quality is 4, the current scale factor... The current scale factor is 8, the number of spatial multiplexing layers changes from 2 to 1. If the value is 2, the resource utilization ratio is not 0, and the weights are 0.4, 0.2, 0.2, and 0.2, then... This result indicates that there is a strong beam state change near the sampling point, which should be considered as a candidate boundary for switching or mismatch analysis.

[0025] The rate correction output module performs weighted correction on the throughput sampling sequence based on the segment type. Specifically, for stable beam segments, the rate correction output module retains the main contribution of the throughput sampling points; for handover transition segments, the rate correction output module reduces the weight of sampling points based on beam event strength and scheduling resource continuity; for beam mismatch segments, the rate correction output module combines retransmission information and beam reference signal quality decline status to classify them into the low-confidence rate set. After the above processing, the rate correction output module outputs the fifth-generation mobile communication network speed detection result decoupled from beam state, and saves the main rate value corresponding to stable beam segments, the compensated rate value corresponding to handover transition segments, and the low-confidence rate value corresponding to beam mismatch segments. The output is not a simple average of the original throughput, but rather a reconstruction of the rate contribution according to segment state within the same detection time window, avoiding beam state disturbances from directly entering the overall detection result.

[0026] Table 1 shows the main fields of the joint sequence and their processing methods in this embodiment, which are used to explain how data from different sources enter the time alignment process under the same detection task identifier, and to explain the purpose of each field in the subsequent decoupling and correction.

[0027] Table 1. Main fields of the joint sequence and their processing methods ; As attached Figure 2In one embodiment, the detection data synchronization acquisition module includes a detection task identifier generation submodule, a radio side parameter acquisition submodule, and a velocity flow parameter acquisition submodule. The detection task identifier generation submodule generates a unique identifier when a detection task is established and writes this identifier into the upload detection task, download detection task, radio side parameter reading task, and result output task. For different sampling points within the same detection time window, the detection task identifier does not change with beam switching; for re-initiated detection tasks, a new detection task identifier is generated even if the serving cell and terminal location remain unchanged. In this way, beam events and throughput fluctuations in the same detection can be categorized into the same analysis object, and data across detection tasks will not be mistakenly incorporated into the current analysis process.

[0028] The wireless side parameter acquisition submodule records the serving cell, beam number, beam reference signal quality, modulation and coding level, spatial division multiplexing layer number, and scheduling resource occupancy ratio according to a unified detection task identifier. The modulation and coding level is used to assist in identifying rate changes caused by link adaptation; the spatial division multiplexing layer number is used to assist in identifying the transmission layer status of massive MIMO at different sampling times; and the scheduling resource occupancy ratio is used to distinguish between insufficient wireless resources and beam quality changes. The velocity flow parameter acquisition submodule records the upload or download direction, sampling time, sampled byte size, and transmission layer retransmission count according to a unified detection task identifier. Upload and download directions are saved separately because the sources of uplink and downlink beam state changes are different; mixing statistics would cause segment determination to lose directionality.

[0029] In a preferred embodiment, the sampled byte count is not directly output as the rate, but is converted into a throughput sample value based on the interval between adjacent sampling times. If the number of retransmissions at a certain sampling point increases in a concentrated manner and the retransmission information of the wireless link also increases synchronously, the velocity flow parameter acquisition submodule marks the sampling point as a retransmission-related sampling point; if the number of retransmissions at the transport layer increases but the retransmission information of the wireless link does not increase, the sampling point is retained as a candidate point for transport layer fluctuations and is not directly included in the beam mismatch segment. The above distinction keeps the wireless side beam state disturbance and transport layer changes separate at the data entry stage, facilitating subsequent processing for technical reasons.

[0030] In this embodiment, the velocity stream parameter acquisition submodule also records the detection direction and velocity stream status. The detection direction indicates whether the sampled value originates from uplink or downlink detection, and the velocity stream status indicates whether the sampling point is in the initial connection establishment stage, stable transmission, or shutdown transition. Sampling points in the initial connection establishment stage and shutdown transition are not directly included in the main rate calculation of the stable beam segment, but are saved as detection boundary samples. This processing can avoid the influence of non-radio side factors on the beam status decoupling results during the velocity stream establishment process. After completing the acquisition, the detection data synchronization acquisition module generates a joint record queue arranged according to the sampling time. Each joint record contains at least the detection task identifier, sampling time, detection direction, throughput sampled value, beam number, beam reference signal quality, spatial division multiplexing layer number, scheduling resource occupancy ratio, and radio link retransmission information.

[0031] Furthermore, as shown in the appendix Figure 3 The beam state decoupling module includes a beam change detection submodule and a segment boundary determination submodule. The beam change detection submodule reads adjacent sampling points from the joint record queue and generates a beam event sequence based on beam number changes, beam reference signal quality steps, scheduling resource continuity interruptions, and spatial multiplexing layer abrupt changes. The beam event sequence is represented by sampling point indices, which correspond to specific positions in the joint record queue. The segment boundary determination submodule aligns the beam event sequence with the rate slope changes of the throughput sampling sequence in time. If, after a beam event occurs, the throughput slope synchronously decreases or reverses within adjacent sampling ranges, the segment boundary determination submodule uses this beam event as the starting boundary of a handover transition segment; if the beam event does not cause a change in the throughput slope, the event is recorded as a weak event and does not form a segment boundary independently.

[0032] The rate slope change of the throughput sampling sequence can be calculated using the following formula: in, Indicates the first The slope of the throughput change at each sampling point relative to the previous sampling point Indicates the first The throughput sampling value of each sampling point Indicates the first The throughput sampling value of each sampling point and These represent the sampling times of two adjacent sampling points. Division represents the change in throughput per unit time. For example, if the first... The throughput sampling value for the first sampling point is 800 megabits per second. The throughput of each sampling point is 620 megabits per second, and the difference between two sampling times is 20 milliseconds. This value indicates a throughput decrease of 9 megabits per second per millisecond. If this decrease is accompanied by a beam number change or a step change in beam reference signal quality, the sampling point is treated as a candidate boundary for a switching transition segment by the segment boundary determination submodule.

[0033] In this embodiment, the segment boundary determination submodule merges consecutive weak events into adjacent stable beam segments, marks the sampling interval near strong events as a handover transition segment, and re-determines the recovery interval after the handover transition segment. If the beam number remains unchanged after the handover transition, the beam reference signal quality recovers to a stable distribution range within the current detection time window, the scheduling resource occupancy ratio returns to continuity, and the throughput slope returns to a low-fluctuation state, then the recovery interval is marked as a new stable beam segment. If the beam number remains unchanged but the beam reference signal quality continues to decline, radio link retransmission information occurs frequently, and the throughput sampling value shows low-level fluctuations, then the recovery interval is marked as a beam mismatch segment. This segment division method allows the detection time window to be divided into multiple continuous intervals with clear radio-side states.

[0034] Table 2 provides the criteria for determining fragment types. The criteria in the table are not independent thresholds, but rather serve to illustrate the logic of the fragment boundary determination submodule in performing combined judgments on state consistency within the same joint record queue.

[0035] Table 2. Criteria for Fragment Type Determination and Explanation of State Combination Judgment ; As attached Figure 4 In one embodiment, the rate correction output module includes a sampling reliability calculation submodule and an abnormal sampling screening submodule. The sampling reliability calculation submodule generates a reliability weight for each throughput sampling point based on segment type, beam reference signal quality stability, scheduling resource occupancy continuity, radio link retransmission density, and spatial multiplexing layer retention time. The reliability weight is represented in a normalized manner; the higher the weight, the closer the sampling point is to a stable beam-bearing state; the lower the weight, the more likely the sampling point is to be affected by handover protection intervals, scheduling gaps, or retransmission concentrations. The abnormal sampling screening submodule does not directly delete all low-rate sampling points, but instead determines whether to screen or downweight them based on the segment where the low-rate sampling point is located and the corresponding radio-side event. Rate drop points located within handover protection intervals are downweighted, zero-rate points located within scheduling gaps are excluded from the main rate calculation, and distortion points located in retransmission concentration segments are retained in the low-reliability rate set.

[0036] Credibility weights can be calculated using the following formula: in, Indicates the first The confidence weight of each throughput sampling point This represents the segment state coefficient. The value of the stable beam segment is higher than that of the switching transition segment, and the value of the switching transition segment is higher than that of the beam mismatch segment. This represents the stability coefficient of the beam reference signal quality. The closer the quality is to the stable distribution of the current detection time window, the higher the value. This represents the continuity coefficient of scheduling resource occupancy. It has a higher value when resources are continuously allocated and a lower value when there is a scheduling window. This represents the retransmission density of the wireless link. The addition in the denominator is used to reduce the weight when the retransmission density increases. For example, if a sampling point is located in a handover transition segment, Beam reference signal quality stability Continuity of scheduling resource usage Wireless link retransmission density ,but This sampling point participates in the calculation with a low weight in the overall rate correction, and its source of transition fragments is preserved.

[0037] Furthermore, the rate correction output module also includes a normalization correction submodule. The normalization correction submodule establishes a set of rate correction factors based on the beam reference signal quality reference interval, the spatial multiplexing layer reference state, the scheduling resource occupancy ratio reference interval, and the radio link retransmission reference interval. The reference intervals are derived from the statistical distribution of stable beam segments within the current detection time window, or from the historical stable segment distribution of the same serving cell, same detection direction, and same beam number combination. The normalization correction submodule maps the set of rate correction factors to the throughput sampling sequence after confidence processing, segment by segment, to obtain the main rate value corresponding to the stable beam segment, the compensation rate value corresponding to the handover transition segment, and the low-confidence rate value corresponding to the beam mismatch segment.

[0038] In this embodiment, the main rate value of the stable beam segment is calculated using a confidence weighting method; the compensation rate value of the handover transition segment is determined by the confidence sampling points within the transition segment and the rate boundaries of the stable segments before and after the transition; the low confidence rate value of the beam mismatch segment is only used for result labeling and anomaly attribution, and does not directly dominate the network speed detection output. The rate correction value can be calculated using the following formula: in, Indicates the first The correction rate value for each segment, Indicates the first The set of sampling points contained in each segment Indicates the first The confidence weight of each sampling point Indicates the first The throughput sampling value of each sampling point Indicates the first The rate correction factor corresponds to each sampling point. For example, a stable beam segment contains 3 sampling points with throughputs of 780, 800, and 760 megabits per second, confidence weights of 0.9, 1.0, and 0.8, and rate correction factors of 1.0, 0.98, and 1.02, respectively. Then the numerator is... The denominator is 2.7. Approximately 780.06 megabits per second. This calculation result serves as a candidate value for the main rate of this stable beam segment.

[0039] In a preferred embodiment, the normalization correction submodule performs fractional processing on the rate correction factor. The beam reference signal quality component is used to correct the impact of quality deviation from the reference range, the spatial multiplexing layer number component is used to correct the bearer differences caused by changes in the number of transmission layers, the scheduling resource component is used to correct the rate changes caused by changes in the resource occupancy ratio, and the radio link retransmission component is used to correct the sampling distortion caused by retransmission. The components are multiplied to form (F_k). When any component is missing, the normalization correction submodule uses the median value of similar sampling points within the current segment to fill in the missing component and sets a low-confidence flag for the filled component. This method ensures that the correction calculation remains traceable even when the radio-side fields are incomplete, avoiding the discarding of the entire data segment due to the missing of individual fields.

[0040] As attached Figure 5 In one embodiment, the detection data synchronization acquisition module further includes an adaptive sampling granularity configuration submodule. The adaptive sampling granularity configuration submodule generates a basic sampling granularity and an event-enhanced sampling granularity for the current network speed detection time window based on the beam event density, scheduling cycle length, beam reference signal quality fluctuation frequency, and throughput sampling dispersion within the historical network speed detection time window. The basic sampling granularity is used for recording throughput and wireless side parameters under normal detection conditions, while the event-enhanced sampling granularity is used for supplementary sampling near beam number changes or beam reference signal quality step changes. The event-enhanced sampling granularity is not a fixed value but is jointly determined by the beam event density of the most recent similar detection tasks and the rate dispersion state of the current detection task.

[0041] Specifically, the adaptive sampling granularity configuration submodule reads the beam event distribution of the same serving cell, detection direction, and frequency band in historical detection samples before detection begins, and determines whether to increase the sampling density near the event based on the concentration of beam events within the time window. If historical samples show that a serving cell frequently experiences beam reselection in the initial stage of detection, the basic sampling granularity remains unchanged, and the event-enhanced sampling granularity is activated near the beam number change location in the initial stage of detection. If a beam number change or a step change in beam reference signal quality occurs during the current detection process, the radio-side parameter acquisition submodule immediately supplements the beam status data and velocity flow parameters according to the event-enhanced sampling granularity. The supplemented data still uses the same detection task identifier and is marked as originating from event-enhanced sampling in the joint record queue.

[0042] In this embodiment, the duration of event enhancement sampling is jointly determined by the beam event strength and the throughput slope. If the beam event strength is high but the throughput slope remains unchanged, event enhancement sampling continues until the beam reference signal quality returns to a stable distribution range; if the beam event strength is high and the throughput slope decreases, event enhancement sampling continues until the slope recovers to a low-fluctuation state. The adaptive sampling granularity configuration submodule sorts the basic sampling records and event enhancement sampling records by time, with records at repeated times using the event enhancement sampling records as the standard, and retains the index of the overwritten basic records. This processing provides finer temporal resolution for short-term identification of beam switching and beam mismatch.

[0043] In one embodiment, the beam state decoupling module further includes a state transition constraint submodule. The state transition constraint submodule constructs a segment state transition matrix based on the beam number preservation ratio within a continuous sampling interval, the beam reference signal quality fall-off magnitude, the spatial multiplexing layer stabilization time, and the number of throughput slope inversions. Segment states include stable beam segments, switching transition segments, and beam mismatch segments. The state transition constraint submodule sets transition constraints on the paths from stable beam segments to switching transition segments, from switching transition segments to beam mismatch segments, and from beam mismatch segments back to stable beam segments. This constraint does not exclude actual state changes but rather avoids frequent jumps in segment states between adjacent sampling points due to single-point noise.

[0044] The state transition cost can be expressed by the following formula: in, Indicates the state of the fragment. Transition to fragment state The cost of transfer, , , , and This represents the weight of each transfer factor. This indicates that the beam numbering remains proportional within the continuous sampling interval. This indicates the magnitude of the beam reference signal quality degradation. Indicates the stabilization time of the spatial multiplexing layer. This indicates the number of times the throughput slope has reversed. and These represent the average throughput slopes of two adjacent candidate segments. For example, if the beam numbering ratio remains at 0.9 within a certain interval, the normalized value for quality dropout amplitude is 0.2, the normalized value for layer stabilization time is 0.1, the number of slope reversals is 1, the average slope difference between adjacent segments is 2, and the weights are 0.2, 0.2, 0.2, 0.2, and 0.2 respectively, then... This cost can serve as a quantitative basis for selecting state transition paths.

[0045] In this embodiment, the state transition constraint submodule searches for a segment path with low cost and conforming to state logic according to the sampling order of the detection time window. A stable beam segment can directly enter a handover transition segment or remain a stable beam segment. When a stable beam segment directly enters a beam mismatch segment, it needs to simultaneously satisfy the continuous decline in beam reference signal quality and the concentrated occurrence of wireless link retransmission information. When a handover transition segment enters a stable beam segment, the beam number needs to be maintained and the scheduling resource occupancy ratio needs to be restored to continuity. When a handover transition segment enters a beam mismatch segment, the quality needs to not recover after the transition and the throughput slope needs to remain at a low level. When a beam mismatch segment returns to a stable beam segment, the quality needs to recover within the continuous sampling interval, the layer number needs to stabilize, and the retransmission density needs to decrease. Through the transition path constraint, a single abnormal sampling point will not individually change the segment state, and continuous abnormal states can be retained as independent segments.

[0046] In one embodiment, the rate correction output module further includes a result labeling submodule. The result labeling submodule generates a detection result confidence interval, beam stability identifier, and anomaly cause label based on the distribution of confidence weights, the deviation between the main rate value and the compensated rate value, the proportion of low-confidence rate values ​​within the network speed detection time window, the wireless link retransmission density, and the continuity of scheduling resource occupancy. The anomaly cause labels include beam switching interference labels, beam mismatch labels, scheduling gap labels, and retransmission concentration labels. The result labeling submodule does not treat these labels as manual judgment explanations but rather as components of the detection result, outputting and storing them along with the corrected rate value.

[0047] Specifically, when the handover transition segment accounts for a high proportion of the detection time window, and there is an identifiable difference between the main rate value and the compensated rate value, the result labeling submodule generates a beam switching interference label; when the beam number remains but the beam reference signal quality continues to decline, the retransmission density is concentrated, and the low confidence rate value remains consistently low, a beam mismatch label is generated; when the scheduling resource occupancy ratio shows consecutive empty windows, and the zero-rate sampling point coincides with the empty window position, a scheduling empty window label is generated; when wireless link retransmission information appears in a concentrated manner, and the throughput slope repeatedly reverses, a retransmission concentration label is generated. Different labels can exist simultaneously, but each label must be associated with a specific sampling interval and corresponding segment type to prevent abnormal causes from being separated from the sampling data.

[0048] Table 3 shows the output fields of the results annotation submodule, illustrating how additional outputs beyond the corrected rate values ​​relate to fragment segmentation and anomaly causes.

[0049] Table 3. Description of output fields and data granularity of the results annotation submodule. ; In one embodiment, the detection system further includes a correction parameter calibration module. The correction parameter calibration module establishes a historical detection sample library based on combinations of serving cell, frequency band, detection direction, beam number combination, and spatial multiplexing layer number. The historical detection sample library does not store a simple collection of original detection results, but rather structured samples containing segment boundaries, segment types, stable beam segment rate distribution, handover transition segment duration distribution, beam mismatch segment retransmission distribution, anomaly cause label distribution, confidence weight, and rate correction factor. After the current detection is completed, the correction parameter calibration module writes the segment data from the current detection result into the corresponding combination of historical detection sample libraries, and retains the time window identifier and detection direction at the time of sample generation for subsequent similar tasks to use.

[0050] The parameter correction calibration module updates the confidence weights, rate correction factor set, and anomaly cause label determination parameters based on the historical distribution of similar detection tasks. The update process is based on stable beam segments, as these represent the bearer state less affected by handover disturbances under the serving cell, detection direction, and beam combination. The handover transition segment duration distribution is used to update the weighting range of the transition segment, the beam mismatch segment retransmission distribution is used to update the identification conditions for low confidence rate values, and the anomaly cause label distribution is used to update the combination conditions for label determination. The parameter correction calibration module avoids directly writing a single anomaly detection into the global parameters; instead, it uses a smooth update method based on current and historical samples.

[0051] Parameter updates can be represented by the following formula: in, Indicates the first Corrected parameters after class detection combination update. Indicates the first Corrected parameters before class detection combination update. Indicates the first Update weights for class detection combinations. This represents the median value of similar samples after being written into the historical detection sample database. For example, if the old confidence parameter for a certain serving cell, detection direction, and beam number combination is 0.7, the median value of similar samples is 0.8, and the update weight is 0.25, then... The updated results are used to calculate confidence weights or correction factors for subsequent similar detection tasks.

[0052] In this embodiment, the parameter calibration module also performs group maintenance on the determination parameters of the anomaly cause tags. The determination parameters for the beam switching interference tag mainly come from the duration of the switching transition segment and the deviation between the main rate value and the compensation rate value; the determination parameters for the beam mismatch tag mainly come from the amplitude of the beam reference signal quality decline, the proportion of low-confidence rate values, and the retransmission distribution; the determination parameters for the scheduling empty window tag mainly come from the overlap relationship between the continuous intervals where the scheduling resource occupancy ratio is 0 and the zero-rate sampling points; the determination parameters for the retransmission concentration tag mainly come from the wireless link retransmission density and the number of throughput slope reversals. The above group update ensures that different anomaly sources correspond to different parameters, and does not use the same anomaly condition to explain all rate decreases.

[0053] Table 4 lists the classification fields of the historical test sample library. This table explains how the correction parameter calibration module maintains the samples according to the wireless side context, ensuring that the calibration results correspond to the operating status of the massive MIMO antenna.

[0054] Table 4. Description of Historical Test Sample Library Classification Fields and Calibration Purposes ; In one embodiment, the detection system further includes a detection consistency verification module. This module marks the current network speed detection time window as a consistent window, a window requiring retesting, or a window requiring encrypted sampling, based on the similarity of beam number change sequences between adjacent network speed detection time windows, throughput differences at the basic sampling granularity, beam reference signal quality differences at the event-enhanced sampling granularity, and wireless link retransmission differences. Adjacent network speed detection time windows can be continuously segmented sub-windows within the same detection task, or two adjacent detection windows initiated under the same serving cell and the same detection direction. The verification module does not modify the already output original sampling records; instead, it classifies the comparability of the detection results and feeds the classification results back to the adaptive sampling granularity configuration sub-module.

[0055] The consistency review score can be calculated using the following formula: in, This represents the consistency review score of the (n)th detection time window relative to the previous detection time window. , , and This represents the weight of each consistency factor. This indicates the similarity of beam number change sequences between adjacent detection time windows. This represents the normalized value of the throughput difference at the basic sampling granularity. This represents the normalized value of the beam reference signal quality difference at the event-enhanced sampling granularity. This represents the normalized difference value for wireless link retransmissions. (The part in parentheses...) , and This indicates that the smaller the difference, the higher the contribution to consistency. For example, if the similarity of beam number change sequences between adjacent windows is 0.8, the normalized value of throughput difference is 0.2, the normalized value of beam reference signal quality difference is 0.1, the normalized value of retransmission difference is 0.3, and all four weights are 0.25, then... This score is used to determine whether the current window can be merged with the previous window for analysis or needs to be reviewed separately.

[0056] In this embodiment, when the consistency verification module detects that the current window and adjacent windows have similar beam number change sequences, low throughput differences, low beam reference signal quality differences, and low retransmission differences, it marks the current window as a consistent window and allows the rate correction output module to use the stable beam segment distributions of the two windows together for main rate calculation. When the beam number change sequences are similar but the throughput and retransmission differences increase significantly, the verification module marks the current window as a window requiring retesting and saves the triggering reason. If the event-enhanced sampling data in the current window is insufficient to cover the beam number change or quality step interval, the verification module marks the current window as a window requiring encrypted sampling and sends configuration instructions to the adaptive sampling granularity configuration submodule to shorten the basic sampling granularity and extend the event-enhanced sampling duration.

[0057] In a preferred embodiment, the retest window does not directly overwrite the original detection results. The detection consistency verification module saves the correction rate value, confidence interval, beam stability identifier, and anomaly cause label of the original window as a record to be verified, and compares the beam event sequence and segment state transition path of the two detections after the retest is completed. If the beam event sequence of the retest window differs significantly from that of the original window, the two detections are saved separately to avoid merging detection results under different beam states. If the beam event sequence of the retest window is similar to that of the original window, but the throughput difference is still significant, the correction parameter calibration module writes this combination into the abnormal sample set and reduces the update weight of this combination in short-term calibration. This process prevents the historical detection sample library from being excessively affected by occasional abnormal segments.

[0058] In one embodiment, the overall system operation is as follows: After the detection task is established, the detection task identifier generation submodule generates a unified detection task identifier. The upload or download speed measurement stream begins collecting the number of bytes transmitted and the number of transport layer retransmissions according to the detection direction. The radio side parameter acquisition submodule synchronously records the serving cell, beam number, beam reference signal quality, modulation and coding level, spatial division multiplexing layer number, scheduling resource occupancy ratio, and radio link retransmission information. The adaptive sampling granularity configuration submodule determines the basic sampling granularity based on the beam event density in the historical detection sample library and the current throughput dispersion. When a beam number change or quality step is detected during the detection process, the event-enhanced sampling granularity is activated. The detection data synchronization acquisition module aligns the speed measurement stream parameters with the radio side parameters according to the main detection time, forming a joint recording queue.

[0059] The beam change detection submodule calculates beam event strengths from the joint recording queue, while the segment boundary determination submodule aligns beam event strengths with throughput slope changes to form candidate segment boundaries. The state transition constraint submodule constructs a segment state transition matrix based on beam number preservation ratio, beam reference signal quality fall-off magnitude, spatial multiplexing layer stabilization time, and throughput slope inversion count, and selects segment paths that conform to state logic. After this path constraint, the joint recording queue is divided into stable beam segments, switching transition segments, and beam mismatch segments. If only stable beam segments exist within the detection time window, the rate correction output module directly calculates the main rate value based on the confidence weight; if switching transition segments or beam mismatch segments exist, the rate correction output module calculates the main rate value, compensation rate value, and low confidence rate value for each segment separately.

[0060] The sampling reliability calculation submodule calculates the reliability weight for each throughput sampling point, while the abnormal sampling screening submodule performs screening or weight reduction on sampling points in handover protection intervals, scheduling gaps, and retransmission concentrated segments. The normalization correction submodule forms a set of rate correction factors based on the beam reference signal quality reference interval, spatial multiplexing layer reference state, scheduling resource occupancy ratio reference interval, and radio link retransmission reference interval, and performs weighted correction on the segment rate. The result labeling submodule generates the corrected network speed detection result, detection result confidence interval, beam stability identifier, and abnormal cause label. The correction parameter calibration module writes the stable beam segment rate distribution, handover transition segment duration distribution, beam mismatch segment retransmission distribution, and abnormal cause label distribution into the historical detection sample library, and updates the reliability weight, rate correction factor set, and abnormal cause label judgment parameters for similar detection tasks. The detection consistency verification module marks the current window as consistent, retested, or encrypted based on the consistency verification score of adjacent windows, and feeds back the verification results to subsequent detection tasks.

[0061] In a specific application scenario, a detection task performs downlink network speed detection within a 5G mobile communication serving cell. Within the detection time window, the beam number of the first-stage sampling points remains the same, the beam reference signal quality changes little, the spatial multiplexing layer number remains constant, the scheduling resource occupancy ratio is continuous, and the throughput sampling value is within a low fluctuation range. At the middle-stage sampling points, the beam number changes, the beam reference signal quality undergoes a step change, the throughput slope changes from gradual to decreasing, and the wireless link retransmission information increases briefly. The beam state decoupling module identifies the middle-stage sampling interval as a handover transition segment. At the rear-stage sampling points, the beam number remains the changed number, the beam reference signal quality returns to a stable distribution, the throughput slope recovers, and the scheduling resource occupancy ratio remains continuous; this interval is identified as a new stable beam segment. The rate correction output module calculates the main rate value for the stable beam segments at the front and rear stages, calculates the compensation rate value for the middle-stage handover transition segment, and labels the output with a beam handover interference tag.

[0062] In another specific application scenario, no beam number change occurred within the detection time window, but the beam reference signal quality declined within a continuous sampling interval, the spatial multiplexing layer number decreased briefly, retransmission information from the wireless link was concentrated, and the throughput sampling value fluctuated repeatedly at a low level. If judged solely based on the beam number, this interval would be mistakenly identified as a stable beam segment. By jointly determining the amplitude of the beam reference signal quality decline, the stabilization time of the spatial multiplexing layer number, the retransmission density, and the number of throughput slope reversals, the state transition constraint submodule identifies this interval as a beam mismatch segment. The rate correction output module writes the rate of this interval into a low-confidence rate value, preventing it from directly dominating the main rate calculation, and the result labeling submodule generates beam mismatch tags and retransmission concentration tags. Through this processing, the system can separate the situation where the beam number has not changed but the beam carrying quality has deteriorated from the stable transmission state.

[0063] In another specific application scenario, detection tasks are executed continuously within a short period. Two adjacent detection time windows share the same serving cell, detection direction, and beam number combination. However, the second window has insufficient event-enhanced sampling records, and beam reference signal quality steps only occur at a few sampling points. After calculating the consistency review scores for adjacent windows, the detection consistency review module found that while the beam number change sequences were similar, the beam reference signal quality differed significantly at the event-enhanced sampling granularity, and the differences in wireless link retransmissions increased synchronously. The detection consistency review module marked the second window as a window requiring encrypted sampling and sent a configuration command to the adaptive sampling granularity configuration submodule, causing the next detection to add sampling records near the beam event. This configuration does not change existing detection records but provides more complete beam event observation data in subsequent detections.

[0064] In the above embodiments, the synchronous acquisition of detection data, beam state decoupling, rate correction output, adaptive sampling, state transition constraints, result labeling, parameter calibration, and consistency verification all revolve around the impact of beam state on network speed detection results in massive MIMO scenarios. The modules are connected through detection task identifiers, joint record queues, segment states, and historical detection sample libraries to avoid disconnection between end-to-end throughput data and wireless side state data. While retaining the original sampling records, the system separately processes the rate contributions corresponding to stable beams, handover transitions, and beam mismatches, enabling the 5G network speed detection results to reflect the transmission capability after beam state decoupling and providing reusable calibration data for subsequent detections under the same serving cell, detection direction, and beam combination.

Claims

1. A 5G network speed detection system based on massive MIMO technology, characterized in that, This includes a detection data synchronous acquisition module, a beam state decoupling module, and a rate correction output module; The detection data synchronous acquisition module is used to synchronously acquire throughput sampling sequence, beam number change sequence, beam reference signal quality, spatial multiplexing layer number, scheduling resource occupancy ratio and wireless link retransmission information within the network speed detection time window; The beam state decoupling module is used to divide the network speed detection time window into a stable beam segment, a switching transition segment, and a beam mismatch segment according to the beam number change sequence and the beam reference signal quality. The rate correction output module is used to perform weighted correction on the throughput sampling sequence according to the segment type, generate 5G network speed detection results decoupled by beam state, and form a 5G network speed detection system based on massive MIMO technology.

2. The 5G network speed detection system based on massive MIMO technology according to claim 1, characterized in that, The detection data synchronization acquisition module includes a detection task identifier generation submodule, a wireless side parameter acquisition submodule, and a velocity flow parameter acquisition submodule; The detection task identifier generation submodule assigns a unified detection task identifier to the throughput sampling sequence and large-scale antenna beam status data within the same network speed detection time window. The wireless side parameter acquisition submodule records the serving cell, beam number, beam reference signal quality, modulation and coding level, spatial multiplexing layer number, and scheduling resource usage ratio according to the unified detection task identifier; The velocity flow parameter acquisition submodule records the upload or download direction, sampling time, sampling byte size, and number of retransmissions at the transport layer according to the unified detection task identifier.

3. The 5G network speed detection system based on massive MIMO technology according to claim 1, characterized in that, The beam state decoupling module includes a beam change detection submodule and a segment boundary determination submodule; The beam change detection submodule generates a beam event sequence based on changes in beam number, step change in beam reference signal quality, interruption of scheduling resource continuity, and sudden change in the number of spatial multiplexing layers at adjacent sampling times. The segment boundary determination submodule aligns the rate slope changes of the beam event sequence with those of the throughput sampling sequence in time, and marks the sampling intervals between adjacent beam events as stable beam segments, switching transition segments, or beam mismatch segments.

4. The 5G network speed detection system based on massive MIMO technology according to claim 2, characterized in that, The rate correction output module includes a sampling reliability calculation submodule and an abnormal sampling screening submodule; The sampling reliability calculation submodule generates a reliability weight for each throughput sampling point based on the segment type, beam reference signal quality stability, scheduling resource occupation continuity, wireless link retransmission density, and spatial multiplexing layer retention time. The abnormal sampling and screening submodule filters or downweights rate drop points within the switching protection interval, zero rate points within the scheduling window, and distortion points in retransmission concentrated segments according to the confidence weight.

5. The 5G network speed detection system based on massive MIMO technology according to claim 4, characterized in that, The rate correction output module also includes a normalization correction submodule; The normalization correction submodule establishes a set of rate correction factors based on the beam reference signal quality reference interval, the spatial multiplexing layer reference state, the scheduling resource occupancy ratio reference interval, and the wireless link retransmission reference interval. It then maps the set of rate correction factors to the throughput sampling sequence processed according to the confidence weight segment by segment to obtain the main rate value corresponding to the stable beam segment, the compensation rate value corresponding to the handover transition segment, and the low confidence rate value corresponding to the beam mismatch segment.

6. The 5G network speed detection system based on massive MIMO technology according to claim 5, characterized in that, The detection data synchronous acquisition module also includes an adaptive sampling granularity configuration submodule; The adaptive sampling granularity configuration submodule generates the basic sampling granularity and event-enhanced sampling granularity for the current network speed detection time window based on the beam event density, scheduling cycle length, beam reference signal quality fluctuation frequency, and throughput sampling dispersion within the historical network speed detection time window. When the beam number change sequence occurs, or when the beam reference signal quality undergoes a step change, the wireless side parameter acquisition submodule supplements the beam status data and velocity flow parameters according to the event enhancement sampling granularity.

7. The 5G network speed detection system based on massive MIMO technology according to claim 6, characterized in that, The beam state decoupling module also includes a state transition constraint submodule; The state transition constraint submodule constructs a segment state transition matrix based on the beam number retention ratio within the continuous sampling interval, the beam reference signal quality drop-off magnitude, the spatial multiplexing layer stabilization time, and the number of throughput slope reversals. The segment state transition matrix is ​​set with transition constraints for entering a switching transition segment from a stable beam segment, entering a beam mismatch segment from a switching transition segment, and returning to a stable beam segment from a beam mismatch segment.

8. The 5G network speed detection system based on massive MIMO technology according to claim 7, characterized in that, The rate correction output module also includes a result labeling submodule; The result labeling submodule generates a confidence interval for the detection result, a beam stability identifier, and an anomaly cause label based on the distribution of the confidence weight, the deviation between the main rate value and the compensation rate value, the proportion of the low confidence rate value within the network speed detection time window, the wireless link retransmission density, and the continuity of scheduling resource occupation. The abnormal cause labels include beam switching interference labels, beam mismatch labels, scheduling gap labels, and retransmission concentration labels.

9. The 5G network speed detection system based on massive MIMO technology according to claim 8, characterized in that, It also includes a parameter calibration module; The correction parameter calibration module establishes a historical detection sample library according to the combination of serving cell, frequency band, detection direction, beam number combination and spatial multiplexing layer combination, and extracts the stable beam segment rate distribution, handover transition segment duration distribution, beam mismatch segment retransmission distribution and abnormal cause label distribution of similar detection tasks from the historical detection sample library. The correction parameter calibration module updates the confidence weight, the rate correction factor set, and the judgment parameters of the anomaly cause label based on the distribution results of the similar detection tasks.

10. The 5G network speed detection system based on massive MIMO technology according to claim 9, characterized in that, It also includes a consistency verification module; The detection consistency verification module marks the current network speed detection time window as a consistent window, a window that needs to be retested, or a window that needs to be encrypted, based on the similarity of beam number change sequences between adjacent network speed detection time windows, the throughput difference at the basic sampling granularity, the beam reference signal quality difference at the event-enhanced sampling granularity, and the wireless link retransmission difference. For the sampling window that needs to be encrypted, the detection consistency verification module sends a configuration instruction to the adaptive sampling granularity configuration submodule to shorten the basic sampling granularity and extend the duration of event enhancement sampling.

Images