A cluster intercom adaptive transmission method based on wireless local area network

Abstract

This invention discloses an adaptive transmission method for trunked intercom based on wireless local area networks (WLANs), comprising the following steps: collecting link-state and service-state indicators to form a state vector; calculating the carrying window and delay budget using sliding window statistics and exponential weighting, generating a resource profile including congestion and risk levels; layering voice frames and labeling key frames with the latest delivery timestamp; selecting representative terminals for group clustering based on access point identifier, channel, and link quality; and generating transmission parameters by calling an improved BBRv3 under resource profile constraints in the unicast pipeline; prioritizing the transmission of key frame copies, aggregating completion summaries, determining majority completion, and retransmitting missing key frames with targeted compensation and updating completion; switching the distribution mode and performing gray-scale rollback when delay exceeds limits, busy-occupancy surges, or roaming occurs; recording indicator and parameter versions, establishing a template library through scenario fingerprint matching, and storing versioned data. This invention improves the reliability of key voice delivery and reduces the risk of delay jitter and packet loss under congestion and roaming conditions.

Description

Technical Field

[0001] This invention relates to the field of wireless communication and voice intercom technology, and in particular to an adaptive transmission method for trunked intercom based on wireless local area network. Background Technology

[0002] With the increasing prevalence of trunked intercom services in scenarios such as emergency command, industrial park security, and large venue support, real-time voice distribution and reliable transmission technologies based on wireless LANs have received widespread attention. Existing WLAN trunked intercom systems typically use multicast / broadcast or all-user unicast methods for voice frame distribution, supplemented by fixed bitrate, fixed retransmission, or simple rate limiting strategies based on packet loss rate to maintain availability. However, in practical applications, the following problems are commonly encountered: When the channel busy ratio suddenly increases, the retransmission rate at the medium access control layer rises, or the transmission queue is backed up, the fixed transmission strategy is prone to buffer expansion, resulting in significant end-to-end latency jitter, frequent voice pops and interruptions, and a lack of controllable constraints on queuing latency. Multicast in wireless environments lacks end-to-end acknowledgment and fine-grained compensation, making it easy for some terminals to continuously miss key voice segments and be difficult to repair in time. Meanwhile, unicast to all terminals will generate sudden loads due to replication fan-out, further aggravating queue growth and congestion spread. During terminal roaming or access point switching, link quality and available bandwidth change rapidly, and existing algorithms cannot maintain a stable transmission rhythm while ensuring key voice segments, often resulting in the problem of "large-scale incomprehensibility during switching". In addition, weak link terminals often drag down the overall distribution strategy, lacking differentiated management and targeted compensation mechanisms for subgroups, leading to a simultaneous deterioration in overall throughput and latency.

[0003] Therefore, how to provide an adaptive transmission method for trunked intercom based on wireless local area networks is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0004] One objective of this invention is to propose an adaptive transmission method for trunked intercom based on wireless local area networks. This invention fully utilizes joint monitoring of link state and service state, sliding time window statistics and exponential weighted smoothing, hierarchical annotation of voice frames and latest delivery constraints of key frames, hierarchical distribution of subgroup-representative-unicast pipeline, and improved BBRv3 congestion control and completion summary directional compensation mechanism to achieve adaptive parameter generation driven by intercom resource profile, priority scheduling of key voice frames and rapid retransmission of missing frames, and switching of distribution mode and gray-scale rollback when congestion or roaming is triggered. It has the advantages of reliable delivery of key frames, stable end-to-end latency, reduced jitter and packet loss, and weak link isolation without affecting the entire group.

[0005] An adaptive transmission method for trunked intercom based on a wireless local area network according to an embodiment of the present invention includes the following steps: Step 1: Collect wireless LAN link state indicators and intercom service state indicators at the intercom terminal and distribution node, and form a state vector; Step 2: Based on the state vector, the sliding time window statistics and exponential weighted smoothing are used to calculate the carrying window and delay budget parameters, and generate a profile of available intercom resources including congestion level and risk level; Step 3: The speaker's speech frames are divided into start segments, energy peak segments and coding reference frames, and the latest delivery timestamp is marked for key speech frames to generate key frame sets, ordinary frame sets and sacrificial frame sets. Step 4: Cluster the groups hierarchically according to access point identifier, channel and link quality to form subgroups, and select representative receivers for each subgroup. For the unicast channel from the distribution node to each representative receiver, use the intercom available resource profile as a constraint, call the improved BBRv3 algorithm to generate a set of transmission parameters. The improved BBRv3 algorithm includes completion summary acquisition, delivery rate sampling, bottleneck bandwidth estimation, minimum propagation delay estimation, transmission rhythm and in-transit upper limit calculation, delay gatekeeping, roaming freeze and weak link isolation. Step 5: Prioritize and control the transmission of voice frame copies in the keyframe set based on the transmission parameter set, aggregate the completion summary, use majority completion judgment and directional unicast compensation to resend the voice frames in the missing keyframe set, and update the group completion status. Step 6: When the trigger queue delay exceeds the limit, the channel busy ratio suddenly increases, or roaming status occurs, switch the distribution mode based on the intercom available resource profile and perform a gray-scale rollback. Step 7: Record the version of the session metrics and transmission parameter set, establish a parameter template mapping library using scenario fingerprint matching, and store it in a versioned manner.

[0006] Optionally, step one specifically includes: When collecting wireless LAN link state indicators on the intercom terminal side, the received signal strength indication, signal-to-noise ratio, current physical layer modulation and coding scheme, media access control layer retransmission count and transmission queue length are recorded using the current access point identifier as an index. The channel busy ratio is also collected, which is a characterization value of the degree of busy occupancy of the wireless LAN carrier sensing results within the collection interval. When collecting intercom service status indicators on the intercom terminal side, record the intercom button status, speaker identifier, group identifier, voice frame sequence number and terminal transmission buffer depth, and use the voice frame sequence number as the primary key to perform time alignment between link status indicators and service status indicators. When receiving link-state indicators and service-state indicators reported by each intercom terminal on the distribution node side, the indicators are merged according to the terminal identifier and group identifier, and missing fields are filled with the same field value from the previous collection cycle to form a state vector containing terminal identifier, group identifier, access point identifier, voice frame sequence number and corresponding link-state indicators and service-state indicators.

[0007] Optionally, step two specifically includes: The state vectors are grouped according to terminal identifier and group identifier, and a sliding time window is constructed within each group according to the acquisition time sequence; Within the sliding time window, statistics are performed on the received signal strength indication, signal-to-noise ratio, physical layer modulation and coding level, media access control layer retransmission count, transmit queue length, and channel busy ratio. The statistical results of the sliding time window are subjected to exponential weighted smoothing. The carryable window is determined based on the exponentially weighted smoothed transmit queue length and the channel busy ratio, and the delay budget parameters are determined based on the smoothed medium access control layer retransmission count and the transmit queue length. The congestion level is determined based on the exponentially weighted smoothed transmit queue length, the increase in the number of retransmissions at the medium access control layer, and the channel busy ratio. Risk levels are determined based on changes in access point identifiers, increases in channel busy percentage, and increases in transmission queue length. The available resources for intercom are generated by merging the capacity window, latency budget parameters, congestion level, and risk level.

[0008] Optionally, step three specifically includes: The microphone input speech is encoded on the speaker's intercom terminal side to obtain continuous speech frames; The starting segment is determined based on the status of the intercom buttons; Within each speech frame, the sum of squared amplitudes of the speech frame sampling points is calculated as an energy feature, and speech frames whose energy features reach a preset energy threshold are marked as keyframe candidates. Extract the encoding reference frame identifier from the speech encoder output, and mark the speech frames with the encoding reference frame identifier as keyframe candidates; Speech frames marked as keyframe candidates are merged into a keyframe set, speech frames not marked as keyframe candidates and located after the start segment are merged into a normal frame set, and speech frames used to compensate for missing redundant repairs and speech frames to be sent that exceed the preset terminal transmission buffer depth threshold are merged into a sacrificial frame set. Label each audio frame in the keyframe set with the latest delivery timestamp, and store the latest delivery timestamp in association with the audio frame sequence number.

[0009] Optionally, the improved BBRv3 algorithm specifically includes: Based on access point identifier, channel, received signal strength indication, media access control layer retransmission count and channel busy ratio, intercom terminals under the same group identifier are hierarchically clustered to form subgroups. Within each subgroup, a representative receiver is selected based on link quality and the risk level of the available intercom resources. A unicast channel is established from the distribution node to each representative receiver, and each representative receiver generates and reports a completion summary according to a preset reporting cycle. For each unicast pipeline, a delivery rate sample is calculated based on two consecutive completion summaries, and a delivery rate sequence is formed. Within a preset bandwidth estimation window, bottleneck bandwidth estimation is performed on the delivery rate sequence. The bottleneck bandwidth estimation involves selecting the maximum delivery rate sample within the bandwidth estimation window as the bottleneck bandwidth and refreshing the bottleneck bandwidth between adjacent bandwidth estimation windows. The delay sample is determined by the distribution node recording the transmission time of the voice frame copy and combining it with the completion summary reporting timestamp, and the minimum delay sample is selected as the minimum propagation delay within the preset delay estimation window; Based on the available resources profile of the intercom system, the transmission rhythm and the upper limit in transit are calculated and basic transmission parameters are generated based on the bottleneck bandwidth and minimum propagation delay. The system performs delay gatekeeping, roaming freeze, and weak link isolation while updating basic transmission parameters. Delay gatekeeping involves using the difference between the current delay sample and the minimum propagation delay as the queuing delay and comparing it with the upper limit of the queuing delay allowed by the distribution node. When the queuing delay exceeds the limit by a preset number of times, the transmission pace is reduced and the in-transit limit is tightened. Roaming freeze involves limiting the transmission pace of the corresponding unicast channel to a preset minimum pace and setting the replication fan-out to a preset lower limit when the access point identifier changes or roaming status is collected. At the same time, the transmission of ordinary frame sets is restricted. Weak link isolation involves setting the replication fan-out to a preset lower limit when the bottleneck bandwidth of the corresponding unicast channel is lower than the preset proportion threshold corresponding to the bottleneck bandwidth statistics value in the group within a preset duration. Only the key frame set is configured with a key frame compensation budget and the transmission of ordinary frame sets is restricted. The transmission parameter set is generated by merging the updated transmission rhythm, in-transit limit, replication fan-out and keyframe compensation budget, and then the transmission parameter set is associated with and stored with the subgroup identifier, the representative receiver identifier and the corresponding unicast pipe identifier.

[0010] Optionally, step five specifically includes: Based on the transmission parameter set, the distribution node establishes a sending queue for the voice frame copies corresponding to the key frame set, the ordinary frame set, and the sacrificial frame set. The voice frame copies corresponding to the key frame set are placed in the high-priority queue, the voice frame copies corresponding to the ordinary frame set are placed in the medium-priority queue, and the voice frame copies corresponding to the sacrificial frame set are placed in the low-priority queue. Based on the transmission rhythm and the in-transit limit in the transmission parameter set, transmission control is performed on each unicast channel; Perform missing bitmap decoding on the completion summary reported by each representative receiver to obtain the set of missing voice frame numbers, and merge the set of missing voice frame numbers according to the subgroup identifier to form the group completion status. A majority completion determination is performed on the group completion status. The majority completion determination is to count the arrival status of the same voice frame number in each subgroup and compare it with a preset majority threshold. When the number of arrivals in the subgroup reaches the preset majority threshold, the voice frame corresponding to the voice frame number is determined to be completed. For voice frames that are incomplete and belong to the keyframe set, directional unicast compensation is performed. The directional unicast compensation is to determine the set of subgroup identifiers that are missing the current voice frame, and to resend the copy of the current voice frame only on the unicast channel of the corresponding subgroup. The number of resends is constrained by the keyframe compensation budget in the transmission parameter set. The group completion status is updated based on the majority completion determination results.

[0011] Optionally, step six specifically includes: Within each preset evaluation cycle, the distribution node reads the profile of available intercom resources and the group completion status, and extracts queue latency, channel busy ratio, access point identifier change and key frame set completion rate as handover criteria. When the queue delay reaches the preset queue delay threshold or the channel busy ratio reaches the preset busy ratio threshold, the distribution mode will be switched to the directional unicast compensation priority mode. When a change in the access point identifier or a high-risk risk level is detected in the intercom available resource profile, the distribution mode will be switched to roaming freeze mode. When the queue latency is lower than the preset recovery threshold and the channel busy percentage is lower than the preset busy percentage recovery threshold, and the key frame set completion rate reaches the preset completion rate threshold, the distribution mode will be switched to the hierarchical replication distribution mode. The keyframe set completion rate and queue latency are compared and verified within the evaluation period after the distribution mode switch. When the keyframe set completion rate decreases by a preset decrease threshold or the queue latency increases by a preset increase threshold compared to the time before the switch, grayscale rollback is performed.

[0012] Optionally, step seven specifically includes: At the end of each cluster intercom session, the distribution node collects session metrics and forms a session record. The session metrics include queue delay statistics, channel busy ratio statistics, media access control layer retransmission count statistics, access point identifier change count, key frame set completion rate, directional unicast compensation count, distribution mode switching count, and gray-scale rollback count. Generate a scenario fingerprint from the session records; Version the set of transmission parameters and the effective distribution mode to form a parameter template; A parameter template mapping library is established using scene fingerprint matching. Scene fingerprint matching involves calculating the similarity between the current session scene fingerprint and historical scene fingerprints in the mapping library, selecting the historical scene fingerprint with the highest similarity exceeding a preset similarity threshold as the matching result, and recording the parameter template corresponding to the matching result as the current session fingerprint. Session records, scenario fingerprints, parameter templates, and version numbers are written to versioned storage, and rollback version numbers and rollback trigger criteria are recorded for each parameter template.

[0013] The beneficial effects of this invention are: This invention generates a state vector by collecting link and service states from the intercom terminal and distribution node, and then generates an available intercom resource profile including congestion and risk levels based on sliding time window statistics and exponential weighted smoothing. This provides the distribution side with clear constraints on the carrying window and latency budget. Furthermore, it clusters groups into subgroups based on access point identifiers, channel and link quality, and selects representative receivers. For unicast channels, under resource profile constraints, it calls an improved BBRv3 algorithm to generate transmission rhythm and on-the-go upper limit, and superimposes latency gatekeeping, roaming freeze, and weak link isolation to actively suppress queue growth and roaming disturbances. Simultaneously, it layers voice frames according to start segments, energy peak segments, and coded reference frames. By forming a keyframe set and combining completion summary aggregation, majority completion judgment, and targeted unicast compensation, missing keyframes are retransmitted and the group completion status is updated. This ensures the timely delivery of key voice segments even under conditions such as a sudden increase in channel busy ratio, increased retransmissions, and access point switching, reducing end-to-end latency peaks and jitter, minimizing pops and interruptions, and preventing weak link terminals from causing the entire group to slow down. Finally, a parameter template mapping library is established through session metrics and parameter versioning, and scenario fingerprint matching, realizing a closed-loop evolution of cross-session reuse and gray-scale rollback. This significantly improves the stability, consistent intelligibility, and engineering maintainability of WLAN trunking intercom in complex environments, and has important application value and promotion significance. Attached Figure Description

[0014] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart of an adaptive transmission method for trunked intercom based on wireless local area network proposed in this invention; Figure 2 This is a schematic diagram of an adaptive transmission method for trunked intercom based on wireless local area network proposed in this invention; Figure 3This is a flowchart of the improved BBRv3 algorithm in a wireless local area network-based adaptive transmission method for trunked intercom proposed in this invention. Detailed Implementation

[0015] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0016] refer to Figures 1-3 An adaptive transmission method for trunked intercom based on wireless local area network includes the following steps: Step 1: Collect wireless LAN link state indicators and intercom service state indicators at the intercom terminal and distribution node, and form a state vector; Step 2: Based on the state vector, the sliding time window statistics and exponential weighted smoothing are used to calculate the carrying window and delay budget parameters, and generate a profile of available intercom resources including congestion level and risk level; Step 3: The speaker's speech frames are divided into start segments, energy peak segments and coding reference frames, and the latest delivery timestamp is marked for key speech frames to generate key frame sets, ordinary frame sets and sacrificial frame sets. Step 4: Cluster the groups hierarchically according to access point identifier, channel and link quality to form subgroups and select representative receivers for each subgroup. For the unicast channel from the distribution node to each representative receiver, use the intercom available resource profile as a constraint to call the improved BBRv3 algorithm to generate a set of transmission parameters. The improved BBRv3 algorithm includes completion summary acquisition, delivery rate sampling, bottleneck bandwidth estimation, minimum propagation delay estimation, transmission rhythm and in-transit upper limit calculation, delay gatekeeping, roaming freeze and weak link isolation. Step 5: Prioritize and control the transmission of voice frame copies in the keyframe set based on the transmission parameter set, aggregate the completion summary, use majority completion judgment and directional unicast compensation to resend the voice frames in the missing keyframe set, and update the group completion status. Step 6: When the trigger queue delay exceeds the limit, the channel busy ratio suddenly increases, or roaming status occurs, switch the distribution mode based on the intercom available resource profile and perform a gray-scale rollback. Step 7: Record the version of the session metrics and transmission parameter set, establish a parameter template mapping library using scenario fingerprint matching, and store it in a versioned manner.

[0017] In this embodiment, step one specifically includes: When collecting wireless LAN link state indicators on the intercom terminal side, the received signal strength indicator, signal-to-noise ratio, current physical layer modulation and coding scheme, media access control layer retransmission count, and transmission queue length are recorded using the current access point identifier as an index. The channel busy ratio is also collected. The channel busy ratio is a characterization value of the degree of busy occupancy of the wireless LAN carrier sensing results within the collection interval. The received signal strength indicator is a characterization value of the received signal power reported by the wireless network card. The signal-to-noise ratio is a characterization value of the ratio of received signal power to noise power. The physical layer modulation and coding scheme is an identifier of the combination of modulation method and coding rate used by the wireless LAN physical layer. The media access control layer retransmission count is the count of repeated transmission of the same data frame in the media access control layer. The transmission queue length is the number of data frames to be transmitted in the intercom terminal's transmission queue. When collecting intercom service status indicators on the intercom terminal side, the intercom button status, speaker identifier, group identifier, voice frame sequence number and terminal transmission buffer depth are recorded. The intercom button status is the identifier of the intercom button being pressed and released. The speaker identifier is the unique identifier of the current speaking terminal. The group identifier is the unique identifier of the intercom session in the same cluster. The voice frame sequence number is the sequence number obtained by incrementing the consecutive voice encoded frames according to the fixed voice frame duration. The terminal transmission buffer depth is the number of voice frames to be transmitted in the buffer on the terminal side. The link status indicators and service status indicators are time-aligned using the voice frame sequence number as the primary key. When receiving link-state indicators and service-state indicators reported by each intercom terminal on the distribution node side, the indicators are merged based on the terminal identifier and group identifier. The terminal identifier is a unique identifier for the intercom terminal. Missing fields are filled with the same field value from the previous collection cycle to form a state vector containing the terminal identifier, group identifier, access point identifier, voice frame sequence number, and corresponding link-state indicators and service-state indicators.

[0018] In this embodiment, step two specifically includes: The state vectors are grouped according to terminal identifier and group identifier, and a sliding time window is constructed within each group according to the acquisition time sequence; Within a sliding time window, statistics are performed on the received signal strength indication, signal-to-noise ratio, physical layer modulation and coding level, media access control layer retransmission count, transmit queue length, and channel busy ratio. For numerical indicators, the mean and maximum values ​​within the window are calculated, and for count indicators, the cumulative values ​​and growth rates within the window are calculated. Perform exponentially weighted smoothing on the statistical results of the sliding time window: the current window statistical value and the smoothed value of the previous window are weighted by a preset weight to obtain the current smoothed value. The preset weight is a constant between zero and one. The carryable window is determined based on the exponentially weighted smoothed transmission queue length and the channel busy ratio. The carryable window is the available transmission period used to constrain the transmission of voice frame copies. The delay budget parameters are determined based on the smoothed medium access control layer retransmission count and the transmission queue length. The delay budget parameters include the upper limit of the queuing delay allowed on the distribution node side and the end-to-end target delay. The congestion level is determined based on the exponentially weighted smoothed transmit queue length, the increase in the number of retransmissions at the medium access control layer, and the channel busy ratio. The increase in the number of retransmissions at the medium access control layer is the difference in the cumulative number of retransmissions within two adjacent sliding time windows. The congestion level consists of mild congestion, moderate congestion, and severe congestion. Queue threshold, retransmission growth threshold, and busy ratio threshold are preset, and low and high thresholds are set respectively. When at least two of the transmit queue length, the increase in the number of retransmissions at the medium access control layer, and the channel busy ratio reach the corresponding high threshold, it is determined to be severe congestion. When at least two of the indicators reach the corresponding low threshold but do not reach the high threshold, it is determined to be moderate congestion. All other cases are determined to be mild congestion. The risk level is determined based on changes in access point identifiers, increases in channel busy ratio, and increases in transmission queue length. Changes in access point identifiers refer to inconsistencies in access point identifiers between adjacent collection periods. Within a preset risk observation window, the number of access point identifier changes is counted, and the increases in channel busy ratio and transmission queue length are calculated. When the number of access point identifier changes reaches a preset threshold and the increases in channel busy ratio or transmission queue length reach a high threshold, it is considered a high-risk situation. When the increases in channel busy ratio or transmission queue length reach a low threshold but do not reach a high threshold, it is considered a warning situation. All other situations are considered normal. The available resources for intercom are generated by merging the capacity window, latency budget parameters, congestion level, and risk level.

[0019] In this embodiment, step three specifically includes: The microphone input voice is encoded on the speaker's intercom terminal side to obtain continuous voice frames. The voice frame is a data unit that is segmented according to a preset frame length and encoded. Each voice frame is assigned an incremental voice frame number. The starting segment is determined based on the state of the intercom button. The starting segment is the audio frame interval corresponding to a preset number of consecutive audio frames after the intercom button switches from the released state to the pressed state, and the audio frames in the starting segment are marked as key frame candidates. Within each speech frame, the sum of squared amplitudes of the speech frame sampling points is calculated as an energy feature, and speech frames whose energy features reach a preset energy threshold are marked as keyframe candidates. Extract the coded reference frame identifier from the speech encoder output. The coded reference frame identifier is the frame type identifier used by the speech encoder for inter-frame prediction or state update. Speech frames with coded reference frame identifiers are marked as keyframe candidates. Speech frames marked as keyframe candidates are merged into a keyframe set, speech frames not marked as keyframe candidates and located after the start segment are merged into a normal frame set, and redundant repair speech frames used to compensate for missing speech frames and speech frames to be sent that exceed the preset terminal transmission buffer depth threshold are merged into a sacrificial frame set, wherein the redundant repair speech frames are speech frame copies generated to compensate for missing speech frames. Mark each audio frame in the keyframe set with the latest delivery timestamp. The latest delivery timestamp is the time point obtained by adding the preset keyframe delay threshold to the audio frame generation time, and the latest delivery timestamp is stored in association with the audio frame sequence number.

[0020] In this embodiment, the improved BBRv3 algorithm specifically includes: Based on access point identifier, channel, received signal strength indicator, media access control layer retransmission count, and channel busy ratio, intercom terminals under the same group identifier are hierarchically clustered to form subgroups. The hierarchical clustering first uses access point identifier and channel as the first-level division condition to obtain a candidate set, and then performs second-level clustering within the candidate set according to the preset grading thresholds of received signal strength indicator, media access control layer retransmission count, and channel busy ratio to obtain subgroups. Within each subgroup, representative receivers are selected based on link quality and risk level profiles of available intercom resources. The selection process is as follows: representative scores are calculated for candidate terminals, and the highest score is selected as the representative receiver. The representative score is obtained by weighting the received signal strength index score, the reverse score of the number of retransmissions at the medium access control layer, the reverse score of the channel busy ratio, and the access point identifier stability score. The access point identifier stability score is determined by comparing the number of access point identifier changes within a preset observation period with a preset threshold. A unicast channel is established from the distribution node to each representative receiver. Each representative receiver generates and reports a completion summary according to a preset reporting period. The completion summary includes the sequence number of the highest continuously arriving voice frame, the missing bitmap, and the reporting timestamp. The missing bitmap is a data structure obtained by setting and encoding the sequence number of voice frames that have not arrived within a preset backtracking window with the highest voice frame sequence number as a reference. For each unicast pipeline, a delivery rate sample is calculated based on two adjacent completion summaries and a delivery rate sequence is formed: the set of newly arrived voice frame numbers is determined based on the highest voice frame number and the missing bitmap of the two completion summaries, and the delivery rate sample is obtained based on the amount of data corresponding to the voice frame number set under the preset frame length and preset coding bit rate and the time interval between the two reporting timestamps. Within a preset bandwidth estimation window, bottleneck bandwidth estimation is performed on the delivery rate sequence. Bottleneck bandwidth estimation involves selecting the maximum delivery rate sample within the bandwidth estimation window as the bottleneck bandwidth and refreshing the bottleneck bandwidth between adjacent bandwidth estimation windows. The distribution node records the transmission time of the voice frame copy and combines it with the completion summary reporting timestamp to determine the delay sample. The delay sample is the time difference obtained by matching the transmission time of the newly arrived voice frame copy in the set of voice frame sequence numbers with the corresponding completion summary reporting timestamp in the unicast channel, and the minimum delay sample is selected as the minimum propagation delay within the preset delay estimation window. Based on the available intercom resource profile as a constraint, the transmission rhythm and the in-transit limit are calculated based on the bottleneck bandwidth and minimum propagation delay, and basic transmission parameters are generated. The transmission rhythm is the constraint on the amount of data that can be transmitted per unit time, and the in-transit limit is the upper limit of the amount of data that is allowed to be in the incomplete confirmation state. When the congestion level is severe, the transmission rhythm and the in-transit limit are respectively limited to the corresponding preset upper limit range, and when the congestion level is mild, they are respectively relaxed to the corresponding preset relaxation range. The system performs delay gatekeeping, roaming freeze, and weak link isolation while updating basic transmission parameters. Delay gatekeeping involves using the difference between the current delay sample and the minimum propagation delay as the queuing delay and comparing it with the upper limit of the queuing delay allowed by the distribution node. When the queuing delay exceeds the limit by a preset number of times, the transmission pace is reduced and the in-transit limit is tightened. Roaming freeze involves limiting the transmission pace of the corresponding unicast channel to a preset minimum pace and setting the replication fan-out to a preset lower limit when the access point identifier changes or roaming status is collected. At the same time, the transmission of ordinary frame sets is restricted. Weak link isolation involves setting the replication fan-out to a preset lower limit when the bottleneck bandwidth of the corresponding unicast channel is lower than the preset proportion threshold corresponding to the bottleneck bandwidth statistics value in the group within a preset duration. Only the key frame set is configured with a key frame compensation budget and the transmission of ordinary frame sets is restricted. The transmission parameter set is generated by merging the updated transmission rhythm, in-transit limit, replication fan-out and keyframe compensation budget, and then the transmission parameter set is associated with and stored with the subgroup identifier, the representative receiver identifier and the corresponding unicast pipe identifier.

[0021] In this embodiment, the improved BBRv3 algorithm is similar to the original BBRv3 algorithm in that it constructs a delivery rate sampling sequence with completion feedback, takes the maximum delivery rate as the bottleneck bandwidth within the preset bandwidth estimation window, takes the minimum delay sample as the minimum propagation delay within the preset delay estimation window, and then calculates the transmission rhythm and the upper limit in transit. By adjusting the transmission rhythm and the upper limit in transit, adaptive control of the link transmission load is achieved. Compared to the original BBRv3 algorithm, the improved BBRv3 algorithm differs in the following ways: First, it introduces a hierarchical structure of "subgroup—representative receiver—unicast channel," transforming the one-to-many distribution of the group into several controllable unicast channels, and using the available intercom resource profile as a constraint to hierarchically limit the transmission rhythm and the on-the-way limit; Second, it adds a latency gatekeeping mechanism, triggering a reduction in the transmission rhythm and tightening the on-the-way limit by counting queue latency exceeding the limit, thus suppressing queue accumulation in the wireless LAN; Third, it adds a roaming freeze mechanism, shrinking the replicated fan-out and restricting the transmission of ordinary frame sets when the access point identifier changes or a roaming state occurs; Finally, it adds a weak link isolation mechanism, ensuring only the key frame set is guaranteed and configuring a key frame compensation budget when the bottleneck bandwidth of the unicast channel is continuously lower than the group statistical threshold; Through improvements, the peak queuing latency can be effectively reduced in scenarios such as sudden increases in the busy ratio of wireless LAN channels, increased retransmissions, and roaming handover. This also reduces voice pops and interruptions and prevents weak link terminals from causing the entire group to slow down. At the same time, key frame sets are prioritized and protected. Combined with targeted compensation of completion summaries, missing key frames can be quickly repaired, improving group consistency and intelligibility and end-to-end latency stability.

[0022] In this embodiment, step five specifically includes: Based on the transmission parameter set, the distribution node establishes a sending queue for the voice frame copies corresponding to the key frame set, the ordinary frame set, and the sacrificial frame set. The voice frame copies corresponding to the key frame set are placed in the high-priority queue, the voice frame copies corresponding to the ordinary frame set are placed in the medium-priority queue, and the voice frame copies corresponding to the sacrificial frame set are placed in the low-priority queue. Based on the transmission rhythm and the in-transit limit in the transmission parameter set, transmission control is performed on each unicast channel. The transmission control limits the amount of data that can be transmitted within each preset transmission cycle to no more than the amount of data corresponding to the transmission rhythm, and limits the amount of transmitted data that is in the incomplete confirmation state to no more than the in-transit limit. Perform missing bitmap decoding on the completion summary reported by each representative receiver to obtain the set of missing voice frame numbers, and merge the set of missing voice frame numbers according to the subgroup identifier to form the group completion status. A majority completion determination is performed on the group completion status. The majority completion determination is to count the arrival status of the same voice frame number in each subgroup and compare it with a preset majority threshold. When the number of arrivals in the subgroup reaches the preset majority threshold, the voice frame corresponding to the voice frame number is determined to be completed. For voice frames that are incomplete and belong to the keyframe set, targeted unicast compensation is performed. Targeted unicast compensation is to determine the set of subgroup identifiers that are missing the current voice frame and to resend the copy of the current voice frame only on the unicast pipeline of the corresponding subgroup. The number of resends is constrained by the keyframe compensation budget in the transmission parameter set. The group completion status is updated based on the majority completion determination results.

[0023] In this embodiment, step six specifically includes: Within each preset evaluation period, the distribution node reads the profile of available intercom resources and the group completion status, and extracts queue delay, channel busy ratio, access point identifier change and key frame set completion rate as switching criteria. The key frame set completion rate is the ratio of the number of key frames that are determined to be completed by the majority to the total number of key frames within the evaluation period. When the queue delay reaches the preset queue delay threshold or the channel busy ratio reaches the preset busy ratio threshold, the distribution mode will be switched to the directional unicast compensation priority mode. The directional unicast compensation priority mode keeps the subgroup and unicast pipeline unchanged and sets the replication fan-out to the preset lower limit. At the same time, it only performs resending on the key frame set and restricts the sending of the ordinary frame set. When a change in the access point identifier or a high-risk risk level is detected in the intercom available resource profile, the distribution mode will be switched to roaming freeze mode. Roaming freeze mode limits the sending rhythm to a preset minimum rhythm and sets the copy fan-out to a preset lower limit within a preset freeze duration. At the same time, only the voice frame copy corresponding to the key frame set is sent and the sending of the ordinary frame set is suspended. When the queue delay is lower than the preset recovery threshold and the channel busy percentage is lower than the preset busy percentage recovery threshold, and the key frame set completion rate reaches the preset completion rate threshold, the distribution mode will be switched to the hierarchical copy distribution mode. The hierarchical copy distribution mode sends voice frame copies to the representative receivers of each subgroup in parallel according to the subgroup division and allows the ordinary frame set to resume transmission. The key frame set completion rate and queue latency are compared and verified within the evaluation period after the distribution mode switch. When the key frame set completion rate decreases by a preset decrease threshold or the queue latency increases by a preset increase threshold compared to the previous switch, a gray rollback is performed. The gray rollback is to roll back to the distribution mode and transmission parameter set version before the switch, and the rollback result and the corresponding trigger criteria are recorded as session metrics.

[0024] In this embodiment, step seven specifically includes: At the end of each cluster intercom session, the distribution node collects session metrics and forms a session record. The session metrics include queue latency statistics, channel busy ratio statistics, media access control layer retransmission count statistics, access point identifier change count, key frame set completion rate, directional unicast compensation count, distribution mode switching count, and gray-scale rollback count. A scenario fingerprint is generated from the session records. The scenario fingerprint is obtained by concatenating the access point identifier sequence, channel sequence, congestion level sequence and risk level sequence in the order of collection time. The scenario fingerprint is then stored in association with the group identifier and the session start and end time. Versioning the transmission parameter set and the effective distribution mode involves assigning a version number to each updated transmission parameter set and distribution mode, and recording the effective time window, applicable subgroup range, and representative receiver range to form a parameter template. A parameter template mapping library is established using scene fingerprint matching. Scene fingerprint matching involves calculating the similarity between the current session scene fingerprint and historical scene fingerprints in the mapping library, selecting the historical scene fingerprint with the highest similarity exceeding a preset similarity threshold as the matching result, and recording the parameter template corresponding to the matching result as the current session fingerprint. Session records, scenario fingerprints, parameter templates, and version numbers are written to versioned storage, and rollback version numbers and rollback trigger criteria are recorded for each parameter template.

[0025] Example 1: To verify the feasibility of this invention in practice, it was applied to a WLAN trunking intercom dispatch system in a large chemical industrial park. The park contains 6 outdoor access points and 2 indoor access points. Workers wearing 48 intercom terminals moved between the equipment area, pipe corridors, and pump rooms. Simultaneously, there was interference from mobile Wi-Fi hotspots, multipath reflections from metal pipe corridors, and personnel roaming across access points. The intercom service generated continuous voice frames of 20 milliseconds each, using a 24kbps voice encoding bit rate. Distribution nodes were deployed in a server room at the park's edge, accessing each access point via wired backhaul. To simulate real workload, multiple people took turns speaking, interrupting, and making emergency calls during the test. Concurrency and movement frequencies were deliberately increased during peak shift changes, causing significant fluctuations in channel busy ratio and retransmission rate.

[0026] Upon system deployment, terminals and distribution nodes continuously collect wireless LAN link state and intercom service state data to form state vectors. The link state data collection interval is set to 100 milliseconds, recording access point identifier, received signal strength indication, signal-to-noise ratio, physical layer modulation and coding scheme, media access control layer retransmission count, and transmission queue length. Channel busy percentage is also collected as a representation of carrier eavesdropping busy occupancy. The service state data synchronously records intercom button status, speaker identifier, group identifier, voice frame sequence number, and terminal transmission buffer depth, aligning the link state and service state with the voice frame sequence number. Distribution nodes merge and report data according to terminal identifier and group identifier, filling missing fields with the same fields from the previous collection cycle to ensure state vector continuity. Subsequently, the distribution nodes perform sliding window statistics and exponentially weighted smoothing on the state vectors: the sliding window length is 1.0 second, the step size is 100 milliseconds, and the exponentially weighted smoothing weight is 0.30. The available window is determined based on the smoothed transmission queue length and channel busy ratio. The upper limit of the allowed queuing delay on the distribution node side is set to 35 milliseconds, and the end-to-end target delay is set to 150 milliseconds. Based on this, a profile of available intercom resources including congestion level and risk level is formed. The congestion level is determined using three thresholds: the queue threshold is set to 18 frames for the low threshold and 30 frames for the high threshold; the retransmission growth threshold is set to 12 times per second for the low threshold and 25 times per second for the high threshold; and the busy ratio threshold is set to 0.55 for the low threshold and 0.75 for the high threshold. When at least two of the three criteria reach the high threshold, it is considered severe congestion; when at least two criteria reach the low threshold but not the high threshold, it is considered moderate congestion; and the rest are considered mild congestion. The risk level uses a 2.0-second risk observation window. The threshold for the number of times the access point identifier changes is set to 2. The low threshold for busy ratio growth and the high threshold for queue length growth are set to 0.08 / 0.15 and +6 frames / +12 frames, respectively. When the number of access point changes reaches the threshold and the busy ratio growth or queue length growth reaches the high threshold, it is judged as high risk; when the busy ratio growth or queue length growth reaches the low threshold but does not reach the high threshold, it is judged as a warning; otherwise, it is normal.

[0027] The speech frame hierarchy uses a joint annotation of "start segment + energy peak segment + coded reference frame". The start segment is the first 300 milliseconds after the button is switched from release to press, i.e., the first 15 consecutive frames are directly used as key frame candidates. The energy threshold is set to 1.8 times the average energy of the most recent 5 seconds, and speech frames exceeding the threshold are used as key frame candidates. The coded reference frame identifier is extracted from the encoder output, and if the reference frame identifier appears, the frame is used as a key frame candidate. Key frame candidates are merged into a key frame set, and the remaining frames are merged into a normal frame set. Redundant repair speech frames used to compensate for missing frames and speech frames to be sent with a terminal transmission buffer depth of more than 40 frames are merged into a sacrificial frame set. The latest delivery timestamp of the key frame is marked as "generation time + 90 milliseconds" and stored in association with the speech frame sequence number, which facilitates boundary control during subsequent priority scheduling and targeted compensation.

[0028] The distribution node first divides the group terminals according to access point identifier and channel, and then performs second-level clustering based on received signal strength indication, retransmission count, and busy percentage to form subgroups. In this test, five subgroups were stably formed, with one representative receiver selected from each subgroup. The representative receiver's score uses a weighted tier system: received signal strength indication score weighted at 0.40, retransmission count weighted at 0.30, busy percentage weighted at 0.20, and access point identifier stability weighted at 0.10. Stability is determined by comparing the number of access point changes within 10 seconds with a threshold. A unicast channel is established between the distribution node and each representative receiver. Each representative receiver reports a completion summary every 80 milliseconds, including the sequence number of the highest consecutively arriving voice frame, missing bitmaps within a 250-millisecond backtracking window, and the reporting timestamp. The distribution node calculates the delivery rate sample based on two consecutive completion summaries. The bandwidth estimation window is 1.5 seconds, and the maximum sample is taken as the bottleneck bandwidth. The latency estimation window is 20 seconds, and the minimum latency sample is taken as the minimum propagation latency. Under the constraints of the available intercom resource profile, the transmission rhythm and the in-transit upper limit are calculated, and latency gatekeeping, roaming freeze, and weak link isolation are superimposed: when the queuing latency exceeds the limit three times, the transmission rhythm is reduced and the in-transit upper limit is tightened; when the access point identifier changes or the risk level is high risk, roaming freeze is entered, the transmission rhythm is limited to the minimum rhythm, and the replication fan-out lower limit is set to 1, while the transmission of ordinary frame sets is restricted; when the bottleneck bandwidth of a certain unicast channel is lower than 0.60 of the median bottleneck bandwidth in the group for 2 consecutive seconds, weak link isolation is triggered, the replication fan-out is set to the lower limit, and only the key frame set is configured with a key frame compensation budget, and the ordinary frame set is transmitted at a reduced frequency in this channel. This forms a transmission parameter set for subsequent scheduling.

[0029] During the transmission scheduling phase, the distribution node establishes high, medium, and low priority queues for the voice frame replicas corresponding to the keyframe set, ordinary frame set, and sacrificial frame set. Based on the transmission parameter set, it performs transmission control on each unicast channel to ensure that the transmission volume per 100 milliseconds does not exceed the quota corresponding to the transmission rhythm, and that the amount of data with incomplete confirmation does not exceed the in-transit limit. After completion summary aggregation, the distribution node decodes the missing bitmap to obtain the missing voice frame sequence number set and merges them by subgroup. A majority completion criterion is used, with a majority threshold of 3 subgroups. If a keyframe is completed in a majority subgroup, it is not retransmitted. If a keyframe is missing in a minority subgroup, targeted unicast compensation is performed only on the unicast channel corresponding to the missing subgroup. The maximum number of retransmissions per frame is set to 1, and is constrained by the keyframe compensation budget. During the mode switching phase, when queue latency exceeds limits, busy ratio suddenly increases, or roaming status is triggered, the system switches the distribution mode based on resource profiles: if the busy ratio reaches 0.75 or queue latency exceeds 35 milliseconds, it enters "directed unicast compensation priority," replicating fanout and shrinking, and restricting ordinary frames; if the risk level is high risk, it enters "roaming freeze"; when queue latency falls below 20 milliseconds and the key frame set completion rate recovers to above 99.0%, it returns to the hierarchical replication distribution mode. If the key frame completion rate drops by more than 0.8 percentage points or the queue latency increases by more than 10 milliseconds after the switch, a gray-scale rollback is triggered, and the system reverts to the previous version's transmission parameter set and distribution mode. After the session ends, session metrics and version are recorded, and a scenario fingerprint is generated by concatenating "access point identifier sequence + channel sequence + congestion level sequence + risk level sequence." After matching historical fingerprints, the fingerprints are written into the parameter template mapping library, and the recommended template can be directly loaded the next time a similar area is entered.

[0030] To demonstrate the beneficial effects, two comparative schemes were tested simultaneously in the same campus and with the same call script: Scheme 1 was "multicast distribution + fixed bitrate + fixed retransmission", and Scheme 2 was "unicast for all users + fixed sending rate limit + retransmission based on packet loss rate". Each scheme was run for 2 hours, and metrics such as end-to-end latency, keyframe completion rate, number of audio pops, and resource usage were statistically analyzed, resulting in the comprehensive data table below: Table 1 Comprehensive Comparison Data Table

[0031] As shown in Table 1, although the multicast scheme has low transmission overhead, it lacks controllable acknowledgment and targeted repair in wireless environments. The keyframe completion rate is only 96.1%, the keyframe loss rate during roaming reaches 6.9%, the end-to-end latency P95 is as high as 286 milliseconds, and the pop / interruption occurs 1.8 times / minute, indicating that the intelligibility of intercoms drops significantly at critical moments. The all-member unicast scheme reduces some of the uncontrollable multicast issues by using a fixed rate cap, and the keyframe completion rate is improved to 97.4%. However, due to the fan-out of replication, the burst load is greater, the average channel busy ratio rises to 0.69, the queue length P95 reaches 46 frames, the retransmission growth increases to 28 times / second, and the maximum end-to-end latency still exceeds 500 milliseconds, indicating that the "fixed cap" cannot effectively suppress queue expansion during interference bursts and mobile handover.

[0032] The improvements of this invention are more directly reflected in the data. First, the resource profiling explicitly defines the carrying window and latency budget, and by overlaying the improved BBRv3 transmission rhythm and on-the-go upper limit, the queue length P95 is reduced from 41 / 46 frames to 27 frames, and the end-to-end latency P95 is reduced to 158 milliseconds, with the maximum value controlled within 322 milliseconds, indicating that queue growth is continuously suppressed. Second, the keyframe set layering and majority completion judgment combined with targeted unicast compensation achieve a keyframe completion rate of 99.3%, reduce the keyframe loss rate during roaming to 1.4%, and reduce popping / interruption to 0.4 times / minute. Although the number of targeted compensation triggers increases, the compensation is limited to the missing subgroup and keyframe range, without causing an increase in the busy ratio. On the contrary, the busy ratio is slightly lower than the two comparison schemes, indicating that the compensation strategy is "precise repair" rather than "indiscriminate retransmission". Finally, the average number of gray-scale rollbacks was 3 times per hour, mainly occurring in the extreme interference area during shift change peaks. After rollback, the keyframe completion rate could recover to above the threshold within one evaluation cycle, indicating that versioning and rollback mechanisms can avoid continuous degradation caused by strategy switching.

[0033] This embodiment demonstrates that the present invention can transform the "replication and retransmission budget of intercom distribution" into a stable transmission rhythm and a controllable in-transit upper limit in the complex environment of real campus WLAN. Through key frame set priority guarantee, completion summary aggregation and targeted unicast compensation, it significantly reduces end-to-end latency peak and jitter in typical pain scenarios such as sudden increase in channel busy ratio, increase in retransmission and roaming handover, reduces popping and interruption, and improves group consistency and intelligibility. At the same time, it realizes cross-session reuse and continuous stable operation through scene fingerprint and parameter template mapping library, meeting the requirements of "controllable, recoverable and evolvable" in engineering deployment.

[0034] The above are merely preferred embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A wireless local area network based cluster intercom adaptive transmission method, characterized in that, Includes the following steps: Step 1: Collect wireless LAN link state indicators and intercom service state indicators at the intercom terminal and distribution node, and form a state vector; Step 2: Based on the state vector, the sliding time window statistics and exponential weighted smoothing are used to calculate the carrying window and delay budget parameters, and generate a profile of available intercom resources including congestion level and risk level; Step 3: The speaker's speech frames are divided into start segments, energy peak segments and coding reference frames, and the latest delivery timestamp is marked for key speech frames to generate key frame sets, ordinary frame sets and sacrificial frame sets. Step 4: Cluster the groups hierarchically according to access point identifier, channel and link quality to form subgroups, and select representative receivers for each subgroup. For the unicast channel from the distribution node to each representative receiver, use the intercom available resource profile as a constraint, call the improved BBRv3 algorithm to generate a set of transmission parameters. The improved BBRv3 algorithm includes completion summary acquisition, delivery rate sampling, bottleneck bandwidth estimation, minimum propagation delay estimation, transmission rhythm and in-transit upper limit calculation, delay gatekeeping, roaming freeze and weak link isolation. Step 5: Prioritize and control the transmission of voice frame copies in the keyframe set based on the transmission parameter set, aggregate the completion summary, use majority completion judgment and directional unicast compensation to resend the voice frames in the missing keyframe set, and update the group completion status. Step 6: When the trigger queue delay exceeds the limit, the channel busy ratio suddenly increases, or roaming status occurs, switch the distribution mode based on the intercom available resource profile and perform a gray-scale rollback. Step 7: Record the version of the session metrics and transmission parameter set, establish a parameter template mapping library using scenario fingerprint matching, and store it in a versioned manner.

2. The wireless LAN-based group intercom adaptive transmission method according to claim 1, characterized by, Step one specifically includes: When collecting wireless LAN link state indicators on the intercom terminal side, the received signal strength indication, signal-to-noise ratio, current physical layer modulation and coding scheme, media access control layer retransmission count and transmission queue length are recorded using the current access point identifier as an index. The channel busy ratio is also collected, which is a characterization value of the degree of busy occupancy of the wireless LAN carrier sensing results within the collection interval. When collecting intercom service status indicators on the intercom terminal side, record the intercom button status, speaker identifier, group identifier, voice frame sequence number and terminal transmission buffer depth, and use the voice frame sequence number as the primary key to perform time alignment between link status indicators and service status indicators. When receiving link-state indicators and service-state indicators reported by each intercom terminal on the distribution node side, the indicators are merged according to the terminal identifier and group identifier, and missing fields are filled with the same field value from the previous collection cycle to form a state vector containing terminal identifier, group identifier, access point identifier, voice frame sequence number and corresponding link-state indicators and service-state indicators.

3. The wireless LAN-based group intercom adaptive transmission method according to claim 1, wherein Step two specifically includes: The state vectors are grouped according to terminal identifier and group identifier, and a sliding time window is constructed within each group according to the acquisition time sequence; Within the sliding time window, statistics are performed on the received signal strength indication, signal-to-noise ratio, physical layer modulation and coding level, media access control layer retransmission count, transmit queue length, and channel busy ratio. The statistical results of the sliding time window are subjected to exponential weighted smoothing. The carryable window is determined based on the exponentially weighted smoothed transmit queue length and the channel busy ratio, and the delay budget parameters are determined based on the smoothed medium access control layer retransmission count and the transmit queue length. The congestion level is determined based on the exponentially weighted smoothed transmit queue length, the increase in the number of retransmissions at the medium access control layer, and the channel busy ratio. Risk levels are determined based on changes in access point identifiers, increases in channel busy percentage, and increases in transmission queue length. The available resources for intercom are generated by merging the capacity window, latency budget parameters, congestion level, and risk level.

4. The wireless LAN-based group intercom adaptive transmission method according to claim 1, wherein Step three specifically includes: The microphone input speech is encoded on the speaker's intercom terminal side to obtain continuous speech frames; The starting segment is determined based on the status of the intercom buttons; Within each speech frame, the sum of squared amplitudes of the speech frame sampling points is calculated as an energy feature, and speech frames whose energy features reach a preset energy threshold are marked as keyframe candidates. Extract the encoding reference frame identifier from the speech encoder output, and mark the speech frames with the encoding reference frame identifier as keyframe candidates; Speech frames marked as keyframe candidates are merged into a keyframe set, speech frames not marked as keyframe candidates and located after the start segment are merged into a normal frame set, and speech frames used to compensate for missing redundant repairs and speech frames to be sent that exceed the preset terminal transmission buffer depth threshold are merged into a sacrificial frame set. Label each audio frame in the keyframe set with the latest delivery timestamp, and store the latest delivery timestamp in association with the audio frame sequence number.

5. The wireless LAN-based group intercom adaptive transmission method according to claim 1, wherein The improved BBRv3 algorithm specifically includes: Based on access point identifier, channel, received signal strength indication, media access control layer retransmission count and channel busy ratio, intercom terminals under the same group identifier are hierarchically clustered to form subgroups. Within each subgroup, a representative receiver is selected based on link quality and the risk level of the available intercom resources. A unicast channel is established from the distribution node to each representative receiver, and each representative receiver generates and reports a completion summary according to a preset reporting cycle. For each unicast pipeline, a delivery rate sample is calculated based on two consecutive completion summaries, and a delivery rate sequence is formed. Within a preset bandwidth estimation window, bottleneck bandwidth estimation is performed on the delivery rate sequence. The bottleneck bandwidth estimation involves selecting the maximum delivery rate sample within the bandwidth estimation window as the bottleneck bandwidth and refreshing the bottleneck bandwidth between adjacent bandwidth estimation windows. The delay sample is determined by the distribution node recording the transmission time of the voice frame copy and combining it with the completion summary reporting timestamp, and the minimum delay sample is selected as the minimum propagation delay within the preset delay estimation window; Based on the available resources profile of the intercom system, the transmission rhythm and the upper limit in transit are calculated and basic transmission parameters are generated based on the bottleneck bandwidth and minimum propagation delay. The system performs delay gatekeeping, roaming freeze, and weak link isolation while updating basic transmission parameters. Delay gatekeeping involves using the difference between the current delay sample and the minimum propagation delay as the queuing delay and comparing it with the upper limit of the queuing delay allowed by the distribution node. When the queuing delay exceeds the limit by a preset number of times, the transmission pace is reduced and the in-transit limit is tightened. Roaming freeze involves limiting the transmission pace of the corresponding unicast channel to a preset minimum pace and setting the replication fan-out to a preset lower limit when the access point identifier changes or roaming status is collected. At the same time, the transmission of ordinary frame sets is restricted. Weak link isolation involves setting the replication fan-out to a preset lower limit when the bottleneck bandwidth of the corresponding unicast channel is lower than the preset proportion threshold corresponding to the bottleneck bandwidth statistics value in the group within a preset duration. Only the key frame set is configured with a key frame compensation budget and the transmission of ordinary frame sets is restricted. The transmission parameter set is generated by merging the updated transmission rhythm, in-transit limit, replication fan-out and keyframe compensation budget, and then the transmission parameter set is associated with and stored with the subgroup identifier, the representative receiver identifier and the corresponding unicast pipe identifier.

6. The wireless LAN-based group intercom adaptive transmission method according to claim 1, wherein Step five specifically includes: Based on the transmission parameter set, the distribution node establishes a sending queue for the voice frame copies corresponding to the key frame set, the ordinary frame set, and the sacrificial frame set. The voice frame copies corresponding to the key frame set are placed in the high-priority queue, the voice frame copies corresponding to the ordinary frame set are placed in the medium-priority queue, and the voice frame copies corresponding to the sacrificial frame set are placed in the low-priority queue. Based on the transmission rhythm and the in-transit limit in the transmission parameter set, transmission control is performed on each unicast channel; Perform missing bitmap decoding on the completion summary reported by each representative receiver to obtain the set of missing voice frame numbers, and merge the set of missing voice frame numbers according to the subgroup identifier to form the group completion status. A majority completion determination is performed on the group completion status. The majority completion determination is to count the arrival status of the same voice frame number in each subgroup and compare it with a preset majority threshold. When the number of arrivals in the subgroup reaches the preset majority threshold, the voice frame corresponding to the voice frame number is determined to be completed. For voice frames that are incomplete and belong to the keyframe set, directional unicast compensation is performed. The directional unicast compensation is to determine the set of subgroup identifiers that are missing the current voice frame, and to resend the copy of the current voice frame only on the unicast channel of the corresponding subgroup. The number of resends is constrained by the keyframe compensation budget in the transmission parameter set. The group completion status is updated based on the majority completion determination results.

7. The wireless LAN-based group intercom adaptive transmission method according to claim 1, wherein Step six specifically includes: Within each preset evaluation cycle, the distribution node reads the profile of available intercom resources and the group completion status, and extracts queue latency, channel busy ratio, access point identifier change and key frame set completion rate as handover criteria. When the queue delay reaches the preset queue delay threshold or the channel busy ratio reaches the preset busy ratio threshold, the distribution mode will be switched to the directional unicast compensation priority mode. When a change in the access point identifier or a high-risk risk level is detected in the intercom available resource profile, the distribution mode will be switched to roaming freeze mode. When the queue latency is lower than the preset recovery threshold and the channel busy percentage is lower than the preset busy percentage recovery threshold, and the key frame set completion rate reaches the preset completion rate threshold, the distribution mode will be switched to the hierarchical replication distribution mode. The keyframe set completion rate and queue latency are compared and verified within the evaluation period after the distribution mode switch. When the keyframe set completion rate decreases by a preset decrease threshold or the queue latency increases by a preset increase threshold compared to the time before the switch, grayscale rollback is performed.

8. The wireless local area network based group intercom adaptive transmission method according to claim 1, wherein, Step seven specifically includes: At the end of each cluster intercom session, the distribution node collects session metrics and forms a session record. The session metrics include queue delay statistics, channel busy ratio statistics, media access control layer retransmission count statistics, access point identifier change count, key frame set completion rate, directional unicast compensation count, distribution mode switching count, and gray-scale rollback count. Generate a scenario fingerprint from the session records; Version the set of transmission parameters and the effective distribution mode to form a parameter template; A parameter template mapping library is established by scene fingerprint matching. The scene fingerprint matching is to calculate the similarity between the current session scene fingerprint and the historical scene fingerprint in the mapping library, select the historical scene fingerprint with the highest similarity and exceeding the preset similarity threshold as the matching result, and record the parameter template corresponding to the matching result as the current one. Session records, scenario fingerprints, parameter templates, and version numbers are written to versioned storage, and rollback version numbers and rollback trigger criteria are recorded for each parameter template.

Images