Emergency rescue video call system and method for people trapped in a malfunctioning elevator

Abstract

This invention relates to the field of edge-cloud collaboration technology, specifically to an emergency rescue video call system and method for people trapped in malfunctioning elevators. The system includes: a state-aware coding module that calculates the signal-to-noise ratio slope to compress macroblocks and construct a video stream; a keyframe loss assessment module that calculates packet loss density to obtain truncation loss rate; an asymmetric error correction scheduling module that stretches the step size to construct a hybrid stream; a delay mapping retransmission module that monitors delay parameters to obtain detection results; and a cross-modal time-domain synchronization module that stretches audio and shifts and calibrates time markers. In this invention, spatial redundancy compression is performed based on dynamic image and signal-to-noise ratio patterns to adapt to physical channel attenuation. The load missing density is analyzed and the step distance is dynamically stretched to perform asymmetric error correction scheduling. Combined with the delay detection results, periodic peak search and non-modulation extension are performed on the waveform to drive the retransmission packet marker shift calibration. In a weak network transmission environment, the audio-visual data is kept aligned in the time domain and the screen stuttering and disconnection faults are eliminated.

Description

Technical Field

[0001] This invention relates to the field of edge-cloud collaboration technology, and in particular to an emergency rescue video call system and method for people trapped in malfunctioning elevators. Background Technology

[0002] The field of edge-cloud collaboration technology mainly involves the architecture of computing task sharing and data interaction between terminal devices distributed at the network edge and centralized cloud servers. The core aspects of this field include continuous collection of environmental status data from edge terminals and centralized scheduling and long-term storage in the cloud data center. It uses transport layer protocols to upload discrete data packets acquired by terminal sensors to the cloud server cluster in real time. The cloud server then distributes the generated downlink control signals and configuration parameters to various distributed physical terminals, establishing a distributed network where hardware computing resources and network transmission bandwidth work together. In contrast, the traditional emergency rescue video call system for trapped elevators refers to a security communication device that establishes a two-way audio and video transmission path between trapped passengers and the building's security monitoring center when an elevator stops due to a sudden mechanical or electrical malfunction and passengers are trapped in the car. It primarily addresses the technical aspects of acquiring internal images and voice recordings and transmitting them to external management personnel in emergency situations within the enclosed metal structure of the elevator car. Traditional systems typically install a fixed-focus miniature camera, a dynamic microphone, and a cone speaker above the elevator car's control panel, using a coaxial cable wound inside the elevator's traveling cable and audio shielding. The line continuously collects analog video and audio electrical signals and transmits them directly to the elevator machine room located on the top floor of the building. The hard-wired terminals in the machine room transfer the electrical signals to the transmitter of the optical transceiver. The optical transceiver converts the signals into optical signals and transmits them to the receiver optical transceiver in the property control room through a pre-embedded single-mode fiber optic line. The restored electrical signals are connected to the video and audio input interfaces of the monitoring host and the images are displayed on the LCD screen. The staff on duty in the monitoring room press the mechanical button on the desktop paging microphone to speak. The voice electrical signals are transmitted back through the same physical line and played in the cone-shaped speaker in the elevator car.

[0003] Traditional architectures rely primarily on analog electrical signals distributed between the edge and the cloud for continuous transmission via direct physical lines. When dealing with sudden emergencies within enclosed metal structures, they depend entirely on pre-embedded optical fibers and coaxial cables to construct fixed communication paths. The fixed transmission channels cannot detect external channel fluctuations and interference changes. Passive data transmission suffers severe losses due to mechanical vibrations. Dispersed audiovisual load signals cause large-scale packet loss and communication interruptions in weak network conditions, resulting in a disconnect between the on-site images and audio obtained from the outside. Summary of the Invention

[0004] To address the technical problems existing in the prior art, embodiments of the present invention provide an emergency rescue video call system and method for people trapped in malfunctioning elevators.

[0005] On the one hand, an emergency rescue video call system for people trapped in malfunctioning elevators was provided, which includes: The state-aware coding module acquires car image frames, calculates the signal-to-noise ratio change rate, filters the set of macroblock coordinates of motion regions where the motion vectors cross the boundary, increases the video coding quantization parameters when the signal-to-noise ratio change rate exceeds the channel attenuation threshold, performs spatial redundancy compression on macroblocks outside the set of macroblock coordinates of motion regions, and constructs the encoded video bitstream. The keyframe loss assessment module analyzes the overlap between the packet loss marker bitmap fed back from the cloud and the keyframe data segments in the encoded video bitstream, calculates the ratio of the maximum number of consecutive packet losses to the total number of keyframe packets, obtains the keyframe consecutive packet loss rate, and generates a keyframe packet loss alarm log when it exceeds the keyframe packet loss threshold. The asymmetric error correction scheduling module, based on the key frame packet loss alarm log, determines when the continuous packet loss rate of the key frame exceeds the standard, performs a multiplier amplification operation on the original forward error correction redundancy to obtain asymmetric forward error correction redundancy, increases the verification packet sending interval, and performs time-series interleaving with the original video payload to construct an enhanced video bitstream with forward error correction. The delay mapping retransmission module calculates the keyframe packet loss rate based on the enhanced video bitstream with forward error correction, generates a retransmission request message when the retransmission request trigger threshold is exceeded, monitors the network round-trip delay and calculates the proportion of the network round-trip delay to the video buffer time, determines the expected arrival time of the retransmission packet, and obtains link delay assessment data. The cross-modal time-domain synchronization module calculates the ratio of network round-trip delay to the remaining audio duration in the audio buffer based on the link delay assessment data to obtain the audio speed ratio. It performs speed-changing without pitch-changing processing on the PCM audio data to generate an extended audio data stream. It shifts the retransmission packet display timestamp to align with the end of the audio data stream and outputs a synchronized audio and video playback sequence.

[0006] As a further aspect of the present invention, the encoded video stream specifically includes a video compression payload, frame header identification information, and a quantization index sequence; The keyframe packet loss alarm log specifically includes out-of-bounds alarm entries, packet loss timestamps, and packet loss type codes; The enhanced video stream with forward error correction specifically includes a forward error correction check packet, an interleaving matrix data block, and a recovery address pointer; The link latency assessment data specifically includes round-trip latency, average network jitter, and predicted arrival interval; The audio and video synchronized playback sequence specifically includes an audio time base, a video rendering time, and a display offset compensation amount.

[0007] As a further aspect of the present invention, the state-aware encoding module includes: The image parsing and detection submodule acquires the original image frame sequence generated by the camera, extracts the spatial coordinate information of each group of pixel blocks in the original image frame sequence, calculates the motion vector offset matrix by comparing the position difference vector of the corresponding pixel blocks in the previous and next frames, obtains the preset motion detection threshold, compares the motion vector offset matrix with the motion detection threshold, filters out the region nodes with the offset exceeding the limit, and establishes a set of macroblock coordinates of the motion region. The channel-aware acquisition submodule reads the physical layer channel quality sampling records reported by the communication link, extracts the signal-to-noise ratio (SNR) of adjacent sampling periods within the sampling records, calculates the difference between adjacent sampling points based on the SNR, and performs a division operation between the difference and the sampling period time interval to obtain the SNR change rate. The redundant compression construction submodule reads the preset channel attenuation threshold. When it is determined that the signal-to-noise ratio change rate exceeds the channel attenuation threshold, it increases the quantization parameters of the video encoder and performs spatial redundant compression on the edge macroblocks outside the mapping range of the motion region macroblock coordinate set to generate an encoded video stream.

[0008] As a further aspect of the present invention, the process of obtaining the preset motion detection threshold specifically includes: The system collects three-axis acceleration signals from the elevator car base, extracts the mechanical vibration amplitude contained in the three-axis acceleration signals, calls the camera focal length parameters to map and calculate the mechanical vibration amplitude as the background global offset compensation value, collects background pixel fluctuation data in the stationary state of the elevator car to generate the basic pixel offset tolerance, performs numerical superposition calculation on the background global offset compensation value and the basic pixel offset tolerance, and sets the motion detection threshold. The process of reading the preset channel attenuation threshold is as follows: A mapping table is established by reading the elevator's full-travel communication noise floor mapping record. The current physical height coordinates of the elevator car are collected. The baseline attenuation value corresponding to the physical height coordinates is retrieved from the mapping table. The environmental interference attenuation compensation value is calculated by monitoring the fluctuation difference of the current received signal strength. The baseline attenuation value and the environmental interference attenuation compensation value are weighted and the channel attenuation threshold is set.

[0009] As a further aspect of the present invention, the keyframe loss evaluation module includes: The packet loss coverage analysis submodule receives a two-dimensional packet loss marker bitmap fed back by the cloud network node, extracts the message sequence number index marked as lost in the packet loss marker bitmap, reads the key frame data segment in the encoded video bitstream, compares the positional distribution relationship between the message sequence number index and the key frame data segment, calculates the overlap ratio of the two in the sequence number interval, and obtains the key frame packet loss coverage rate. The loss rate calculation submodule calls the keyframe packet loss coverage rate, defines the location of the packet loss segment based on the keyframe packet loss coverage rate, slides the statistical window along the sequence number within the segment location, counts the number of consecutively lost data packets in each window, extracts the maximum value to generate the maximum number of consecutive packet losses, reads the total number of packets contained in the keyframe data segment, performs a division operation between the maximum number of consecutive packet losses and the total number of packets, and obtains the keyframe consecutive packet loss rate. The out-of-bounds judgment and evaluation submodule calls the key frame continuous packet loss rate, reads the preset key frame packet loss threshold, compares the key frame continuous packet loss rate with the key frame packet loss threshold, and when it is determined that the key frame continuous packet loss rate exceeds the key frame packet loss threshold, it extracts the time node and lost packet sequence number information of the out-of-bounds occurrence, aggregates the time node and lost packet sequence number information, and generates a key frame packet loss alarm log.

[0010] As a further aspect of the present invention, the asymmetric error correction scheduling module includes: The redundancy adjustment submodule parses the keyframe packet loss alarm log to extract the keyframe continuous packet loss rate, reads the preset keyframe packet loss rate threshold, performs a comparison operation between the keyframe continuous packet loss rate and the keyframe packet loss rate threshold, and when it is determined that the value exceeds the limit, calls the original forward error correction redundancy associated with the encoded video bitstream, performs a multiplier amplification operation on the original forward error correction redundancy according to the keyframe continuous packet loss rate, and obtains the asymmetric forward error correction redundancy. The verification packet interval allocation submodule extracts the initial forward error correction verification packet corresponding to the key frame in the encoded video bitstream, reads the message transmission interval corresponding to the initial forward error correction verification packet, increases the message transmission interval in combination with the asymmetric forward error correction redundancy, obtains the basic interval of the verification packet associated with the conventional prediction frame, allocates a transmission interval greater than the basic interval of the verification packet for the initial forward error correction verification packet, and obtains the extended forward error correction verification packet sequence. The hybrid stream combination submodule extracts the original video payload contained in the encoded video stream, performs a timing concatenation and merging operation on the extended forward error correction check packet sequence and the original video payload, performs interleaving arrangement according to the message sequence number, builds a data packet transmission queue containing check redundancy information, and generates an enhanced video stream with forward error correction.

[0011] As a further aspect of the present invention, the process of reading the preset key frame packet loss rate threshold specifically includes: Extract a set of image frames from the encoded video stream, divide the set of image frames by the video encoding frame rate to obtain the key frame impact duration, collect the current link packet loss tolerance extreme value fed back by the network probe, perform weighted fitting between the current link packet loss tolerance extreme value and the key frame impact duration to obtain the basic packet loss rate parameter, analyze the historical packet loss records of the communication link to generate the channel degradation compensation coefficient, perform multiplication operation between the basic packet loss rate parameter and the channel degradation compensation coefficient to establish the key frame packet loss rate threshold in percentage form; The process of performing a scaling operation on the original forward error correction redundancy based on the continuous packet loss rate of keyframes is as follows: The continuous packet loss rate of the keyframe is input into a pre-fitted exponential mapping function to obtain the redundancy amplification factor. The original forward error correction redundancy and the redundancy amplification factor are multiplied to generate the initial amplified redundancy. The physical communication channel is monitored to obtain the real-time available bandwidth of the network. The difference between the real-time available bandwidth of the network and the basic transmission code rate is calculated. The upper limit threshold of the forward error correction bandwidth is set. When it is determined that the initial amplified redundancy exceeds the upper limit threshold of the forward error correction bandwidth, the upper limit truncation operation is performed on the initial amplified redundancy to set the asymmetric forward error correction redundancy.

[0012] As a further aspect of the present invention, the delay mapping retransmission module includes: The cache sequence analysis submodule reads the message sequence number records associated with the enhanced video stream with forward error correction in the cloud receiving cache, detects the continuity of the message sequence number values ​​on the time axis, counts the number of lost messages based on the continuity interruption, and performs a division operation between the number of lost messages and the total number of messages to obtain the key frame packet loss rate. The missing retransmission request submodule calls the key frame packet loss rate, reads the preset retransmission request trigger threshold, compares the key frame packet loss rate with the retransmission request trigger threshold, and when it is determined that the key frame packet loss rate exceeds the retransmission request trigger threshold, extracts the missing sequence number at the location where the packet loss occurred, encapsulates the missing sequence number into a node communication message, and generates a retransmission request message. The latency detection and evaluation submodule activates the node's listening port based on the retransmission request message, monitors the network round-trip latency generated by the network probe in the communication link, extracts the video buffer duration in the video playback buffer, calculates the ratio of network round-trip latency to video buffer duration, determines the expected arrival time range of the retransmitted data packet based on the ratio, and obtains link latency evaluation data.

[0013] As a further aspect of the present invention, the cross-modal time-domain synchronization module includes: The stretching coefficient calculation submodule extracts the network round-trip delay included in the link delay evaluation data, reads the remaining effective duration of voice recorded in the audio playback buffer, performs a division operation between the network round-trip delay and the remaining effective duration of voice, extracts the ratio as a time-domain stretching reference value, and obtains the audio speed ratio. The waveform non-pitch processing submodule acquires the received PCM audio data, extracts the repeating bands in the PCM audio data, performs pitch period search and data block insertion operations on the repeating bands, adjusts the playback speed of the PCM audio data according to the audio speed change ratio, performs speed change without pitch change processing operations in combination with waveform characteristics, and generates an audio data stream with extended duration. The time-domain aligned output submodule acquires the retransmission data packets sent by the network transmission node, extracts the display timestamp carried by the retransmission data packets, extracts the termination time point of the duration-extended audio data stream, performs a numerical translation operation on the display timestamp to align the time coordinate of the display timestamp with the termination time point, combines the audio-visual load signals, and outputs an audio-visual synchronized playback sequence. The process of performing pitch period search and data block insertion operations for repeating bands is as follows: Extract the fundamental frequency signal within the repeating band, perform cross-correlation operation on the adjacent sampling point array within the fundamental frequency signal to generate a cross-correlation sequence matrix, retrieve the peak coordinate node with the largest value in the cross-correlation sequence matrix to establish the overlapping splicing point, copy and connect the repeating band at the overlapping splicing point to construct the waveform segment to be inserted, and embed the waveform segment to be inserted into the time domain gap of the PCM audio data. The process of performing speed-changing without pitch-changing processing based on waveform features is as follows: The connection boundary interval of the PCM audio data after embedding the waveform segment to be inserted is extracted, the local phase spectrum corresponding to the connection boundary interval is extracted, a Hanning window smoothing mapping function spanning the connection boundary interval is constructed, the attenuation and growth slopes of the Hanning window smoothing mapping function are aligned with the two ends of the connection boundary interval and cross-fade-in and fade-out weighted operations are performed, the time domain spacing of adjacent pitch periods is recombined proportionally based on the audio speed ratio, and the local phase spectrum is locked to perform coherent superposition operation.

[0014] On the other hand, the emergency rescue video call method for people trapped in a malfunctioning elevator, which is based on the aforementioned emergency rescue video call system for people trapped in a malfunctioning elevator, includes the following steps: S1: Collect car image frames, calculate the signal-to-noise ratio change rate, filter the set of macroblock coordinates of motion regions where the motion vectors cross the boundary, increase the video coding quantization parameters when the signal-to-noise ratio change rate exceeds the channel attenuation threshold, perform spatial redundancy compression on macroblocks outside the set of macroblock coordinates of motion regions, and construct the encoded video bitstream; S2: Analyze the overlap between the packet loss marker bitmap fed back from the cloud and the key frame data segment in the encoded video bitstream, calculate the ratio of the maximum number of consecutive packet losses to the total number of key frame packets, obtain the key frame consecutive packet loss rate, and generate a key frame packet loss alarm log when the key frame packet loss threshold is exceeded. S3: Based on the key frame packet loss alarm log, when it is determined that the continuous packet loss rate of the key frame exceeds the standard, the original forward error correction redundancy is magnified by a multiplier to obtain asymmetric forward error correction redundancy, the verification packet sending interval is increased, and the original video payload is interleaved in time to construct an enhanced video bitstream with forward error correction. S4: Calculate the keyframe packet loss rate based on the enhanced video bitstream with forward error correction, generate a retransmission request message when the retransmission request trigger threshold is exceeded, monitor the network round-trip time and calculate the proportion of network round-trip time to video buffer time, determine the expected arrival time of the retransmission packet, and obtain link delay assessment data. S5: Based on the link delay assessment data, calculate the ratio of network round-trip delay to the remaining audio duration in the audio buffer to obtain the audio speed ratio, perform speed-changing without pitch-changing processing on the PCM audio data to generate a duration-extended audio data stream, shift the retransmission packet display timestamp to align with the end of the audio data stream, and output the audio and video synchronized playback sequence.

[0015] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following: By performing spatial redundancy compression based on dynamically acquired image features and signal-to-noise ratio variation patterns, it actively adapts to the attenuation of the underlying physical channel. By constructing a packet loss distribution matrix to deeply analyze the payload missing density and dynamically stretching the step distance of the verification packet to perform asymmetric error correction scheduling, it effectively avoids the destruction of key frame fragmentation caused by network environment deterioration. Combined with the bidirectional link delay detection comparison results, it performs periodic peak search and non-modulated audio extension operation on the underlying pulse waveform data, drives the retransmission packet time stamp shift calibration, maintains the temporal hard alignment of audio-visual data in the severely fluctuating weak network transmission environment, and completely eliminates the interactive faults of screen stuttering and voice tearing at the monitoring terminal. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a schematic diagram of an emergency rescue video call system for people trapped in a malfunctioning elevator provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the system framework of the present invention; Figure 3 This is a flowchart of the state-aware coding module in this invention; Figure 4 This is a flowchart of the keyframe loss assessment module in this invention; Figure 5 This is a flowchart of the asymmetric error correction scheduling module in this invention; Figure 6 This is a flowchart of the delay mapping retransmission module in this invention; Figure 7 This is a flowchart of the cross-modal time-domain synchronization module in this invention; Figure 8 This is a flowchart of an emergency rescue video call method for people trapped in a malfunctioning elevator, provided in an embodiment of the present invention. Detailed Implementation

[0018] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0019] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0020] This invention provides an emergency rescue video call system for people trapped in a malfunctioning elevator, such as... Figure 1-2 The diagram shown illustrates an emergency rescue video call system for people trapped in a malfunctioning elevator. The system includes: The state-aware coding module acquires car image frames, calculates the signal-to-noise ratio change rate, filters the set of macroblock coordinates of motion regions where the motion vectors cross the boundary, increases the video coding quantization parameters when the signal-to-noise ratio change rate exceeds the channel attenuation threshold, performs spatial redundancy compression on macroblocks outside the set of macroblock coordinates of motion regions, and constructs the encoded video bitstream. The keyframe loss assessment module analyzes the overlap between the packet loss marker bitmap fed back from the cloud and the keyframe data segments in the encoded video bitstream, calculates the ratio of the maximum number of consecutive packet losses to the total number of keyframe packets, obtains the keyframe consecutive packet loss rate, and generates a keyframe packet loss alarm log when it exceeds the keyframe packet loss threshold. The asymmetric error correction scheduling module, based on the key frame packet loss alarm log, determines when the continuous packet loss rate of key frames exceeds the standard. It performs a multiplier amplification operation on the original forward error correction redundancy to obtain asymmetric forward error correction redundancy, increases the verification packet sending interval, and performs time-series interleaving with the original video payload to construct an enhanced video bitstream with forward error correction. The delay mapping retransmission module calculates the keyframe packet loss rate based on the enhanced video bitstream with forward error correction. When the retransmission request trigger threshold is exceeded, a retransmission request message is generated. The network round-trip delay is monitored and the proportion of the network round-trip delay to the video buffer time is calculated. The estimated arrival time of the retransmission packet is determined and the link delay assessment data is obtained. The cross-modal time-domain synchronization module calculates the ratio of network round-trip delay to the remaining audio duration in the audio buffer based on link delay assessment data to obtain the audio speed ratio. It then performs speed-changing without pitch-changing processing on the PCM audio data to generate a duration-extended audio data stream. The module shifts the retransmission packet display timestamp to align with the end of the audio data stream and outputs a synchronized audio and video playback sequence.

[0021] The encoded video stream specifically includes the video compression payload, frame header identification information, and quantization index sequence; The keyframe packet loss alarm log specifically includes out-of-bounds alarm entries, packet loss timestamps, and packet loss type codes; The enhanced video stream with forward error correction specifically includes a forward error correction check packet, an interleaving matrix data block, and a recovery address pointer; Link latency assessment data specifically includes round-trip time, average network jitter, and predicted arrival interval; The audio and video synchronized playback sequence specifically includes the audio time base, the video rendering time, and the display offset compensation amount.

[0022] Specifically, such as Figure 2 , 3 As shown, the state-aware coding module includes: The image parsing and detection submodule acquires the original image frame sequence generated by the camera, extracts the spatial coordinate information of each group of pixel blocks in the original image frame sequence, calculates the motion vector offset matrix by comparing the position difference vector of the corresponding pixel blocks in the previous and next frames, obtains the preset motion detection threshold, compares the motion vector offset matrix with the motion detection threshold, filters out the region nodes with the offset exceeding the limit, and establishes a set of macroblock coordinates of the motion region. The system acquires the original image frame sequence generated by the camera and divides the single-frame image with a resolution of 1920 pixels by 1080 pixels into a 16-pixel by 16-pixel macroblock grid matrix. Image frames acquired in two adjacent sampling periods are read, and the basic spatial coordinate information of the upper left corner of the macroblocks with the same number in each frame is extracted. For corresponding pixel blocks in the two frames, the absolute error of the brightness residual is calculated. The similarity of pixel blocks is measured by subtracting and summing the brightness values ​​of each pixel. After finding the pixel block with the highest matching degree, the pixel value of each macroblock moving along the two-dimensional plane is calculated using the spatial difference matrix of the horizontal and vertical coordinates, thereby constructing a motion vector offset matrix covering the entire image. Subsequently, the dynamic calculation logic of the motion detection threshold is initiated, and the three-axis acceleration signal of the elevator car base is acquired. The mechanical vibration amplitude contained in this signal is extracted; in this embodiment, the mechanical vibration amplitude is taken as 0.15g. The camera focal length parameter is used to map and calculate the mechanical vibration amplitude into a background global offset compensation value, which in this embodiment is converted into a displacement of 2 pixels. Simultaneously, background pixel fluctuation data of the elevator car in a stationary state at a floor level is collected to generate a basic pixel offset tolerance, which is set to a displacement of 1 pixel in this embodiment. The background global offset compensation value and the basic pixel offset tolerance are numerically superimposed, resulting in 3, and a motion detection threshold of 3 pixels is dynamically generated and set. All vector values ​​contained in the motion vector offset matrix are compared with this motion detection threshold one by one. When the offset of a node in a specific region reaches 4 pixels or more, the node is identified as an out-of-bounds node. The geometric center coordinates of all out-of-bounds nodes are extracted and compiled into a structured mapping list. A set of macroblock coordinates of motion regions representing the real motion state is established in system memory. The process of establishing this set eliminates global background interference caused by slight shaking of the elevator car, providing a basis for regional localization for subsequent video compression.

[0023] The channel-aware acquisition submodule reads the physical layer channel quality sampling records reported by the communication link, extracts the signal-to-noise ratio (SNR) of adjacent sampling periods within the sampling records, calculates the difference between adjacent sampling points based on the SNR, and performs a division operation between the difference and the sampling period time interval to obtain the SNR change rate. The system reads the physical layer channel quality sampling records reported by the communication link, with a fixed sampling period of 50 milliseconds. The signal-to-noise ratio (SNR), reflecting channel quality, is extracted from the collected sampling records. Based on the aforementioned 50-millisecond time interval mechanism, a baseline SNR value of 25 dB measured in the previous period is retrieved, while the actual SNR value obtained in the latest period is extracted as 15 dB. A subtraction operation is performed on these two period values, subtracting the previous period's 25 dB from the current period's 15 dB, resulting in a difference of -10 dB between adjacent sampling points. Next, this -10 dB difference is extracted and divided by the fixed 50-millisecond sampling period, yielding a SNR change rate of -0.2 dB per millisecond, representing the physical channel deterioration trend. This change rate, as a quantitative indicator reflecting the signal attenuation rate, can quickly reveal the signal drop trend caused by the sudden increase in the metal shielding effect within the elevator shaft. The calculated change rate is stored in a designated address for subsequent compression and judgment processes to read and retrieve at any time.

[0024] The redundant compression construction submodule reads the preset channel attenuation threshold. When it is determined that the signal-to-noise ratio change rate exceeds the channel attenuation threshold, it increases the quantization parameters of the video encoder and performs spatial redundant compression on the edge macroblocks outside the mapping range of the macroblock coordinate set of the motion region to generate the encoded video bitstream. The pre-configured channel attenuation threshold, set to -0.1 dB per millisecond, is read. The signal-to-noise ratio change rate (SNR) calculated in the previous step (-0.2 dB per millisecond) is compared with the threshold value of -0.1 dB per millisecond. If the comparison confirms that the attenuation rate of -0.2 dB per millisecond exceeds the limit of -0.1 dB per millisecond in absolute negative value, a warning condition is triggered based on the channel condition. A soft interrupt control command is then generated to modify the underlying video encoding environment parameters. The quantization parameters in the video encoding stage are increased from the default 28th order value to a 35th order value representing a heavily compressed state. Under the constraint of this higher-order quantization parameter, the previously generated set of motion region macroblock coordinates is read, and a 35th-order quantization truncation is applied to all peripheral edge macroblocks in the image space not mapped by this set. By filtering out high-frequency spatial details over a large area, deep spatial redundancy compression is performed on static background walls and other areas. This compression process reduces the overall video frame size to less than 40% of its original size while ensuring that the image clarity in areas of personnel movement is not compromised. The resulting encapsulated video stream possesses resilience against weak network conditions, reducing the risk of bandwidth congestion in confined elevator shaft environments.

[0025] Specifically, such as Figure 2 , 4 As shown, the keyframe loss assessment module includes: The packet loss coverage analysis submodule receives a two-dimensional packet loss marker bitmap fed back from the cloud network node, extracts the packet sequence number index marked as lost in the packet loss marker bitmap, reads the key frame data segment in the encoded video bitstream, compares the positional distribution relationship between the packet sequence number index and the key frame data segment, calculates the overlap ratio of the two in the sequence number interval, and obtains the key frame packet loss coverage rate. The system receives a two-dimensional packet loss marker bitmap sent by the cloud network node via the feedback control channel. This bitmap records the liveness and loss status of message sequences in the network transmission channel. The bitmap is scanned to extract the sequence number indices of all messages marked as lost. For example, the extracted lost sequence numbers are messages 1020 to 1025. Simultaneously, the headers of the locally waiting encoded video stream are parsed to extract the consecutive numbered intervals corresponding to keyframe data segments, confirming that the sequence number range covered by the keyframe data segments is 1000 to 1050. The extracted set of lost sequence number indices is compared with the complete keyframe data segment sequence to confirm the specific distribution location of the lost points within the keyframe data blocks. Statistical confirmation shows that the six lost packets located in the 1020 to 1025 interval all fall within the keyframe range of 1000 to 1050. The number of these six packets is used as the numerator, and a division is performed using the total number of packets spanning the keyframe interval (51 packets) as the denominator. The final keyframe packet loss coverage rate, calculated numerically, is 0.1176. This ratio reflects the proportion of underlying physical packet loss events occurring on core video frames, providing a reliable data foundation for assessing the potential probability of video decoding corruption.

[0026] The loss rate calculation submodule calls the keyframe packet loss coverage rate, defines the location of the packet segment that has lost packets based on the keyframe packet loss coverage rate, slides the statistical window along the sequence number within the segment location, counts the number of consecutively lost data packets in each window, extracts the maximum value to generate the maximum number of consecutive packet losses, reads the total number of packets contained in the keyframe data segment, performs a division operation between the maximum number of consecutive packet losses and the total number of packets, and obtains the keyframe consecutive packet loss rate. The generated keyframe packet loss coverage rate is used to define the spatial location of consecutive packet loss segments, specifically the interval from sequence number 1020 to 1025, based on the damaged boundary index defined by this rate. To assess the impact of consecutive packet loss, a sliding statistical window with a width of 10 packet units is constructed at the beginning of this segment location. This window is then moved sequentially along the sequence number coordinate axis of the keyframe data segment, with a sliding step size of 1 packet unit. During each mapping anchoring period, the total number of packets exhibiting consecutive loss within the window is counted. After the window has scanned the entire interval, the statistical peaks generated by each window are traversed, and the consecutive loss count with the largest value is selected and extracted. In this example, the largest consecutive loss count is 6 packets, which are extracted separately to generate the maximum consecutive packet loss count. Subsequently, the total number of packets contained in this keyframe data segment is read, which is constant at 51. The maximum consecutive packet loss count of 6 is substituted into the dividend position, and the total number of packets of 51 is substituted into the divisor position, and a division operation is performed to obtain a quotient of 0.1176. Convert the quotient to a percentage format of 11.76% to obtain the keyframe continuous packet loss rate, which is used to characterize the degree of continuous packet loss.

[0027] Table 1. Distribution of Keyframe Packet Loss Analysis

[0028] As shown in Table 1, the specific data flow of using a sliding statistical window to count the maximum number of consecutive packet losses and convert it into the consecutive packet loss rate of key frames quantifies the packet loss status when the communication environment deteriorates.

[0029] The out-of-bounds judgment and evaluation submodule calls the key frame continuous packet loss rate, reads the preset key frame packet loss threshold, compares the key frame continuous packet loss rate with the key frame packet loss threshold, and when it is determined that the key frame continuous packet loss rate exceeds the key frame packet loss threshold, it extracts the time node and lost packet sequence number information of the out-of-bounds occurrence, aggregates the time node and lost packet sequence number information, and generates a key frame packet loss alarm log. The keyframe continuous packet loss rate, calculated using the preceding steps, is 11.76%. A pre-coded keyframe packet loss threshold is read from the system configuration file. Based on the high reliability requirements of the elevator emergency rescue passage, this threshold is limited to 10%. The dynamically acquired 11.76% and the threshold 10% are simultaneously substituted into a numerical comparison function for verification. The verification results show that the dynamic parameter exceeds the allowable range of the keyframe packet loss threshold, triggering a numerical out-of-bounds interruption signal. During the interruption response, the system clock record is extracted to obtain the time node information for determining the out-of-bounds event as 45 seconds and 720 milliseconds. The initial and final sequence numbers of the lost packets that triggered this out-of-bounds event are simultaneously captured, i.e., sequence numbers 1020 to 1025. Finally, the captured time node timestamp information and these lost packet sequence numbers are packaged and aggregated, assigning this data a unique event identifier. A keyframe packet loss alarm log containing time and sequence number records is generated according to the preset format specifications. This log will be stored in a specific cache area as a criterion for triggering subsequent forward error correction and compensation mechanisms.

[0030] Specifically, such as Figure 2 , 5 As shown, the asymmetric error correction scheduling module includes: The redundancy adjustment submodule parses the keyframe packet loss alarm log to extract the keyframe continuous packet loss rate, reads the preset keyframe packet loss rate threshold, performs a comparison operation between the keyframe continuous packet loss rate and the keyframe packet loss rate threshold, and when it is determined that the value exceeds the limit, it calls the original forward error correction redundancy associated with the encoded video bitstream, performs a multiplier amplification operation on the original forward error correction redundancy based on the keyframe continuous packet loss rate, and obtains the asymmetric forward error correction redundancy. The keyframe packet loss alarm log stored in memory is parsed, and the continuous packet loss rate of 11.76% is extracted. The keyframe packet loss rate threshold for measuring communication link status, limited to 8%, is read from the underlying network protocol specification table. A subtraction comparison is performed between 11.76% and 8%. If the difference is greater than zero (indicating a numerical out-of-bounds condition), the original forward error correction redundancy bound to the current encoded video stream is invoked. This parameter is set to 2 basic redundancy packet distances by default. The exponential mapping function is enabled, and the continuous packet loss rate of the keyframes is input into this function to calculate the redundancy amplification factor. For a packet loss rate of 11.76%, the calculated amplification factor is 1.5. Based on this factor, the original forward error correction redundancy 2 is amplified. The original parameter value 2 is multiplied by the amplification factor 1.5 to obtain result 3. This value 3 is extracted and set as the asymmetric forward error correction redundancy used by the system to combat sudden channel packet loss. The value of 3 indicates that the insertion interval of the verification packet will change dynamically when assembling the data stream, which improves the pertinence and effectiveness of the subsequent data recovery mechanism.

[0031] The verification packet interval allocation submodule extracts the initial forward error correction verification packet corresponding to the key frame in the encoded video bitstream, reads the message transmission interval corresponding to the initial forward error correction verification packet, increases the message transmission interval in combination with the asymmetric forward error correction redundancy, obtains the basic interval of the verification packet associated with the regular prediction frame, allocates a transmission interval greater than the basic interval of the verification packet for the initial forward error correction verification packet, and obtains the extended forward error correction verification packet sequence. The protocol encapsulation shell of the encoded video stream is unpacked to extract the initial forward error correction (FEC) check packet payload block attached to the keyframe data segment. The default message transmission interval corresponding to this initial FEC check packet, recorded in the transmission queue manager, is read; the base value is 20 milliseconds. The specific value of the asymmetric forward error correction redundancy (3) generated in the previous step is retrieved and used as a multiplier factor in the transmission interval adjustment calculation. The original message transmission interval is increased by multiplying 20 milliseconds by 3, extending it to a transmission span of 60 milliseconds. Based on this calculation, the base interval of the check packet associated with the predicted frame under normal lossless conditions is queried and obtained; its value is fixed at 2 message positions. Using the calculated new 60-millisecond interval, a special transmission interval larger than the base interval of the check packet is reallocated and assigned to the initial FEC check packet, effectively spanning a distance of at least 4 regular video message units. Through this time-span stretching and isolation allocation, a sparser extended FEC check packet sequence data pool is output. This operation separates the checksum data from the original data during transmission, reducing the probability of both the original packet and the checksum packet being lost simultaneously within the same congestion window.

[0032] The hybrid stream combination submodule extracts the original video payload contained in the encoded video stream, performs a timing splicing and merging operation on the extended forward error correction check packet sequence and the original video payload, performs interleaving arrangement according to the message sequence number, builds a data packet transmission queue containing check redundancy information, and generates an enhanced video stream with forward error correction. The raw video payload data stream is read in batches from the backbone encoding buffer. Two parallel concurrent time axes are established: one aligned with the raw video payload, and the other aligned with the extended forward error correction (FEC) packet sequence generated in the previous process. These two data streams are interleaved, spliced, and merged on a unified time dimension. During this process, based on the message sequence number index issued by the underlying communication socket, the interleaving arrangement is performed according to the newly set asymmetric redundancy logic. Specifically, the sorting rule is adjusted so that after every 3 or 4 raw video payload data packets are sent, one data packet carrying the check redundancy information is inserted. Through this staggered deployment method, the extended FEC packet sequence is merged into the raw video sequence, and finally combined in the transmission buffer ring to form an enhanced video bitstream with forward error correction.

[0033] Table 2 Enhanced Video Stream Transmission Queue Interleaving Sort Table

[0034] As shown in Table 2, by implementing interleaving and combination arrangements at specific intervals, the enhanced video stream transmission queue changes the conventional continuous payload arrangement pattern. Distributing and interleaving check packets between normal audio and video payloads improves data survivability in elevator shaft environments with weak network connectivity, ensuring the retention of critical video information.

[0035] Specifically, such as Figure 2 , 6 As shown, the delay mapping retransmission module includes: The cache sequence analysis submodule reads the message sequence number records associated with the enhanced video bitstream with forward error correction in the cloud receive cache, detects the continuity of the message sequence number values ​​on the time axis, counts the number of lost messages based on the continuity interruption, and performs a division operation between the number of lost messages and the total number of messages to obtain the keyframe packet loss rate. The latest message snapshot stored in the cloud-based receive buffer of the elevator security monitoring room is read via a remote procedure call interface. For this snapshot data, a targeted scan is performed on the message sequence number record set associated with the generated enhanced video stream with forward error correction, for example, extracting a total of 200 mixed message data captured in the current batch. Following a unidirectional increasing order along the timeline, a character-by-character continuity comparison is performed on the sequence number records of these 200 messages. When detecting the numerical continuity of the sequence numbers, hidden numerical discontinuities within the message sequence are identified, such as a jump from 1003 to 1006, or from 1050 to 1062. Based on the missing space reflected by these numerical discontinuities, an accumulation and statistical operation is initiated to calculate the number of missing messages, accumulating to 15 missing independent data packets. A division operation is then called, using the statistically calculated number of missing messages (15) as the numerator and the total number of messages sent in this detection cycle (200) as the denominator, to perform a division operation, yielding a result of 0.075. Convert 0.075 to 7.5% as a percentage, output it, and label it as a keyframe packet loss rate parameter used to characterize the current state of video frame loss.

[0036] The missing retransmission request submodule calls the keyframe packet loss rate, reads the preset retransmission request trigger threshold, compares the keyframe packet loss rate with the retransmission request trigger threshold, and when it is determined that the keyframe packet loss rate exceeds the retransmission request trigger threshold, it extracts the missing sequence number at the location where the packet loss occurred, encapsulates the missing sequence number into a node communication message, and generates a retransmission request message. The calculated 7.5% keyframe packet loss rate parameter is extracted and read from memory. The system's protocol configuration center is accessed to retrieve the pre-set retransmission request trigger threshold for the elevator-end-cloud collaborative communication architecture, which is fixed at 5%. The dynamically acquired packet loss rate of 7.5% and the fixed retransmission request trigger threshold of 5% are compared in the comparison logic. The result indicates that 7.5% exceeds the 5% protection limit, confirming that the packet loss rate exceeds the limit. Upon confirmation of the excessive value, the protocol stack's index backtracking mechanism is triggered to reverse-search and extract the missing sequence number index group corresponding to the location of the fault hole from the underlying network transmission log. For example, message sequence numbers 1004, 1005, 1051 to 1061 discovered during the preceding comparison are extracted. For these clearly located missing sequence numbers, the underlying message formatting and reassembly operation is performed to generate node communication messages based on the negative acknowledgment protocol format. A high-priority processing identifier is assigned to the header of this communication message, and it is combined and encapsulated to generate a retransmission request message. This message is ready to initiate a data retransmission request to the elevator car transmitter at any time.

[0037] The latency detection and evaluation submodule activates the node listening port based on the retransmission request message, monitors the network round-trip latency generated by the network probe in the communication link, extracts the video buffer duration in the video playback buffer, calculates the ratio of network round-trip latency to video buffer duration, determines the expected arrival time range of retransmitted data packets based on the ratio, and obtains link latency evaluation data. Based on the captured retransmission request message, the activation code carried in the message is used to wake up the listening port located at the core node of the transmission link. After the port is activated, the network probes deployed in the communication link are continuously monitored. The network probes measure the network round-trip time generated by electromagnetic waves between the sender and receiver by sending lightweight echo request messages and waiting for acknowledgment responses; for example, the current round-trip time is measured to be 120 milliseconds. At the same time, the video buffer duration parameter reserved for smoothing network jitter in the video playback buffer is extracted, and the value is read as 300 milliseconds. The measured 120-millisecond network round-trip time and 300-millisecond video buffer duration are extracted, and 120 milliseconds is divided by 300 milliseconds to calculate that the time ratio of this delay to the total buffer duration is 40%. Based on the judgment condition that this 40% time ratio has not reached the 100% exhaustion threshold, it is estimated that the target retransmission data packet can arrive before the buffer is exhausted. Therefore, the expected arrival time range of the retransmission data packet is established to be within the window range of 150 to 200 milliseconds after the command is issued. By integrating various time-consuming parameters and judgment intervals, link latency assessment data records are generated.

[0038] Specifically, such as Figure 2 , 7 As shown, the cross-modal time-domain synchronization module includes: The stretching coefficient calculation submodule extracts the network round-trip delay included in the link delay assessment data, reads the remaining effective duration of voice recorded in the audio playback buffer, performs a division operation between the network round-trip delay and the remaining effective duration of voice, extracts the ratio as a time-domain stretching reference value, and obtains the audio speed ratio. The measured network round-trip time (RTD) was extracted from the link latency assessment data records and confirmed to be 120 milliseconds. Using memory read technology to delve into the audio processing, the remaining effective audio duration recorded in the temporary queue of the audio playback buffer was obtained and confirmed to be 800 milliseconds. After obtaining these two time-domain boundary data, a division operation was performed between the 120-millisecond RTD and the 800-millisecond remaining effective audio duration. Substituting the RTD of 120 into the numerator and the remaining effective audio duration of 800 into the denominator, the division yielded a decimal ratio of 0.15, which served as the reference value for time-domain scaling. To ensure smooth and continuous audio playback, a natural playback ratio constant of 1 was set, and the extracted reference value of 0.15 was summed with the constant 1. The calculated result was 1.15, which was marked as the audio speed ratio. This ratio of 1.15 represents the need to amplify the original playback time by 15% in subsequent audio waveform processing stages to fill the time gap generated during the waiting period for video retransmission.

[0039] The waveform non-pitch processing submodule acquires the received PCM audio data, extracts the repeating bands in the PCM audio data, performs pitch period search and data block insertion operations on the repeating bands, adjusts the playback speed of the PCM audio data according to the audio speed change ratio, performs speed change without pitch change processing operations in combination with waveform characteristics, and generates an audio data stream with extended duration. The system acquires PCM audio data received and reconstructed by the underlying audio decoder, with a fixed base sampling frequency of 8000 Hz. A sliding detection method is used on the time-domain coordinate axis to locate and extract band regions containing repetitive features. For the captured repetitive bands, a pitch period search and waveform truncation operation is performed to extract a 10-millisecond waveform segment to be inserted and copied. The calculated audio speed ratio of 1.15 is then read. Based on this 1.15 speed ratio, the playback time axis of the original PCM audio data is redistributed, extending the original 800-millisecond physical duration to a 920-millisecond time domain. During this time extension, the copied waveform segment is embedded into the time-domain gaps of the PCM audio data. Combining the phase envelope characteristics of the band splicing points, an overlap-add algorithm is used to perform a smooth transition operation with unchanged pitch. This extension operation ensures that the fundamental frequency characteristics of the original waveform remain unchanged, generating a continuous and distortion-free extended audio data stream.

[0040] Table 3. Tone Period Band Stretch Control Table

[0041] As shown in Table 3, the dynamic interpolation and scaling control process based on pitch period data can smoothly transition the playback pauses caused by network latency.

[0042] The time-domain aligned output submodule acquires the retransmission data packets sent by the network transmission node, extracts the display timestamp carried by the retransmission data packets, extracts the termination time point of the duration-extended audio data stream, performs a numerical translation operation on the display timestamp to align the time coordinate of the display timestamp with the termination time point, combines the audio-visual payload signals, and outputs an audio-visual synchronized playback sequence. The system acquires high-priority retransmission data packets delivered by network transmission nodes. It performs header decomposition on the arriving retransmission packets, extracting the embedded display timestamp parameter for decoder reference. The original display timestamp read at this point is at the 800ms node on the timeline. The system then switches to the audio processing channel, extracting the specific termination time point on the timeline of the previously generated duration-extended audio data stream. Information extraction confirms this termination time point is at the 920ms node on the timeline. Due to the delay in the arrival of the retransmission packets, the system performs a numerical shift operation on the read 800ms display timestamp using the arithmetic compensation process. A 120ms delay compensation offset is directly added to this display timestamp, causing the display timestamp of the retransmission data packets to shift and align from 800ms to the 920ms termination time point. Subsequently, in the rendering pipeline, the time-aligned visual payload and auditory audio signal are combined. The audio and video synchronous playback sequence is merged and output at the hardware output port to ensure that the final monitoring screen and sound are consistent in time, thus improving the audio-visual disconnection phenomenon that occurs during retransmission.

[0043] Please see Figure 8 The emergency rescue video call method for people trapped in a malfunctioning elevator is based on the aforementioned emergency rescue video call system for people trapped in a malfunctioning elevator and includes the following steps: S1: Collect car image frames, calculate the signal-to-noise ratio change rate, filter the set of macroblock coordinates of motion regions where the motion vectors cross the boundary, increase the video coding quantization parameters when the signal-to-noise ratio change rate exceeds the channel attenuation threshold, perform spatial redundancy compression on macroblocks outside the set of macroblock coordinates of motion regions, and construct the encoded video bitstream; S2: Analyze the overlap between the packet loss marker bitmap fed back from the cloud and the key frame data segments in the encoded video bitstream, calculate the ratio of the maximum number of consecutive packet losses to the total number of key frame packets, obtain the key frame consecutive packet loss rate, and generate a key frame packet loss alarm log when the key frame packet loss threshold is exceeded. S3: Based on the keyframe packet loss alarm log, when it is determined that the continuous packet loss rate of keyframes exceeds the standard, the original forward error correction redundancy is magnified to obtain asymmetric forward error correction redundancy, the verification packet sending interval is increased, and the original video payload is interleaved in time to construct an enhanced video bitstream with forward error correction. S4: Calculate the keyframe packet loss rate based on the enhanced video bitstream with forward error correction, generate a retransmission request message when the retransmission request trigger threshold is exceeded, monitor the network round-trip latency and calculate the proportion of network round-trip latency to video buffer time, determine the expected arrival time of the retransmission packet, and obtain link latency assessment data. S5: Based on the link delay assessment data, calculate the ratio of network round-trip delay to the remaining audio duration in the audio buffer to obtain the audio speed ratio. Perform speed-changing without pitch-changing processing on the PCM audio data to generate a duration-extended audio data stream. Align the timestamp of the retransmission packet display with the end of the audio data stream and output the audio and video synchronized playback sequence.

[0044] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. An emergency rescue video call system for people trapped in a malfunctioning elevator, characterized in that, The system includes: The state-aware coding module acquires car image frames, calculates the signal-to-noise ratio change rate, filters the set of macroblock coordinates of motion regions where the motion vectors cross the boundary, increases the video coding quantization parameters when the signal-to-noise ratio change rate exceeds the channel attenuation threshold, performs spatial redundancy compression on macroblocks outside the set of macroblock coordinates of motion regions, and constructs the encoded video bitstream. The keyframe loss assessment module analyzes the overlap between the packet loss marker bitmap fed back from the cloud and the keyframe data segments in the encoded video bitstream, calculates the ratio of the maximum number of consecutive packet losses to the total number of keyframe packets, obtains the keyframe consecutive packet loss rate, and generates a keyframe packet loss alarm log when it exceeds the keyframe packet loss threshold. The asymmetric error correction scheduling module, based on the key frame packet loss alarm log, determines when the continuous packet loss rate of the key frame exceeds the standard, performs a multiplier amplification operation on the original forward error correction redundancy to obtain asymmetric forward error correction redundancy, increases the verification packet sending interval, and performs time-series interleaving with the original video payload to construct an enhanced video bitstream with forward error correction. The delay mapping retransmission module calculates the keyframe packet loss rate based on the enhanced video bitstream with forward error correction, generates a retransmission request message when the retransmission request trigger threshold is exceeded, monitors the network round-trip delay and calculates the proportion of the network round-trip delay to the video buffer time, determines the expected arrival time of the retransmission packet, and obtains link delay assessment data. The cross-modal time-domain synchronization module calculates the ratio of network round-trip delay to the remaining audio duration in the audio buffer based on the link delay assessment data to obtain the audio speed ratio. It performs speed-changing without pitch-changing processing on the PCM audio data to generate an extended audio data stream. It shifts the retransmission packet display timestamp to align with the end of the audio data stream and outputs a synchronized audio and video playback sequence.

2. The emergency rescue video call system for people trapped in a malfunctioning elevator according to claim 1, characterized in that, The encoded video stream specifically includes a video compression payload, frame header identification information, and a quantization index sequence; The keyframe packet loss alarm log specifically includes out-of-bounds alarm entries, packet loss timestamps, and packet loss type codes; The enhanced video stream with forward error correction specifically includes a forward error correction check packet, an interleaving matrix data block, and a recovery address pointer; The link latency assessment data specifically includes round-trip latency, average network jitter, and predicted arrival interval; The audio and video synchronized playback sequence specifically includes an audio time base, a video rendering time, and a display offset compensation amount.

3. The emergency rescue video call system for people trapped in a malfunctioning elevator according to claim 1, characterized in that, The state-aware coding module includes: The image parsing and detection submodule acquires the original image frame sequence generated by the camera, extracts the spatial coordinate information of each group of pixel blocks in the original image frame sequence, calculates the motion vector offset matrix by comparing the position difference vector of the corresponding pixel blocks in the previous and next frames, obtains the preset motion detection threshold, compares the motion vector offset matrix with the motion detection threshold, filters out the region nodes with the offset exceeding the limit, and establishes a set of macroblock coordinates of the motion region. The channel-aware acquisition submodule reads the physical layer channel quality sampling records reported by the communication link, extracts the signal-to-noise ratio (SNR) of adjacent sampling periods within the sampling records, calculates the difference between adjacent sampling points based on the SNR, and performs a division operation between the difference and the sampling period time interval to obtain the SNR change rate. The redundant compression construction submodule reads the preset channel attenuation threshold. When it is determined that the signal-to-noise ratio change rate exceeds the channel attenuation threshold, it increases the quantization parameters of the video encoder and performs spatial redundant compression on the edge macroblocks outside the mapping range of the motion region macroblock coordinate set to generate an encoded video stream.

4. The emergency rescue video call system for people trapped in a malfunctioning elevator according to claim 3, characterized in that, The process of obtaining the preset motion detection threshold is as follows: The system collects three-axis acceleration signals from the elevator car base, extracts the mechanical vibration amplitude contained in the three-axis acceleration signals, calls the camera focal length parameters to map and calculate the mechanical vibration amplitude as the background global offset compensation value, collects background pixel fluctuation data in the stationary state of the elevator car to generate the basic pixel offset tolerance, performs numerical superposition calculation on the background global offset compensation value and the basic pixel offset tolerance, and sets the motion detection threshold. The process of reading the preset channel attenuation threshold is as follows: A mapping table is established by reading the elevator's full-travel communication noise floor mapping record. The current physical height coordinates of the elevator car are collected. The baseline attenuation value corresponding to the physical height coordinates is retrieved from the mapping table. The environmental interference attenuation compensation value is calculated by monitoring the fluctuation difference of the current received signal strength. The baseline attenuation value and the environmental interference attenuation compensation value are weighted and the channel attenuation threshold is set.

5. The emergency rescue video call system for people trapped in a malfunctioning elevator according to claim 3, characterized in that, The keyframe loss assessment module includes: The packet loss coverage analysis submodule receives a two-dimensional packet loss marker bitmap fed back by the cloud network node, extracts the message sequence number index marked as lost in the packet loss marker bitmap, reads the key frame data segment in the encoded video bitstream, compares the positional distribution relationship between the message sequence number index and the key frame data segment, calculates the overlap ratio of the two in the sequence number interval, and obtains the key frame packet loss coverage rate. The loss rate calculation submodule calls the keyframe packet loss coverage rate, defines the location of the packet loss segment based on the keyframe packet loss coverage rate, slides the statistical window along the sequence number within the segment location, counts the number of consecutively lost data packets in each window, extracts the maximum value to generate the maximum number of consecutive packet losses, reads the total number of packets contained in the keyframe data segment, performs a division operation between the maximum number of consecutive packet losses and the total number of packets, and obtains the keyframe consecutive packet loss rate. The out-of-bounds judgment and evaluation submodule calls the key frame continuous packet loss rate, reads the preset key frame packet loss threshold, compares the key frame continuous packet loss rate with the key frame packet loss threshold, and when it is determined that the key frame continuous packet loss rate exceeds the key frame packet loss threshold, it extracts the time node and lost packet sequence number information of the out-of-bounds occurrence, aggregates the time node and lost packet sequence number information, and generates a key frame packet loss alarm log.

6. The emergency rescue video call system for people trapped in a malfunctioning elevator according to claim 5, characterized in that, The asymmetric error correction scheduling module includes: The redundancy adjustment submodule parses the keyframe packet loss alarm log to extract the keyframe continuous packet loss rate, reads the preset keyframe packet loss rate threshold, performs a comparison operation between the keyframe continuous packet loss rate and the keyframe packet loss rate threshold, and when it is determined that the value exceeds the limit, calls the original forward error correction redundancy associated with the encoded video bitstream, performs a multiplier amplification operation on the original forward error correction redundancy according to the keyframe continuous packet loss rate, and obtains the asymmetric forward error correction redundancy. The verification packet interval allocation submodule extracts the initial forward error correction verification packet corresponding to the key frame in the encoded video bitstream, reads the message transmission interval corresponding to the initial forward error correction verification packet, increases the message transmission interval in combination with the asymmetric forward error correction redundancy, obtains the basic interval of the verification packet associated with the conventional prediction frame, allocates a transmission interval greater than the basic interval of the verification packet for the initial forward error correction verification packet, and obtains the extended forward error correction verification packet sequence. The hybrid stream combination submodule extracts the original video payload contained in the encoded video stream, performs a timing concatenation and merging operation on the extended forward error correction check packet sequence and the original video payload, performs interleaving arrangement according to the message sequence number, builds a data packet transmission queue containing check redundancy information, and generates an enhanced video stream with forward error correction.

7. The emergency rescue video call system for people trapped in a malfunctioning elevator according to claim 6, characterized in that, The process of reading the preset keyframe packet loss rate threshold is as follows: Extract a set of image frames from the encoded video stream, divide the set of image frames by the video encoding frame rate to obtain the key frame impact duration, collect the current link packet loss tolerance extreme value fed back by the network probe, perform weighted fitting between the current link packet loss tolerance extreme value and the key frame impact duration to obtain the basic packet loss rate parameter, analyze the historical packet loss records of the communication link to generate the channel degradation compensation coefficient, perform multiplication operation between the basic packet loss rate parameter and the channel degradation compensation coefficient to establish the key frame packet loss rate threshold in percentage form; The process of performing a scaling operation on the original forward error correction redundancy based on the continuous packet loss rate of keyframes is as follows: The continuous packet loss rate of the keyframe is input into a pre-fitted exponential mapping function to obtain the redundancy amplification factor. The original forward error correction redundancy and the redundancy amplification factor are multiplied to generate the initial amplified redundancy. The physical communication channel is monitored to obtain the real-time available bandwidth of the network. The difference between the real-time available bandwidth of the network and the basic transmission code rate is calculated. The upper limit threshold of the forward error correction bandwidth is set. When it is determined that the initial amplified redundancy exceeds the upper limit threshold of the forward error correction bandwidth, the upper limit truncation operation is performed on the initial amplified redundancy to set the asymmetric forward error correction redundancy.

8. The emergency rescue video call system for people trapped in a malfunctioning elevator according to claim 6, characterized in that, The delay mapping retransmission module includes: The cache sequence analysis submodule reads the message sequence number records associated with the enhanced video stream with forward error correction in the cloud receiving cache, detects the continuity of the message sequence number values ​​on the time axis, counts the number of lost messages based on the continuity interruption, and performs a division operation between the number of lost messages and the total number of messages to obtain the key frame packet loss rate. The missing retransmission request submodule calls the key frame packet loss rate, reads the preset retransmission request trigger threshold, compares the key frame packet loss rate with the retransmission request trigger threshold, and when it is determined that the key frame packet loss rate exceeds the retransmission request trigger threshold, extracts the missing sequence number at the location where the packet loss occurred, encapsulates the missing sequence number into a node communication message, and generates a retransmission request message. The latency detection and evaluation submodule activates the node's listening port based on the retransmission request message, monitors the network round-trip latency generated by the network probe in the communication link, extracts the video buffer duration in the video playback buffer, calculates the ratio of network round-trip latency to video buffer duration, determines the expected arrival time range of the retransmitted data packet based on the ratio, and obtains link latency evaluation data.

9. The emergency rescue video call system for people trapped in a malfunctioning elevator according to claim 8, characterized in that, The cross-modal time-domain synchronization module includes: The stretching coefficient calculation submodule extracts the network round-trip delay included in the link delay evaluation data, reads the remaining effective duration of voice recorded in the audio playback buffer, performs a division operation between the network round-trip delay and the remaining effective duration of voice, extracts the ratio as a time-domain stretching reference value, and obtains the audio speed ratio. The waveform non-pitch processing submodule acquires the received PCM audio data, extracts the repeating bands in the PCM audio data, performs pitch period search and data block insertion operations on the repeating bands, adjusts the playback speed of the PCM audio data according to the audio speed change ratio, performs speed change without pitch change processing operations in combination with waveform characteristics, and generates an audio data stream with extended duration. The time-domain aligned output submodule acquires the retransmission data packets sent by the network transmission node, extracts the display timestamp carried by the retransmission data packets, extracts the termination time point of the duration-extended audio data stream, performs a numerical translation operation on the display timestamp to align the time coordinate of the display timestamp with the termination time point, combines the audio-visual load signals, and outputs an audio-visual synchronized playback sequence. The process of performing pitch period search and data block insertion operations for repeating bands is as follows: Extract the fundamental frequency signal within the repeating band, perform cross-correlation operation on the adjacent sampling point array within the fundamental frequency signal to generate a cross-correlation sequence matrix, retrieve the peak coordinate node with the largest value in the cross-correlation sequence matrix to establish the overlapping splicing point, copy and connect the repeating band at the overlapping splicing point to construct the waveform segment to be inserted, and embed the waveform segment to be inserted into the time domain gap of the PCM audio data. The process of performing speed-changing without pitch-changing processing based on waveform features is as follows: The connection boundary interval of the PCM audio data after embedding the waveform segment to be inserted is extracted, the local phase spectrum corresponding to the connection boundary interval is extracted, a Hanning window smoothing mapping function spanning the connection boundary interval is constructed, the attenuation and growth slopes of the Hanning window smoothing mapping function are aligned with the two ends of the connection boundary interval and cross-fade-in and fade-out weighted operations are performed, the time domain spacing of adjacent pitch periods is recombined proportionally based on the audio speed ratio, and the local phase spectrum is locked to perform coherent superposition operation.

10. An emergency rescue video call method for people trapped in a malfunctioning elevator, characterized in that, The emergency rescue video call system for people trapped in a malfunctioning elevator according to any one of claims 1-9 includes the following steps: S1: Collect car image frames, calculate the signal-to-noise ratio change rate, filter the set of macroblock coordinates of motion regions where the motion vectors cross the boundary, increase the video coding quantization parameters when the signal-to-noise ratio change rate exceeds the channel attenuation threshold, perform spatial redundancy compression on macroblocks outside the set of macroblock coordinates of motion regions, and construct the encoded video bitstream; S2: Analyze the overlap between the packet loss marker bitmap fed back from the cloud and the key frame data segment in the encoded video bitstream, calculate the ratio of the maximum number of consecutive packet losses to the total number of key frame packets, obtain the key frame consecutive packet loss rate, and generate a key frame packet loss alarm log when the key frame packet loss threshold is exceeded. S3: Based on the key frame packet loss alarm log, when it is determined that the continuous packet loss rate of the key frame exceeds the standard, the original forward error correction redundancy is magnified by a multiplier to obtain asymmetric forward error correction redundancy, the verification packet sending interval is increased, and the original video payload is interleaved in time to construct an enhanced video bitstream with forward error correction. S4: Calculate the keyframe packet loss rate based on the enhanced video bitstream with forward error correction, generate a retransmission request message when the retransmission request trigger threshold is exceeded, monitor the network round-trip time and calculate the proportion of network round-trip time to video buffer time, determine the expected arrival time of the retransmission packet, and obtain link delay assessment data. S5: Based on the link delay assessment data, calculate the ratio of network round-trip delay to the remaining audio duration in the audio buffer to obtain the audio speed ratio, perform speed-changing without pitch-changing processing on the PCM audio data to generate a duration-extended audio data stream, shift the retransmission packet display timestamp to align with the end of the audio data stream, and output the audio and video synchronized playback sequence.

Images