Liquid crystal panel self-adaptive signal receiving and reconstructing method and system for emergency communication

Abstract

The invention discloses a liquid crystal panel self-adaptive signal receiving and reconstructing method and system for emergency communication, and relates to the technical field of emergency communication. According to the method, a signal-noise joint feature space is constructed, and adaptive filtering is performed by adopting variational mode decomposition, so that the problem of weak signal capture is solved; a grouping parameter adaptive adjustment mechanism is established based on a fuzzy control theory, and intelligent aggregation and recombination of signal fragments are realized in combination with an auto-encoder-clustering hybrid model; further performing compensation enhancement on the damaged signal through an adversarial learning model, and ensuring signal coherence by using a multi-scale convolution attention mechanism; finally, error correction and semantic recovery are achieved through distributed verification and a pre-training language model, a complete guarantee system from a physical layer to a semantic layer is formed, and the reliability and efficiency of emergency communication are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of emergency communication technology, specifically to a method and system for adaptive signal reception and reconstruction of a liquid crystal panel for emergency communication. Background Technology

[0002] Emergency communications are a crucial support for ensuring the safety of life and property during sudden disasters. Their importance lies in their ability to quickly restore information flow and prevent the escalation of secondary disasters. In extreme scenarios such as earthquakes and floods, traditional communication base stations are often the first to be damaged, leading to the interruption of rescue information transmission and becoming the biggest bottleneck restricting emergency response. Current emergency communication solutions mainly rely on dedicated receiving equipment or external antennas. Although these methods can cover basic signals, they have significant drawbacks: the equipment is bulky and inconvenient to carry, deployment requires professional personnel, and it is difficult to adapt to the ever-changing terrain in the field. More importantly, they are highly dependent on a stable power supply and a complete link. Once environmental electromagnetic interference increases or the signal path is blocked, the receiving performance drops sharply, making it impossible to flexibly cope with real-world scenarios with complex interference.

[0003] The core technical challenge in this field lies in the coupling between the electromagnetic response characteristics of the LCD panel pixel array and the dynamic grouping mechanism. The electromagnetic response characteristics of the pixel array determine the differences in the panel's sensitivity to capturing signals of different frequency bands. This characteristic makes it difficult for a single pixel to stably lock onto weak signals when noise intensity fluctuates. This further leads to the accuracy bottleneck of the dynamic grouping mechanism, namely, the group size adjustment lags behind environmental changes and cannot match the signal attenuation level in real time. For example, at a flood site, when a portable display screen faces a sudden shortwave signal, some pixels on the panel may respond out of tune due to noise interference, causing the grouping algorithm to misjudge the group size as too small. This results in fragmentation and breakage of the entire signal acquisition process, with rescue voice information being repeatedly lost during transmission, creating a continuous interruption of service. This service interruption not only prolongs the rescue window but also amplifies the chain reaction of disasters. Especially when multiple panels need to work together, the lack of a signal cross-verification mechanism further exacerbates the spread of interference, making the establishment of temporary communication links inefficient and making it difficult to restore voice or text communication within minutes. Therefore, how to achieve real-time adaptive capture of complex electromagnetic environments by pixel arrays without additional hardware, and accurately stitch together signal fragments to ensure continuous transmission, has become a key issue in the application of emergency communication LCD panels. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for adaptive signal reception and reconstruction of liquid crystal panels for emergency communication. By constructing a complete adaptive processing chain from signal preprocessing to semantic error correction, the invention enables the liquid crystal panel to intelligently receive and reconstruct emergency signals with high reliability in complex electromagnetic environments.

[0005] The objective of this invention can be achieved through the following technical solutions: This application provides a method for adaptive signal reception and reconstruction of a liquid crystal panel for emergency communication, comprising the following steps: S1. By collecting electromagnetic response data of the pixel array, a signal-noise joint feature space is constructed. An adaptive filtering algorithm based on variational mode decomposition is adopted to dynamically decompose and reconstruct the effective components according to the signal frequency band characteristics, and simultaneously extract the time-frequency features and signal-to-noise ratio of the signal. S2. Based on time-frequency characteristics and signal-to-noise ratio, an adaptive adjustment mechanism for grouping parameters is established through fuzzy control theory. The matching degree between signal attenuation and group size is quantified through membership function. When a mismatch between signal attenuation and group size is detected, the grouping strategy is dynamically adjusted based on feature similarity. S3. Input the signal fragments in the grouping strategy into the autoencoder-clustering hybrid model, learn the deep feature representation of the signal fragments through the encoder, perform feature clustering in the latent space, obtain the signal block features of intelligent aggregation and recombination, evaluate the grouping effect through the clustering results, and form a collaborative mechanism for mutual verification between grouping and clustering. S4. Based on the characteristics of the signal block, an adversarial learning model of interference intensity and signal quality is constructed. The generator network compensates and enhances the damaged signal block, and the discriminator network evaluates the enhancement quality to obtain the enhanced signal block. S5. Utilize the enhanced signal blocks and learn the correlation of boundary features of adjacent signal blocks through a multi-scale convolutional attention mechanism. During the reconstruction process, continuously monitor the signal continuity index. When a decrease in splicing quality is detected, immediately trigger the regrouping and clustering process. S6. Spatiotemporal double check codes are embedded in the generated continuous signal stream. An improved practical Byzantine fault-tolerant algorithm is used for distributed check between multiple panels. The error patterns discovered by the check are used to optimize the compensation strategy of the adversarial network, forming a dual guarantee of error prevention and correction.

[0006] This invention provides a liquid crystal panel adaptive signal receiving and reconstruction system for emergency communication, which implements a liquid crystal panel adaptive signal receiving and reconstruction method for emergency communication, including: The signal preprocessing and feature extraction module is used to collect electromagnetic response data of pixel array, perform signal denoising and reconstruction through variational mode decomposition and adaptive filtering algorithms, and simultaneously extract time-frequency features and signal-to-noise ratio indicators. The adaptive adjustment module for grouping strategy is based on fuzzy control theory to construct a grouping parameter adjustment mechanism. It quantifies the matching degree between signal attenuation and group size through membership function, and dynamically optimizes the grouping strategy using hierarchical clustering algorithm to match the grouping configuration with the current signal state. The deep feature clustering and verification module adopts an autoencoder-clustering hybrid model to perform deep feature clustering and intelligent recombination of signal fragments in the latent space. The grouping effect is evaluated by indicators such as intra-cluster variance and inter-cluster distance, forming a collaborative mechanism for mutual verification between grouping and clustering. The signal enhancement and compensation module constructs an adversarial learning model of interference intensity and signal quality. It uses a generator network to compensate and enhance damaged signal blocks, and a discriminator network to evaluate the enhancement quality. The multi-scale coherence reconstruction module uses a multi-scale convolutional attention mechanism to learn the correlation of boundary features of adjacent signal blocks and continuously monitors the signal continuity index. When a decrease in splicing quality is detected, a regrouping process is immediately triggered. The distributed verification and error optimization module embeds a spatiotemporal dual verification code into the signal stream, uses an improved Byzantine fault-tolerant algorithm for distributed verification between multiple panels, and optimizes the adversarial network compensation strategy by discovering error patterns, thus forming a dual guarantee of error prevention and correction. The semantic-level error correction and information generation module combines a pre-trained language model to perform semantic-level error correction and completion on the signal stream. It guides the adjustment of parameters throughout the process by identifying semantic anomalies during the recovery process, and finally outputs accurate and complete rescue information.

[0007] The beneficial effects of this invention are as follows: This invention effectively solves the problem of weak signal capture caused by the difference in electromagnetic response characteristics of pixel arrays in liquid crystal panels by constructing a signal-noise joint feature space and adopting an adaptive filtering algorithm based on variational mode decomposition. It can dynamically decompose and reconstruct effective components according to the signal frequency band characteristics, significantly improving the sensitivity of capturing weak emergency signals in complex electromagnetic environments. This enables liquid crystal panels to achieve stable signal reception without additional hardware, overcoming the defect of traditional equipment whose performance drops sharply when the signal path is blocked. By adopting fuzzy control theory to establish an adaptive adjustment mechanism for grouping parameters and combining it with an autoencoder-clustering hybrid model, the technical bottleneck of insufficient accuracy of dynamic grouping mechanism was successfully overcome. The matching degree between signal attenuation and group size was quantified in real time through membership function, and the grouping strategy was dynamically adjusted based on feature similarity. This solved the problem of group size adjustment lagging behind environmental changes, ensuring that signal fragments can be intelligently aggregated and reassembled, and effectively avoiding fragmentation and breakage of rescue information during transmission. By constructing a distributed verification system with a multi-scale convolutional attention mechanism and spatiotemporal dual check codes, a complete error prevention and correction guarantee system is formed. It can monitor signal continuity indicators in real time and immediately trigger a reprocessing process when a degradation in splicing quality is detected. At the same time, through distributed verification optimization compensation strategies among multiple panels, semantic-level error correction is finally achieved by combining a pre-trained language model. This greatly improves the reliability and integrity of emergency communication, ensures that usable voice or text communication can be restored within minutes, and significantly shortens the rescue response time. Attached Figure Description

[0008] To better understand and implement this application, the technical solution is described in detail below with reference to the accompanying drawings.

[0009] Figure 1 This is a flowchart illustrating an adaptive signal reception and reconstruction method for an emergency communication liquid crystal panel provided in Embodiment 1 of this application. Figure 2 This is a flowchart illustrating step S2 of a liquid crystal panel adaptive signal receiving and reconstruction method for emergency communication provided in Embodiment 1 of this application. Figure 3 This is a flowchart illustrating step S3 of a liquid crystal panel adaptive signal receiving and reconstruction method for emergency communication provided in Embodiment 1 of this application. Figure 4 This is a schematic diagram of a liquid crystal panel adaptive signal receiving and reconstruction system for emergency communication provided in Embodiment 2 of this application. Detailed Implementation

[0010] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, exemplary embodiments will be described in detail below, examples of which are illustrated in the accompanying drawings. In the following description, when referring to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application.

[0011] The terminology used in this application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. The singular forms “a,” “the,” and “the” used herein are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any and all possible combinations of one or more of the associated listed items.

[0012] The following detailed description of the specific implementation methods, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided in detail.

[0013] Example 1 Please see Figures 1-3 This embodiment provides a method for adaptive signal reception and reconstruction of a liquid crystal panel for emergency communication, including the following steps: S1. By collecting electromagnetic response data of the pixel array, a signal-noise joint feature space is constructed. An adaptive filtering algorithm based on variational mode decomposition is adopted to dynamically decompose and reconstruct the effective components according to the signal frequency band characteristics. The time-frequency features and signal-to-noise ratio of the signal are extracted simultaneously, and a feature transmission channel from signal preprocessing to intelligent grouping is established to provide feature input for subsequent grouping decisions. Further, step S1 specifically includes: Electromagnetic response data of the pixel array is collected, normalized using a standardization method to obtain a standardized dataset, and then signal and noise features are extracted to construct a joint feature space of signal and noise, resulting in a joint feature set. Variational mode decomposition algorithm is used to decompose the signal frequency band of the joint feature set to obtain the frequency band decomposition result. Effective components are extracted from the frequency band decomposition result and reconstructed using an adaptive filtering algorithm to obtain the reconstructed component set. Time-frequency characteristics and signal-to-noise ratio are calculated based on the reconstructed component set to obtain the feature index set.

[0014] In this process, when using variational mode decomposition (VMD) to decompose the signal frequency band of the joint feature set, key parameters are initialized based on the overall frequency distribution characteristics of the signals in the joint feature set. These parameters include the preset number of modes to be decomposed (i.e., the number of potential frequency band components), a penalty factor to balance reconstruction accuracy and computational efficiency, and a convergence threshold to control the stopping of iterations. Subsequently, a variational optimization model is constructed with the objective of minimizing the sum of the bandwidths of each mode signal and minimizing the error between the reconstructed signals and the original joint feature set signal, transforming the frequency band decomposition problem into a constrained optimization problem. Then, through cross-validation... The Alternating Direction Multiplier Method (ADMM) iteratively solves the optimization model. In each iteration, the time-domain signal and corresponding center frequency of each mode are updated, so that each mode gradually focuses on its own frequency band, while continuously reducing the reconstruction error with the original signal. When the iteration process meets the preset convergence condition (such as the change in the modal signal in two adjacent iterations being less than the convergence threshold), the iteration stops. The multiple independent modal signals with non-overlapping frequency bands obtained at this time are the final frequency band decomposition results. Each mode corresponds to a component of a different frequency range in the joint feature set signal.

[0015] After obtaining the reconstructed component set using an adaptive filtering algorithm, when calculating the time-frequency characteristics, a short-time Fourier transform or wavelet transform is applied to each reconstructed component. Signal segments are extracted through a sliding time window and Fourier analysis is performed to obtain the spectral distribution that changes over time. Features such as instantaneous frequency (frequency corresponding to the peak of the spectrum at each moment), amplitude spectral density (signal amplitude intensity at different frequency points), and time-frequency energy entropy (disorder of energy distribution in the time-frequency plane) are extracted from this data. When calculating the signal-to-noise ratio (SNR), the signal power is first estimated using the mean square value of the reconstructed components (assuming that the effective signal after reconstruction is the main component). Then, the residual noise power is estimated using the error signal obtained during the adaptive filtering process (the difference between the original component and the filtered component). Finally, the SNR in decibels is calculated. Ultimately, the time-frequency characteristics of all reconstructed components and their corresponding SNR values ​​are integrated into a feature index set.

[0016] Specifically, by constructing a signal-noise joint feature space and employing an adaptive filtering algorithm based on variational mode decomposition, the problem of weak signal capture caused by differences in pixel response characteristics in complex electromagnetic environments of LCD panels is effectively solved. This achieves accurate separation and reconstruction of non-stationary signals, significantly improving the reliability and quality of initial signal acquisition in emergency communications.

[0017] S2. Based on time-frequency characteristics and signal-to-noise ratio, an adaptive adjustment mechanism for grouping parameters is established through fuzzy control theory. The matching degree between signal attenuation and group size is quantified through membership function. When a mismatch between signal attenuation and group size is detected, the grouping strategy is dynamically adjusted based on feature similarity. Further, step S2 specifically includes: S21. Obtain the time-frequency characteristics and signal-to-noise ratio of the input signal, construct the input variables of the fuzzy control system, use a preset fuzzy rule base, quantify the matching degree between signal attenuation and group size through the membership function, and obtain the matching degree score; if the matching degree score is lower than the preset threshold, first extract the feature similarity from the time-frequency feature matrix, use the hierarchical clustering algorithm to perform a first clustering of the signal features, and generate a clustering tree containing multi-granularity grouping results (covering the clustering division under different group sizes). S22. Based on the clustering tree, select the results corresponding to the initial group size, calculate the centroid features of each cluster group, use Euclidean distance to evaluate the suitability of each group with the current group size, determine the group size adjustment parameters, extract the grouping results of the corresponding granularity from the clustering tree through the adjustment parameters (without re-clustering), and redistribute the signal groups using the weighted average method to obtain the updated grouping configuration. S23. Based on the updated grouping configuration, recalculate the time-frequency feature matrix and signal-to-noise ratio index, and quantify the matching degree again through the membership function to obtain a new matching degree score. If the new matching degree score is still lower than the preset threshold, iteratively adjust the weight of feature similarity, select a more suitable granular grouping result from the clustering tree (update the suitability evaluation standard based on the weight), and repeat the grouping strategy adjustment until the matching degree score reaches the preset threshold to obtain the final grouping strategy.

[0018] The matching degree between signal attenuation and group size is quantified by a membership function to obtain a matching degree score. This includes defining signal attenuation and group size as fuzzy subsets, with signal attenuation divided into low, medium, and high attenuation based on signal-to-noise ratio, and group size divided into small, medium, and large based on the number of signal fragments. The membership degree of the current signal attenuation is calculated using a Gaussian membership function, and the membership degree of the current group size is calculated using a triangular membership function. Then, corresponding rules are triggered according to a preset fuzzy rule base and the trigger strength is determined. Finally, the fuzzy output is defuzzified using the centroid method to convert it into a matching degree score of 0-100.

[0019] The clustering tree is used to locate the hierarchical division that matches the initial group size. All clusters under that level are extracted. The mean of all signal feature vectors in each cluster is then calculated to obtain the center point features of each group. Finally, the Euclidean distance between the center point features of each cluster and the ideal center is calculated with reference to the ideal feature center corresponding to the initial group size. The smaller the distance, the higher the fit between the group and the current group size.

[0020] Specifically, by establishing a fuzzy control-based adaptive adjustment mechanism for grouping parameters, the problem of mismatched grouping sizes caused by dynamic signal attenuation in emergency communications was effectively solved, achieving real-time and accurate matching between grouping strategies and signal quality, and significantly improving the efficiency and accuracy of signal fragment aggregation.

[0021] S3. Input the signal fragments in the grouping strategy into the autoencoder-clustering hybrid model, learn the deep feature representation of the signal fragments through the encoder, perform feature clustering in the latent space, obtain the signal block features of intelligent aggregation and recombination, evaluate the grouping effect through the clustering results, and form a collaborative mechanism for mutual verification between grouping and clustering. Further, step S3 specifically includes: S31. Obtain signal fragments in the grouping strategy, including temporal markers and spatial location information of pixel arrays. Encode them using a stacked autoencoder with residual connections. The encoder extracts local time-frequency texture features of the signal fragments through convolutional layers. After the dimension is compressed by pooling layers, it is mapped to a low-dimensional latent space through fully connected layers to generate a deep feature representation that integrates temporal correlation, spatial distribution features and noise robustness. The decoder optimizes the encoder parameters in reverse through reconstruction error to ensure that the features retain the key structural information of the original signal. S32. Perform improved K-means clustering on the deep feature representation in the latent space. Initialize the centroid based on the temporal continuity of signal fragments to avoid intra-cluster temporal disorder caused by random initialization. Use weighted Euclidean distance to calculate feature similarity, where the weights are dynamically adjusted with the signal-to-noise ratio. High signal-to-noise ratio features are given higher weights. Iteratively optimize the cluster partitioning to obtain the feature clustering results. S33. Calculate the intra-cluster variance of each cluster. If the variance is lower than the preset threshold, the cluster features are reorganized into aggregated signal block features according to temporal and spatial correlation. If the variance is higher than the threshold, the model is retrained by increasing the latent space dimension of the autoencoder and adjusting the convolution kernel size to generate a more discriminative deep feature representation. Among them, the intra-cluster variance is the weighted variance of the fusion of time-frequency feature fluctuation and spatial location deviation; the preset threshold is dynamically set according to the minimum effective continuous length of the emergency communication signal; the convolution kernel size is used to enhance the local feature capture capability; S34. The grouping effect is evaluated by using aggregated signal block features. The cosine distance between clusters (quantifying the feature differences between different signal blocks) and the average density within clusters (the mean Euclidean distance between features within a cluster and the cluster center, reflecting the consistency within the block) are calculated to form a quantitative index of grouping effect (the larger the distance between clusters and the smaller the density within clusters, the better the grouping effect). A collaborative verification mechanism is constructed to calculate the verification consistency score between the clustering results and the original grouping strategy from three dimensions: feature distribution overlap (the crossover ratio of the feature space of the clustered clusters and the original grouping), temporal continuity matching degree (the time stamp continuity deviation of fragments within the signal block), and signal-to-noise ratio distribution consistency (the standard deviation of the signal-to-noise ratio within the same group / cluster). If the score is higher than the preset threshold, the grouping and clustering are deemed to be effective, and the final signal block features are output.

[0022] When evaluating the grouping effect using aggregated signal block features, the inter-cluster cosine distance between the depth feature vectors of different signal blocks (corresponding to clusters) is first calculated. The larger the distance value, the more significant the feature differences between different signal blocks and the higher the discriminative power. Then, the mean Euclidean distance between all features within each signal block and the center of that cluster is calculated to obtain the average cluster compactness. The smaller this value, the stronger the consistency of features within the block and the more stable the aggregation effect. Finally, the inter-cluster cosine distance and the average cluster compactness are used as core dimensions to form a quantitative index of grouping effect. Both of them together satisfy the condition that the inter-cluster distance is large and the cluster compactness is small, indicating that the grouping strategy can effectively distinguish different signal blocks and ensure the coherence of features within the block, resulting in a better grouping effect.

[0023] Specifically, by using an autoencoder-clustering hybrid model to perform deep feature learning and intelligent recombination of signal fragments, the problem of signal fragmentation caused by grouping strategy bias is effectively solved, and the accurate aggregation and integrity reconstruction of signal blocks are achieved, significantly improving the accuracy and reliability of information recovery in emergency communications.

[0024] S4. Based on the characteristics of the signal block, an adversarial learning model of interference intensity and signal quality is constructed. The generator network compensates and enhances the damaged signal block, and the discriminator network evaluates the enhancement quality to obtain the enhanced signal block. Further, step S4 specifically includes: The signal block features are obtained, including time-frequency distribution, spatial correlation and clustering labels. Principal component analysis is used to extract key feature vectors from the signal blocks to form a signal block feature vector set for interference intensity and signal quality assessment. If the interference intensity index in the feature vector set exceeds a preset threshold, it is marked as a damaged signal block. An adversarial learning model of interference intensity-signal quality is constructed. The damaged signal block is input into the generator network of the generative adversarial network. The generator generates a targeted compensation signal by learning the time-frequency-spatial distribution law of the normal signal block. After being fused with the damaged signal block, a preliminary enhanced signal block is obtained. The interference intensity index is calculated based on the time-frequency energy distortion rate and spatial noise distribution. First, using the time-frequency energy distribution of a normal signal block as a benchmark, the absolute difference between the energy value of the damaged signal block at the same time-frequency point and the benchmark value is calculated. All differences are summed and divided by the total benchmark energy to obtain the time-frequency energy distortion rate. The larger the value, the more severe the interference in the time-frequency domain. Next, for the LCD panel pixel array corresponding to the signal block, the square of the deviation between the signal value of each pixel and the mean value of normal pixels is calculated, which is the noise energy. The proportion of pixels with noise energy exceeding a preset noise threshold is counted to obtain the spatial noise distribution coefficient. The higher the coefficient proportion, the wider the range of spatial domain interference. Finally, the two are weighted and summed according to preset weights, and the result is the interference intensity index. The initial enhanced signal block is evaluated by a discriminator network, and a binary classifier is used to determine the similarity between the enhanced signal block and the original signal to obtain an enhancement quality score. If the enhancement quality score is lower than a preset threshold, the compensation parameters are readjusted by a generator network to generate an enhanced signal block.

[0025] The constructed interference strength-signal quality adversarial learning model includes: extracting key feature vectors from signal block features through principal component analysis, and calculating interference strength indices based on time-frequency energy distortion rate and spatial noise distribution; when a damaged signal block is identified, the generator network generates a targeted compensation signal and fuses it with the damaged signal by deep learning the time-frequency-spatial distribution pattern of the normal signal block; subsequently, the discriminator network performs binary classification evaluation on the enhanced signal block and calculates its similarity score with the original signal; if the quality score is not up to standard, the compensation parameters of the generator will be readjusted for iterative optimization until an enhanced signal block that meets the quality requirements is output, thereby achieving intelligent repair and quality improvement of the damaged signal.

[0026] Specifically, by constructing an adversarial learning model of interference intensity and signal quality, the problem of signal block damage caused by electromagnetic interference in emergency communications is effectively solved, and intelligent compensation and quality enhancement of damaged signals are achieved, significantly improving the accuracy and reliability of signal reconstruction in strong interference environments.

[0027] S5. Utilize the enhanced signal blocks and learn the correlation of boundary features of adjacent signal blocks through a multi-scale convolutional attention mechanism. During the reconstruction process, continuously monitor the signal continuity index. When a decrease in splicing quality is detected, immediately trigger the regrouping and clustering process to ensure the integrity and coherence of the signal flow. Further, step S5 specifically includes: The enhanced signal blocks (carrying time stamps, spatial coordinates, and enhanced feature labels) are obtained. A multi-scale convolutional network is used to extract features in the boundary regions of adjacent signal blocks (feature fragments of 20% length of the block edge, containing time-frequency and spatial pixel domain information). A 3×3 convolutional kernel focuses on local fine boundary features (such as the amplitude gradient of time-frequency jump points and the spatial coordinate offset of pixel value abrupt changes), and a 5×5 convolutional kernel captures wide-range boundary correlation features (such as the consistency of frequency drift trends of adjacent blocks and the gradient continuity of spatial pixel distribution). After batch normalization and activation function processing, a boundary feature matrix that integrates multi-scale information is generated. The feature matrix is ​​dynamically weighted and optimized through a convolutional attention mechanism. The importance weights of each feature channel are calculated through global average pooling and fully connected layers (prioritizing the weights of channels with concentrated time-frequency energy and channels with significant spatial correlation). Spatial weight maps are generated through max pooling and average pooling in the channel dimension. Higher weights are assigned to key locations in the boundary region (such as signal abrupt change points and high noise interference areas). Finally, the feature correlation weights that fuse channel priority and spatial criticality are obtained, quantifying the matching strength of adjacent block boundaries in terms of time-frequency trends and spatial distribution. In the process of dynamically weighting and optimizing the boundary feature matrix generated by multi-scale convolution, the spatial dimension information of each channel in the feature matrix is ​​first compressed into a single-valued feature through global average pooling (preserving the global time-frequency and spatial information at the channel level). This single-valued feature is then input into a fully connected network consisting of a dimensionality reduction layer, an activation function, and an up-dimensionality layer. The dimensionality reduction layer reduces the feature dimension to lower computational complexity, the activation function introduces a nonlinear mapping, and the up-dimensionality layer outputs weight coefficients consistent with the number of channels in the original feature matrix. These weight coefficients will prioritize strengthening channels with concentrated time-frequency energy and significant spatial correlation (corresponding to the channels where key signal features are located in emergency communication), thereby achieving dynamic weighting of the channel dimension of the feature matrix and allowing subsequent calculations to focus more on the core information of boundary correlation.

[0028] The signal continuity index is calculated based on feature correlation weights. The index consists of three parts: boundary feature matching degree (the weighted average of correlation weights, which are positively correlated with the feature channel contribution rate, reflecting the continuity at the feature level), temporal continuity (the ratio of the difference in timestamps between adjacent blocks to the preset minimum effective interval; a ratio ≤1 is considered acceptable, assessing the continuity in the temporal dimension), and spatial connection error (the root mean square error of pixel values ​​in overlapping areas, reflecting the splicing accuracy in the spatial dimension). These are weighted and summed according to the priority of emergency communication scenarios to obtain a comprehensive continuity index. The comprehensive continuity index is compared with a preset threshold: if the index meets the standard, the signal blocks are spliced ​​into a complete signal stream according to the mapping relationship between the timestamps in ascending order and the spatial coordinates; if the index is lower than the threshold (indicating a decline in splicing quality), a regrouping and clustering process is immediately triggered.

[0029] Specifically, this includes feeding back the boundary feature matching deviation, temporal breakpoint coordinates, and spatial connection error value to S2 to optimize the grouping strategy (such as reducing the group size to reduce cross-block boundaries and adjusting feature similarity weights to strengthen temporal correlation features); simultaneously feeding back the current latent space feature distribution deviation to S3 to readjust the clustering parameters (such as optimizing the K value and updating the distance metric weights); after being enhanced again by S4, returning to this step for re-evaluation, iterating until the continuity index meets the standard, and finally outputting a complete and coherent signal stream that meets the requirements of emergency communication; The complete signal stream carries a timing stamp, LCD panel spatial coordinates, and enhanced feature tags optimized by S4. It has passed multiple checks on timing continuity, spatial connection accuracy, and boundary feature matching degree, meeting the basic requirements of emergency communication for signal integrity and coherence.

[0030] Specifically, by using a multi-scale convolutional attention mechanism to perform correlation analysis on the boundary features of signal blocks and establishing a real-time continuity monitoring mechanism, the problem of splicing inconsistencies and breaks in the signal reconstruction process was effectively solved, achieving complete reconstruction and stable transmission of the signal stream, and significantly improving the reliability and continuity of information transmission in emergency communications.

[0031] S6. Spatiotemporal double check codes are embedded in the generated continuous signal stream. An improved practical Byzantine fault-tolerant algorithm is used for distributed verification between multiple panels. The error patterns discovered by the verification are used to optimize the compensation strategy of the adversarial network, forming a dual guarantee of error prevention and correction. Further, step S6 specifically includes: Continuous signal stream data is acquired, and a spatiotemporal check code embedding method is used to insert timestamps and spatial identifiers into data packets to generate check code embedded signal streams. Distributed verification is performed among multiple panels using the Byzantine fault-tolerant algorithm to compare the consistency of check codes of each panel. If the consistency is lower than a preset threshold, it is marked as a potential error signal, thus obtaining a set of error signals. In the distributed verification process among multiple panels, one master panel and multiple slave panels are first designated from the LCD panels participating in emergency communication. The total number of panels meets the fault tolerance requirements of the improved practical Byzantine fault-tolerant algorithm, i.e., at least 3f+1, where f is the allowed number of faulty panels. The master panel first extracts the spatiotemporal check code from the continuous signal stream. The spatiotemporal check code contains a timestamp and a spatial identifier, and sends a verification request and its own calculated check code to all slave panels. After receiving the request, each slave panel independently calculates the spatiotemporal check code based on the same segment of continuous signal stream stored locally, and then feeds back the calculation result to the master panel. After the master panel collects the check codes of all panels (including itself), it counts the proportion of identical check codes. If the proportion is ≥ a preset consistency threshold (e.g., 2 / 3, which meets the consensus ratio of Byzantine fault tolerance), the check codes are considered to be consistent. If the proportion is lower than the threshold, the panel feedback results that are different from the majority of check codes are further marked, and the corresponding signal segment is initially identified as a potential error signal.

[0032] For the error signal set, a pattern analysis algorithm is used to extract error pattern features and generate error pattern descriptions. An adversarial network model is used to input error features, adjust network parameters, and generate compensation strategy parameters. The spatiotemporal check code embedding rules are updated through the compensation strategy parameters. If the error pattern features deviate from the preset range, the check code density is adjusted to obtain the updated check code embedding signal stream. The updated checksum is embedded in the signal stream, and distributed verification is performed again among multiple panels to determine whether the error is lower than the preset threshold. A verification pass signal stream is obtained. Once the verification pass signal stream is passed, the compensation strategy parameters and checksum embedding rules are solidified to generate an error prevention and correction mechanism.

[0033] Specifically, by introducing spatiotemporal dual check codes and an improved Byzantine fault tolerance mechanism, a multi-panel collaborative distributed check system was constructed, which effectively solved the problems of insufficient reliability of single-panel check and error propagation, realized early detection and accurate location of errors, and continuously optimized the compensation strategy through error pattern learning, forming a dual guarantee of forward prevention and backward correction, which significantly improved the overall fault tolerance capability and signal transmission reliability of the emergency communication system.

[0034] S7. Combining a pre-trained language model, semantic-level error correction and completion are performed on the verified signal stream. The semantic anomalies discovered during the recovery process guide the parameter adjustment of the entire signal processing process, and output accurate and complete rescue information.

[0035] Further, step S7 specifically includes: Semantic analysis is performed on the verified signal stream using a pre-trained language model to identify semantic errors and omissions, resulting in a set of semantic anomalies. If the set of semantic anomalies is not empty, an anomaly detection algorithm is used to locate the error type and the location of the omission, and to determine the anomaly feature set. The signal processing parameters are adjusted according to the abnormal feature set, the signal flow is optimized by an adaptive filter to obtain the optimized signal flow, semantic features are extracted from the optimized signal flow, and a pre-trained language model is used for semantic completion to generate complete semantic content. If the complete semantic content matches the preset semantic template, rescue information is generated through model reasoning to obtain preliminary rescue information. The semantic content of the preliminary rescue information and the signal context are integrated using an information generation algorithm to generate accurate rescue information. The completeness and consistency of the accurate rescue information are verified through data stream processing, and the final rescue information is output.

[0036] Specifically, by introducing a pre-trained language model to perform semantic-level analysis and processing on the verified signal stream, the problem of semantic errors and omissions in traditional methods is effectively solved, and intelligent error correction and completion of rescue information is achieved, significantly improving the accuracy and completeness of information content in emergency communications.

[0037] Example 2 Please see Figure 4 This embodiment provides an adaptive signal receiving and reconstruction system for liquid crystal panels used in emergency communication, which implements an adaptive signal receiving and reconstruction method for liquid crystal panels used in emergency communication, including: The signal preprocessing and feature extraction module is used to collect electromagnetic response data of pixel array, perform signal denoising and reconstruction through variational mode decomposition and adaptive filtering algorithms, and simultaneously extract time-frequency features and signal-to-noise ratio indicators. The adaptive adjustment module for grouping strategy is based on fuzzy control theory to construct a grouping parameter adjustment mechanism. It quantifies the matching degree between signal attenuation and group size through membership function, and dynamically optimizes the grouping strategy using hierarchical clustering algorithm to ensure that the grouping configuration is best matched with the current signal state. The deep feature clustering and verification module adopts an autoencoder-clustering hybrid model to perform deep feature clustering and intelligent recombination of signal fragments in the latent space. The grouping effect is evaluated by indicators such as intra-cluster variance and inter-cluster distance, forming a collaborative mechanism for mutual verification between grouping and clustering. The signal enhancement and compensation module constructs an adversarial learning model of interference intensity and signal quality. It uses a generator network to compensate and enhance damaged signal blocks, and a discriminator network to evaluate the enhancement quality, thereby achieving effective repair of damaged signals. The multi-scale coherence reconstruction module uses a multi-scale convolutional attention mechanism to learn the correlation of boundary features of adjacent signal blocks and continuously monitors the signal continuity index. When a degradation in splicing quality is detected, a regrouping process is immediately triggered to ensure the integrity and coherence of the signal flow. The distributed verification and error optimization module embeds a spatiotemporal dual verification code into the signal stream, uses an improved Byzantine fault-tolerant algorithm for distributed verification between multiple panels, and optimizes the adversarial network compensation strategy by discovering error patterns, thus forming a dual guarantee of error prevention and correction. The semantic-level error correction and information generation module combines a pre-trained language model to perform semantic-level error correction and completion on the signal stream. It guides the adjustment of parameters throughout the process by identifying semantic anomalies during the recovery process, and finally outputs accurate and complete rescue information.

[0038] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for adaptive signal reception and reconstruction of a liquid crystal panel for emergency communication, characterized in that: Includes the following steps: S1. By collecting electromagnetic response data of the pixel array, a signal-noise joint feature space is constructed. An adaptive filtering algorithm based on variational mode decomposition is adopted to dynamically decompose and reconstruct the effective components according to the signal frequency band characteristics, and simultaneously extract the time-frequency features and signal-to-noise ratio of the signal. S2. Based on time-frequency characteristics and signal-to-noise ratio, an adaptive adjustment mechanism for grouping parameters is established through fuzzy control theory. The matching degree between signal attenuation and group size is quantified through membership function. When a mismatch between signal attenuation and group size is detected, the grouping strategy is dynamically adjusted based on feature similarity. S3. Input the signal fragments in the grouping strategy into the autoencoder-clustering hybrid model, learn the deep feature representation of the signal fragments through the encoder, perform feature clustering in the latent space, obtain the signal block features of intelligent aggregation and recombination, evaluate the grouping effect through the clustering results, and form a collaborative mechanism for mutual verification between grouping and clustering. S4. Based on the characteristics of the signal block, an adversarial learning model of interference intensity and signal quality is constructed. The generator network compensates and enhances the damaged signal block, and the discriminator network evaluates the enhancement quality to obtain the enhanced signal block. S5. Utilize the enhanced signal blocks and learn the correlation of boundary features of adjacent signal blocks through a multi-scale convolutional attention mechanism. During the reconstruction process, continuously monitor the signal continuity index. When a decrease in splicing quality is detected, immediately trigger the regrouping and clustering process. S6. Spatiotemporal double check codes are embedded in the generated continuous signal stream. An improved practical Byzantine fault-tolerant algorithm is used for distributed check between multiple panels. The error patterns discovered by the check are used to optimize the compensation strategy of the adversarial network, forming a dual guarantee of error prevention and correction.

2. The method for adaptive signal reception and reconstruction of a liquid crystal panel for emergency communication according to claim 1, characterized in that: Step S1 specifically includes: Electromagnetic response data of the pixel array is collected, normalized using a standardization method to obtain a standardized dataset, and then signal and noise features are extracted to construct a joint feature space of signal and noise, resulting in a joint feature set. Variational mode decomposition algorithm is used to decompose the signal frequency band of the joint feature set to obtain the frequency band decomposition result. Effective components are extracted from the frequency band decomposition result and reconstructed using an adaptive filtering algorithm to obtain the reconstructed component set. Time-frequency characteristics and signal-to-noise ratio are calculated based on the reconstructed component set to obtain the feature index set.

3. The method for adaptive signal reception and reconstruction of a liquid crystal panel for emergency communication according to claim 1, characterized in that: Step S2 specifically includes: S21. Obtain the time-frequency characteristics and signal-to-noise ratio of the input signal, construct the input variables of the fuzzy control system, use the preset fuzzy rule base, quantify the matching degree between signal attenuation and group size through the membership function, obtain the matching degree score, extract the feature similarity from the time-frequency feature matrix, use the hierarchical clustering algorithm to perform a clustering of the signal features, and generate a clustering tree containing multi-granularity grouping results. S22. Based on the clustering tree, select the results corresponding to the initial group size, calculate the centroid features of each cluster group, use Euclidean distance to evaluate the suitability of each group with the current group size, determine the group size adjustment parameters, extract the grouping results of the corresponding granularity from the clustering tree through the adjustment parameters, and redistribute the signal groups using the weighted average method to obtain the updated grouping configuration. S23. Based on the updated grouping configuration, recalculate the time-frequency feature matrix and signal-to-noise ratio index, and quantify the matching degree again through the membership function to obtain a new matching degree score. If the new matching degree score is still lower than the preset threshold, iteratively adjust the weight of feature similarity, select a more suitable granular grouping result from the clustering tree, and repeat the grouping strategy adjustment until the matching degree score reaches the preset threshold to obtain the final grouping strategy.

4. The method for adaptive signal reception and reconstruction of a liquid crystal panel for emergency communication according to claim 1, characterized in that: Step S3 specifically includes: S31. Obtain signal fragments in the grouping strategy and encode them using a stacked autoencoder with residual connections. The encoder extracts the local time-frequency texture features of the signal fragments through convolutional layers, compresses the dimension through pooling layers, and maps them to a low-dimensional latent space through fully connected layers to generate a deep feature representation that integrates temporal correlation, spatial distribution features and noise robustness. S32. Perform improved K-means clustering on the deep feature representation in the latent space, initialize the centroid based on the temporal continuity of signal fragments, calculate the feature similarity using weighted Euclidean distance, iteratively optimize the cluster partitioning, and obtain the feature clustering results. S33. Calculate the intra-cluster variance of each cluster. If the variance is lower than the preset threshold, the cluster features are reorganized into aggregated signal block features according to temporal and spatial correlation. If the variance is higher than the threshold, the model is retrained by increasing the latent space dimension of the autoencoder and adjusting the convolution kernel size to generate a more discriminative deep feature representation. S34. The grouping effect is evaluated by using aggregated signal block features. The cosine distance between clusters and the average density within clusters are calculated to form a quantitative index of grouping effect. A collaborative verification mechanism is constructed. The verification consistency score between the clustering results and the original grouping strategy is calculated from three dimensions: feature distribution overlap, temporal continuity matching degree, and signal-to-noise ratio distribution consistency. If the score is higher than the preset threshold, the grouping and clustering are deemed to be effective, and the final signal block features are output.

5. The method for adaptive signal reception and reconstruction of a liquid crystal panel for emergency communication according to claim 1, characterized in that: Step S4 specifically includes: The signal block features, including time-frequency distribution, spatial correlation and clustering labels, are obtained. Principal component analysis is used to extract key feature vectors from the signal blocks to form a signal block feature vector set for interference intensity and signal quality assessment. If the interference intensity index in the feature vector set exceeds a preset threshold, it is marked as a damaged signal block. The damaged signal block is input into the generator network of the generative adversarial network. The generator generates a targeted compensation signal by learning the time-frequency-spatial distribution law of the normal signal block. After being fused with the damaged signal block, a preliminary enhanced signal block is obtained. The initial enhanced signal block is evaluated by a discriminator network, and a binary classifier is used to determine the similarity between the enhanced signal block and the original signal to obtain an enhancement quality score. If the enhancement quality score is lower than a preset threshold, the compensation parameters are readjusted by a generator network to generate an enhanced signal block.

6. The method for adaptive signal reception and reconstruction of a liquid crystal panel for emergency communication according to claim 1, characterized in that: Step S5 specifically includes: The enhanced signal blocks are obtained, and a multi-scale convolutional network is used to extract hierarchical features from the boundary regions of adjacent signal blocks. A 3×3 convolutional kernel is used to focus on local fine boundary features, and a 5×5 convolutional kernel is used to capture wide-range boundary correlation features. After batch normalization and activation function processing, a boundary feature matrix that integrates multi-scale information is generated. The feature matrix is ​​dynamically weighted and optimized through convolutional attention mechanism. The importance weight of each feature channel is calculated through global average pooling and fully connected layers. A spatial weight map is generated through max pooling and average pooling in the channel dimension. Higher weights are assigned to key positions in the boundary region to obtain feature correlation weights that fuse channel priority and spatial keyness. Signal continuity indicators are calculated based on feature correlation weights, including boundary feature matching degree, temporal continuity and spatial connection error. The weighted sums are calculated according to the priority of emergency communication scenarios to obtain a comprehensive continuity indicator. The comprehensive continuity indicator is compared with a preset threshold. If the indicator meets the standard, the signal blocks are spliced ​​into a complete signal stream according to the mapping relationship between the time stamp and spatial coordinates. If the indicator is lower than the threshold, the regrouping and clustering process is triggered immediately.

7. The method for adaptive signal reception and reconstruction of a liquid crystal panel for emergency communication according to claim 1, characterized in that: Step S6 specifically includes: Continuous signal stream data is acquired, and a spatiotemporal check code embedding method is used to insert timestamps and spatial identifiers into data packets to generate check code embedded signal streams. Distributed verification is performed among multiple panels using the Byzantine fault-tolerant algorithm to compare the consistency of check codes of each panel. If the consistency is lower than a preset threshold, it is marked as a potential error signal, thus obtaining a set of error signals. Error pattern features are extracted using a pattern analysis algorithm to generate error pattern descriptions. An adversarial network model is used to input error features, adjust network parameters, and generate compensation strategy parameters. The spatiotemporal check code embedding rules are updated using the compensation strategy parameters. If the error pattern features deviate from the preset range, the check code density is adjusted to obtain the updated check code embedding signal stream. The updated checksum is embedded in the signal stream, and distributed verification is performed again among multiple panels to determine whether the error is lower than the preset threshold. A verification pass signal stream is obtained. Once the verification pass signal stream is passed, the compensation strategy parameters and checksum embedding rules are solidified to generate an error prevention and correction mechanism.

8. The method for adaptive signal reception and reconstruction of a liquid crystal panel for emergency communication according to claim 1, characterized in that: Also includes: S7. Combining a pre-trained language model, semantic-level error correction and completion are performed on the verified signal stream. The semantic anomalies discovered during the recovery process guide the parameter adjustment of the entire signal processing process, and output accurate and complete rescue information.

9. A method for adaptive signal reception and reconstruction of a liquid crystal panel for emergency communication according to claim 8, characterized in that: Step S7 specifically includes: Semantic analysis is performed on the verified signal stream using a pre-trained language model to identify semantic errors and omissions, resulting in a set of semantic anomalies. If the set of semantic anomalies is not empty, an anomaly detection algorithm is used to locate the error type and the location of the omission, and to determine the anomaly feature set. The signal processing parameters are adjusted according to the abnormal feature set, the signal flow is optimized by an adaptive filter to obtain the optimized signal flow, semantic features are extracted from the optimized signal flow, and a pre-trained language model is used for semantic completion to generate complete semantic content. If the complete semantic content matches the preset semantic template, rescue information is generated through model reasoning to obtain preliminary rescue information. The semantic content of the preliminary rescue information and the signal context are integrated using an information generation algorithm to generate accurate rescue information. The completeness and consistency of the accurate rescue information are verified through data stream processing, and the final rescue information is output.

10. A liquid crystal panel adaptive signal receiving and reconstruction system for emergency communication, used to implement the liquid crystal panel adaptive signal receiving and reconstruction method for emergency communication as described in any one of claims 1-9, characterized in that: include: The signal preprocessing and feature extraction module is used to collect electromagnetic response data of pixel array, perform signal denoising and reconstruction through variational mode decomposition and adaptive filtering algorithms, and simultaneously extract time-frequency features and signal-to-noise ratio indicators. The adaptive adjustment module for grouping strategy is based on fuzzy control theory to construct a grouping parameter adjustment mechanism. It quantifies the matching degree between signal attenuation and group size through membership function, and dynamically optimizes the grouping strategy using hierarchical clustering algorithm to match the grouping configuration with the current signal state. The deep feature clustering and verification module adopts an autoencoder-clustering hybrid model to perform deep feature clustering and intelligent recombination of signal fragments in the latent space. The grouping effect is evaluated by intra-cluster variance and inter-cluster distance, forming a collaborative mechanism for mutual verification between grouping and clustering. The signal enhancement and compensation module constructs an adversarial learning model of interference intensity and signal quality. It uses a generator network to compensate and enhance damaged signal blocks, and a discriminator network to evaluate the enhancement quality. The multi-scale coherence reconstruction module uses a multi-scale convolutional attention mechanism to learn the correlation of boundary features of adjacent signal blocks and continuously monitors the signal continuity index. When a decrease in splicing quality is detected, a regrouping process is immediately triggered. The distributed verification and error optimization module embeds a spatiotemporal dual verification code into the signal stream, uses an improved Byzantine fault-tolerant algorithm for distributed verification between multiple panels, and optimizes the adversarial network compensation strategy by discovering error patterns, thus forming a dual guarantee of error prevention and correction. The semantic-level error correction and information generation module combines a pre-trained language model to perform semantic-level error correction and completion on the signal stream. It guides the adjustment of parameters throughout the process by identifying semantic anomalies during the recovery process, and finally outputs accurate and complete rescue information.

Images