A data transmission method of a wireless communication device in an industrial scenario

Abstract

The application discloses a data transmission method of a wireless communication device in an industrial scene and belongs to the technical field of industrial wireless communication. The method comprises the following steps: obtaining historical transmission data packets of a plurality of wireless communication devices in an industrial communication network, performing time-frequency domain decomposition on the historical transmission data packets, extracting a plurality of time-frequency domain sparse characteristic values corresponding to each historical transmission data packet to form an original characteristic value sequence, selecting a reference characteristic value according to the statistical distribution characteristics of the original characteristic value sequence, calculating the offset of each characteristic value to generate an offset sequence, constructing an adaptive quantization interval according to the numerical fluctuation range of the offset sequence, mapping the offset sequence to an index code to obtain a compressed characteristic index sequence, replacing original load data with the compressed characteristic index sequence to generate a compressed data frame, and performing data transmission between wireless communication devices based on the compressed data frame.

Description

Technical Field

[0001] This invention relates to the field of industrial wireless communication technology, specifically to a data transmission method for wireless communication devices in industrial settings. Background Technology

[0002] Industrial wireless communication equipment plays a crucial role in real-time data exchange on automated production lines, transmitting payload data such as sensor readings, control commands, and equipment status information. Traditional data transmission methods directly send raw payload data packets, which consume significant channel resources when the data volume is large. In densely deployed industrial wireless networks, this can easily lead to channel congestion and transmission delays, impacting production cycle time and equipment coordination efficiency.

[0003] Existing technologies include compression schemes for industrial wireless data transmission. One type uses general data compression algorithms to compress the payload data as a whole. However, general compression algorithms have high computational overhead, making it difficult to guarantee real-time performance on resource-constrained industrial embedded wireless devices. Furthermore, the compression ratio is significantly limited by the redundancy of the payload data itself. Another type of scheme simplifies specific industrial protocol fields, such as omitting fixed padding bytes or shortening address fields. This type of compression relies on prior knowledge of the protocol format, resulting in limited compression effectiveness and a lack of universality. In addition, existing compression methods often use fixed quantization parameters or fixed code tables, failing to dynamically adjust the compression strategy based on changes in the characteristics of the data packets. This makes it difficult to simultaneously achieve compression efficiency and recovery accuracy in complex, time-varying industrial channel environments.

[0004] Industrial wireless transmission data packets exhibit sparse characteristics in the time-frequency domain, meaning energy is concentrated around a few time-frequency points. Traditional compression methods do not fully utilize this sparsity for efficient representation. How to capture these sparse time-frequency features from historical data packets and perform compact encoding constitutes a problem to be solved. During encoding, the numerical distribution of the offset sequence fluctuates with changes in the communication environment and equipment operating conditions. Using a fixed quantization interval can lead to increased quantization errors or coding redundancy. How to adaptively match the division of the quantization interval to the actual fluctuation state of the offset sequence constitutes another problem to be solved. Summary of the Invention

[0005] To address the issues of high channel resource consumption, insufficient real-time compression, and poor adaptability caused by fixed quantization parameters in data transmission of industrial wireless communication equipment, a data transmission method based on time-frequency domain sparse feature extraction and adaptive quantization coding is proposed. By mining the time-frequency sparsity of the payload data and adaptively quantizing the offset, the method can reduce the amount of data transmission while adapting to the dynamically changing industrial channel environment.

[0006] To achieve the above objectives, the present invention provides the following technical solution: The present invention provides a data transmission method for wireless communication devices in industrial scenarios. At the transmitting end, historical transmission data packets from multiple wireless communication devices within an industrial communication network are acquired. The historical transmission data packets are decomposed in the time-frequency domain, and multiple time-frequency sparse feature values ​​corresponding to each historical transmission data packet are extracted to form an original feature value sequence. A benchmark feature value is selected based on the statistical distribution characteristics of the original feature value sequence, and the offset of each feature value from the benchmark feature value is calculated to generate an offset sequence. An adaptive quantization interval is constructed based on the numerical fluctuation range of the offset sequence, and the offset sequence is mapped to an index code to obtain a compressed feature index sequence. The compressed feature index sequence is used to replace the original payload data in the historical transmission data packets to generate compressed data frames, and data transmission is performed between wireless communication devices based on the compressed data frames. This method utilizes the time-frequency sparse characteristics of historical transmission data packets to compress the original payload data containing a large amount of redundant information into a feature index sequence. It replaces the large payload field with a very low index code, significantly reducing the amount of transmitted data, saving wireless communication bandwidth resources, and improving the efficiency of concurrent transmission between multiple devices in industrial scenarios.

[0007] As a preferred embodiment of the present invention, when performing time-frequency domain decomposition on historical transmitted data packets, a short-time Fourier transform is performed to obtain the time-frequency spectrum of each data packet within a preset time window. Energy peak points are detected on this time-frequency spectrum, and the center frequency and occurrence time corresponding to each energy peak point are recorded. The center frequency and occurrence time are combined into a two-dimensional feature vector, and the two-dimensional feature vectors corresponding to all energy peak points within a data packet are sorted according to their occurrence time and concatenated to form the original feature value sequence. By extracting energy peak points as sparse features, the most representative signal components in the time-frequency spectrum can be retained, low-energy noise and redundant frequency components can be filtered out, and the sequence dimension can be further reduced. Furthermore, the length of the preset time window is dynamically adjusted according to the channel coherence time of the industrial communication network, so that the time-frequency analysis window matches the channel variation characteristics, ensuring a balance between the time resolution and frequency domain resolution of feature extraction, and enhancing the adaptability of the feature sequence to the time-varying characteristics of the channel.

[0008] When generating the offset sequence, the median of the original feature value sequence is preferably calculated as the reference feature value. The reference feature value is then subtracted from each feature value in the original feature value sequence to obtain the original offset corresponding to each feature value. The original offsets are then symbolically encoded: positive offsets are encoded as the first symbol, negative offsets as the second symbol, and zero offsets as the third symbol. The encoded symbols are combined with the absolute values ​​of the original offsets to generate the offset sequence. Using the median as the reference reduces the interference of extreme peaks on the reference value, making the offset distribution more concentrated, thereby improving the compactness of subsequent quantization encoding.

[0009] One implementation of this invention for obtaining the compressed feature index sequence is as follows: The maximum and minimum values ​​of all offsets in the offset sequence are statistically analyzed, and the numerical fluctuation range is determined using these two values ​​as boundaries. This numerical fluctuation range is then uniformly divided into multiple quantization levels, each corresponding to a unique index code. The offset sequence is traversed, the quantization level to which each offset belongs is determined, and the index code corresponding to that quantization level is extracted. These are then arranged in traversal order to obtain the compressed feature index sequence. Uniform quantization and index mapping simplify the offset sequence into a finite number of discrete symbols, effectively removing information redundancy caused by subtle numerical fluctuations. Preferably, the number of quantization levels is adaptively determined based on the ratio of the standard deviation of the offset sequence to a preset quantization step size. This allows for the automatic increase of quantization levels to ensure reconstruction accuracy when offset fluctuations are severe, and the reduction of levels to obtain a higher compression ratio when fluctuations are mild, achieving a dynamic balance between compression performance and reconstruction accuracy.

[0010] When generating a compressed data frame, the transmitting wireless communication device parses and retains the frame header information of historically transmitted data packets; it replaces the original payload data field entirely with a compressed feature index sequence, and adds a termination identifier to the end of the replaced payload data field, thus generating a compressed data frame; subsequently, it sends this compressed data frame to the receiving wireless communication device via the industrial wireless communication protocol stack. Retaining the frame header information ensures that the compressed data frame remains compatible with existing industrial wireless communication protocol formats without additional signaling overhead, while the termination identifier clearly defines the end position of the compressed payload, facilitating accurate extraction of the valid sequence by the receiving end. The termination identifier is preferably a specific bit sequence of a preset length, which will not naturally appear in the compressed feature index sequence, avoiding ambiguity with the index code and ensuring the reliability of frame parsing.

[0011] At the receiving wireless communication device, after receiving the compressed data frame, the payload data field is parsed to extract the compressed feature index sequence and the termination identifier. Based on each index code in the compressed feature index sequence, the corresponding quantization level center value is searched in a pre-constructed local codebook table to reconstruct the offset reconstruction sequence. Then, based on the offset reconstruction sequence and the pre-negotiated baseline feature value, the original feature value reconstruction sequence is recovered through inverse offset operations. Through codebook mapping and inverse operations, the feature sequence can be accurately recovered from the compressed index without relying on additional transmitted information, achieving transparent data compression and decompression without altering the upper-layer application's understanding of the data.

[0012] Furthermore, during offset reconstruction, the receiving end acquires the same numerical fluctuation range and quantization level division method as the sending end, and constructs a local codebook table recording the mapping relationship between each index code and a quantization level center value. Each index code in the compressed feature index sequence is read sequentially, and the local codebook table is searched using the index code as the key to obtain the corresponding quantization level center value. All the obtained quantization level center values ​​are arranged in order to form the offset reconstruction sequence. This search method based on a shared codebook has extremely low computational complexity, making it suitable for the limited processor resources of industrial equipment.

[0013] During the reverse offset operation, the receiving end obtains the baseline feature value agreed upon with the sending end, as well as the symbol code corresponding to each reconstructed offset in the offset reconstruction sequence. For each reconstructed offset, the operation direction is determined according to its symbol code: the first symbol corresponds to addition, the second symbol to subtraction, and the third symbol to adding zero. The baseline feature value and each reconstructed offset are calculated according to the operation direction to obtain the reconstructed feature value. All the reconstructed feature values ​​are arranged in order to obtain the original feature value reconstruction sequence. This process is entirely based on simple rules pre-agreed upon by both parties, without the need for online transmission of symbol information, further improving compression efficiency.

[0014] After recovering the original eigenvalue reconstruction sequence, an inverse time-frequency domain transform is performed on the original eigenvalue reconstruction sequence to convert each two-dimensional eigenvector in the sequence back to its corresponding energy peak point information. Based on all energy peak point information and combined with the timestamps in the frame header information of historical transmitted data packets, a reconstructed time-frequency spectrum is synthesized. An inverse short-time Fourier transform is performed on the reconstructed time-frequency spectrum to recover the original payload data, completing the data decompression process. This complete reconstruction path enables the receiving end to restore the original data packet that is highly consistent with the sending end, ensuring the accurate transmission of industrial control commands.

[0015] As an enhancement of this invention, after obtaining the compressed feature index sequence, the channel quality indicator value of the current industrial communication network is detected to determine whether it is lower than a preset robustness threshold. When the channel quality indicator value is lower than the threshold, forward error correction coding is performed on the compressed feature index sequence to generate an enhanced index sequence with redundant check information. This enhanced index sequence is then used to replace the original compressed feature index sequence to generate a compressed data frame. When the wireless channel deteriorates rapidly, the introduction of error correction coding can effectively combat transmission errors, ensuring the correct reception and reconstruction of the compressed feature index sequence. This improves the reliability of data transmission without increasing bandwidth usage, meeting the stringent requirements for communication robustness in industrial environments.

[0016] The technical effects and advantages provided by the present invention in the above technical solution are as follows: By performing time-frequency domain decomposition on historical transmitted data packets, the time-spectrum graph of each data packet is obtained, and the center frequency and occurrence time of the energy peak points are extracted to form the original feature value sequence. The energy distribution of industrial payload data in the time-frequency domain is highly sparse, requiring only a small number of energy peak points to characterize the core information of the data packet. Using this sparse feature sequence instead of the complete original payload data as the compression target significantly reduces the volume of information to be encoded at the source, resulting in a substantial reduction in the amount of data subsequently encoded and transmitted, and lowering the wireless channel occupancy time.

[0017] The statistical distribution center of the original feature value sequence is selected as the reference feature value. Each feature value in the sequence is converted into an offset relative to the reference, and a symbolic code is added to preserve directional information. The numerical dynamic range of the offset sequence is significantly narrowed compared to the original feature value sequence, reducing the numerical span that the quantization interval needs to cover. At the same quantization precision, the narrowed dynamic range means that fewer quantization levels can be used to complete the full description of the offset sequence, further reducing the length of the compressed index code.

[0018] The maximum and minimum value boundaries are determined based on the numerical fluctuation range of the offset sequence itself. Then, the number of quantization levels is adaptively determined based on the sequence statistical characteristics, constructing a dynamic quantization interval that matches the current data packet characteristic fluctuation state. When the data packet characteristic values ​​fluctuate drastically, the number of quantization levels automatically increases to maintain recovery accuracy; when the data packet characteristic values ​​tend to stabilize, the number of quantization levels automatically decreases to reduce index code redundancy. This adaptive mechanism allows the compression strategy to follow the changes in the inherent characteristics of the payload data in real time, dynamically adjusting between compression efficiency and recovery fidelity, avoiding the performance degradation problem of fixed quantization parameters when facing diverse industrial data.

[0019] At the receiving end, using the same reference feature values ​​and local codebook table as the sending end, the offset sequence is reconstructed by retrieving the quantization level center value through the index code. Then, the original feature value sequence is recovered by performing an inverse offset operation based on the symbol encoding. The entire reconstruction process only requires table lookups and simple arithmetic operations, resulting in extremely low computational overhead, meeting the stringent requirements of real-time decompression for industrial wireless communication equipment. The time-frequency sparse features obtained from the reconstruction can be losslessly recovered from the original payload data through inverse short-time Fourier transform, with a clear data restoration path and controllable computational load. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0021] Figure 1This is a flowchart of the data transmission method for wireless communication devices in industrial settings. Figure 2 This is a flowchart of the offset sequence generation process; Figure 3 It is a flowchart of adaptive forward error correction coding and data frame generation based on compressed feature index sequence of channel quality indication; Figure 4 This is a comparison chart of the probability density of reconstruction error amplitudes for different historical data packets. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] See Figure 1 This invention provides a data transmission method for wireless communication devices in an industrial setting, comprising: acquiring historical transmission data packets of multiple wireless communication devices within an industrial communication network; performing time-frequency domain decomposition on the historical transmission data packets; extracting multiple time-frequency domain sparse feature values ​​corresponding to each historical transmission data packet to form an original feature value sequence; selecting a reference feature value based on the statistical distribution characteristics of the original feature value sequence; calculating the offset of each feature value in the original feature value sequence from the reference feature value to generate an offset sequence; constructing an adaptive quantization interval based on the numerical fluctuation range of the offset sequence; mapping the offset sequence to an index code to obtain a compressed feature index sequence; replacing the original payload data in the historical transmission data packets with the compressed feature index sequence to generate a compressed data frame; and performing data transmission between wireless communication devices based on the compressed data frame.

[0024] Example 1:

[0025] In practical implementation, historical transmission data packets from multiple wireless communication devices within the industrial communication network are acquired. These historical transmission data packets are complete data packets that these wireless communication devices have sent or received during past communications. Each historical transmission data packet contains frame header information and raw payload data, which is the valid information bit stream to be transmitted.

[0026] A short-time Fourier transform is performed on each historical transmitted data packet. The window function used for performing the short-time Fourier transform is the Hamming window, and its expression is:

[0027] in, Indicates the Hamming window in the index Window coefficient at the location, Indicates the length of the preset time window. This is the index of the sampling point within the window. Pi is a constant.

[0028] Preset time window length The channel coherence time of the industrial communication network is dynamically adjusted. The adjustment process is as follows: The current channel coherence time of the industrial communication network is obtained through a channel estimation algorithm. The channel estimation algorithm uses pilot symbol-based least squares channel estimation. Using the known pilot symbols inserted at the transmitting end and the received signal at the corresponding pilot position at the receiving end, the channel frequency response at the pilot position is calculated. An inverse Fourier transform is performed on the channel frequency response to obtain the channel impulse response. The autocorrelation function of the channel impulse response is calculated, and the time delay corresponding to the decrease in the amplitude of the autocorrelation function to 0.7 times its maximum amplitude is taken as the channel coherence time. After obtaining the channel coherence time, it is used as the upper limit of the preset time window length, ensuring that the preset time window length is always less than or equal to the channel coherence time. Specifically, the preset time window length is... The value is the product of the channel coherence time and a preset coefficient. The preset coefficient is set to 0.9, based on the following: Monte Carlo simulation was used to statistically analyze the energy concentration of the time-frequency spectra obtained from historical transmitted data packets under different preset coefficient values. The energy concentration is defined as the ratio of the sum of the energy values ​​of the top 10% of time-frequency points with the highest energy values ​​in the time-frequency spectra to the sum of the energy values ​​of all time-frequency points. The energy concentration statistical results are optimal when the preset coefficient is 0.9, hence 0.9 is chosen. The sliding step size between adjacent windows in the short-time Fourier transform is set to the preset time window length. Half of it, so that there is an overlap between adjacent analysis windows.

[0029] After the short-time Fourier transform is completed, the time-spectrum diagram of the historical transmitted data packets within a preset time window is obtained. The time-spectrum diagram is a two-dimensional real matrix, with the rows of the matrix corresponding to the frequency axis and the columns corresponding to the time axis. The value of each element in the matrix is ​​the energy amplitude value at the corresponding time-frequency point.

[0030] Energy peak points are detected on the time-frequency spectrum. The detection method combines local maximum search with an adaptive threshold. The adaptive threshold is set as the average of all energy amplitude values ​​in the time-frequency spectrum plus twice the standard deviation of the energy amplitude values. Each time-frequency point in the time-frequency spectrum is traversed. If the energy amplitude value of the traversed time-frequency point is greater than the adaptive threshold and also greater than the energy amplitude values ​​of all other time-frequency points within a neighborhood window centered on the traversed time-frequency point, then the traversed time-frequency point is determined to be an energy peak point. The neighborhood window covers 3 time units in the time dimension and 5 frequency units in the frequency dimension. The time unit is one time sampling interval, and the frequency unit is one frequency sampling interval. The size of the neighborhood window is set based on the fact that the diffusion range of spectral energy in the frequency direction is usually greater than that in the time direction in industrial scenarios. For each determined energy peak point, the corresponding center frequency and occurrence time are recorded. The center frequency is the frequency value corresponding to the energy peak point on the frequency axis, in Hertz; the occurrence time is the time value corresponding to the energy peak point on the time axis, in seconds.

[0031] The center frequency and occurrence time of each recorded energy peak point are combined into a two-dimensional feature vector, which has the following form: ,in Indicates the first The center frequency of each energy peak point Indicates the first The occurrence times of each energy peak point. For a two-dimensional feature vector corresponding to all energy peak points within a historical transmission data packet, sorted by occurrence time... The values ​​are sorted in ascending order, and all the sorted two-dimensional feature vectors are concatenated end to end in order to form the original feature value sequence.

[0032] The above processing is performed on all historical transmission data packets of multiple wireless communication devices in the industrial communication network, and the original feature value sequence corresponding to each historical transmission data packet is extracted one by one.

[0033] Example 2:

[0034] In specific implementation, please refer to Figure 2The benchmark eigenvalue is selected based on the statistical distribution characteristics of the original eigenvalue sequence. The center frequency values ​​and occurrence time values ​​contained in all two-dimensional eigenvectors of the original eigenvalue sequence are uniformly sorted. The original eigenvalue sequence is composed of multiple two-dimensional eigenvectors concatenated in the order of their occurrence time. Each two-dimensional eigenvector contains a center frequency value and an occurrence time value. All center frequency values ​​and occurrence time values ​​of all two-dimensional eigenvectors in the original eigenvalue sequence are extracted to form a one-dimensional numerical set. All values ​​in the one-dimensional numerical set are sorted in ascending order to obtain an ordered numerical sequence. The total number of values ​​in the ordered numerical sequence is counted. When the total number is odd, the value in the middle position of the ordered numerical sequence is taken as the median of the original eigenvalue sequence; when the total number is even, the average of the values ​​in the two middle positions of the ordered numerical sequence is taken as the median of the original eigenvalue sequence. The calculated median is determined as the benchmark eigenvalue.

[0035] Calculate the offset using each eigenvalue in the original eigenvalue sequence and the reference eigenvalue. Iterate through each two-dimensional eigenvector in the original eigenvalue sequence, performing the same offset calculation operation on each. For each iterated two-dimensional eigenvector, subtract the center frequency value and the occurrence time value from the reference eigenvalue to obtain the original offset corresponding to the center frequency value and the original offset corresponding to the occurrence time value, respectively. The formula for calculating the original offset corresponding to the center frequency value is:

[0036] in, Indicates the first The original offset corresponding to the center frequency value in each two-dimensional feature vector. Indicates the first The center frequency value in a two-dimensional feature vector This represents the baseline characteristic value. The calculation of the original offset corresponding to the value at the occurrence time uses the same method, substituting the value at the occurrence time into... The position is calculated. After traversing all two-dimensional feature vectors in the original feature value sequence, the original offsets corresponding to the two values ​​in each two-dimensional feature vector are obtained, and all the original offsets are arranged in the order of the original feature value sequence.

[0037] Each original offset is symbolically encoded. The symbol encoding rule is as follows: positive offsets are encoded as the first symbol, negative offsets as the second symbol, and zero offsets as the third symbol. The first symbol is represented by the binary bit sequence "00", the second symbol by the binary bit sequence "01", and the third symbol by the binary bit sequence "10". The allocation of the above symbol encoding bit sequences is based on the following: using fixed-length binary encoding facilitates parsing by the receiving end, and reserving the binary bit sequence "11" as a reserved encoding for specific extended purposes. When symbolically encoding the original offset, the relationship between the value of the original offset and zero is determined. If the value of the original offset is greater than zero, the original offset is determined to be a positive offset, and the first symbol "00" corresponding to the positive offset is assigned to the original offset; if the value of the original offset is less than zero, the original offset is determined to be a negative offset, and the second symbol "01" corresponding to the negative offset is assigned to the original offset; if the value of the original offset is equal to zero, the original offset is determined to be a zero offset, and the third symbol "10" corresponding to the zero offset is assigned to the original offset.

[0038] The encoded symbol is combined with the absolute value of the original offset to generate a single offset element in the offset sequence. For each original offset, the absolute value of the original offset is calculated, and the symbol encoding assigned to the original offset is concatenated with the calculated absolute value. The concatenated offset element contains a symbol encoding field and an amplitude value field, with the symbol encoding field preceding the amplitude value field. The symbol encoding field stores the binary bit sequence of the first, second, or third symbol, and the amplitude value field stores the binary representation of the absolute value of the original offset. All offset elements corresponding to all two-dimensional feature vectors in the original feature value sequence are arranged sequentially according to the original order of the feature value sequence to generate the offset sequence. The order of elements in the offset sequence is consistent with the order of the two-dimensional feature vectors in the original feature value sequence.

[0039] Example 3:

[0040] In practice, an adaptive quantization interval is constructed based on the numerical fluctuation range of the offset sequence. The offset sequence consists of multiple offset elements arranged sequentially. Each offset element contains a symbol encoding field and an amplitude value field, where the amplitude value field stores the binary representation of the absolute value of the original offset. Each offset element in the offset sequence is traversed, and the value stored in the amplitude value field of each offset element is extracted. All extracted amplitude values ​​are then combined into an amplitude value set. All values ​​in the amplitude value set are compared to determine the maximum and minimum values. The comparison method is as follows: An initial maximum value variable and an initial minimum value variable are defined. The initial maximum value variable is assigned the first amplitude value in the amplitude value set, and the initial minimum value variable is also assigned the first amplitude value in the amplitude value set. Each remaining amplitude value in the amplitude value set is read sequentially. If the read amplitude value is greater than the current maximum value variable, the maximum value variable is updated with the read amplitude value; if the read amplitude value is less than the current minimum value variable, the minimum value variable is updated with the read amplitude value. After traversal, the value stored in the maximum value variable is the maximum value of the amplitude value set, and the value stored in the minimum value variable is the minimum value of the amplitude value set. Using the defined maximum and minimum values ​​as boundaries, the numerical fluctuation range of the offset sequence is determined. The interval of the numerical fluctuation range is represented as... ,in This represents the minimum value in the set of amplitude values. This represents the maximum value in the set of amplitude values.

[0041] The numerical fluctuation range is evenly divided into multiple quantization levels, each corresponding to a unique index code. The number of quantization levels is adaptively determined based on the ratio of the standard deviation of the offset sequence to the preset quantization step size, calculated using the following formula:

[0042] in, Indicates the number of quantization levels. This represents the floor operator, which rounds the result up to the nearest integer in the direction of positive infinity. This represents the standard deviation of all amplitude values ​​in the offset sequence. This indicates the preset quantization step size.

[0043] Standard deviation The calculation process is as follows: Calculate the average of all amplitude values ​​in the amplitude value set; calculate the square of the difference between each amplitude value and the average; sum the squares of all differences and divide by the total number of amplitude values ​​in the amplitude value set; calculate the arithmetic square root of the quotient; the result of the arithmetic square root is the standard deviation. .

[0044] Preset quantization step size Set to a fixed value of 8. The basis for this fixed value is as follows: considering that the typical resolution of analog-to-digital converters in industrial wireless communication equipment is 8 to 12 bits, the minimum quantization interval corresponding to 8-bit resolution is selected as a reference. Setting the preset quantization step size to 8 can achieve a balance between quantization accuracy and compression efficiency. When the preset quantization step size is less than 8, the number of quantization levels increases, which leads to an increase in the index code encoding length. When the preset quantization step size is greater than 8, the number of quantization levels decreases, which leads to a decrease in reconstruction accuracy.

[0045] The process of uniformly dividing the quantization levels is as follows: dividing the range of numerical fluctuations into intervals... Average score A series of continuous and non-overlapping sub-intervals, each representing a quantization level, with a width of [missing information]. The calculation method is as follows . No. The sub-interval range corresponding to each quantization level is: ,in This is the sequence number of the quantification level. The value is The last quantification level is the first The sub-interval range corresponding to each quantization level is: .

[0046] Each quantization level is assigned a unique index code. The index code is represented using binary encoding, with the number of bits depending on the total number of quantization levels. Determine the number of bits in the binary encoding. The calculation method is as follows From binary encoding " "Starting as the first" Each quantization level is assigned an index code, and the binary code value is incremented sequentially according to the ascending order of the quantization level number, so that each quantization level is assigned a unique index code.

[0047] Iterate through all offset elements in the offset sequence, determine the quantization level of each offset element, and extract the corresponding index code. The traversal process is as follows: read each offset element sequentially according to the order of the offset elements in the offset sequence, and extract the amplitude value stored in the amplitude value field of the read offset element. Compare the amplitude value with the sub-interval range corresponding to each quantization level, starting from the first... Starting with each quantization level, the amplitude value is sequentially checked to see if it belongs to the sub-interval range of the current quantization level. If the amplitude value is greater than or equal to the lower limit of the sub-interval range and less than the upper limit, or if the amplitude value of the last quantization level is greater than or equal to the lower limit of the sub-interval range and less than or equal to the upper limit, then the read offset element is determined to belong to the current quantization level. Based on the determination result, the index code corresponding to the quantization level to which the read offset element belongs is extracted.

[0048] The extracted index codes are arranged sequentially according to the traversal order of the offset elements in the offset sequence to obtain the compressed feature index sequence. The compressed feature index sequence is a sequence composed of consecutively arranged index codes, and the order of the index codes in the compressed feature index sequence corresponds one-to-one with the order of the offset elements in the offset sequence.

[0049] Example 4:

[0050] In practice, at the transmitting wireless communication device, the frame header information of historically transmitted data packets is parsed. The parsing process involves reading the bitstream of historically transmitted data packets and locating the frame header information according to the frame format specified by the industrial wireless communication protocol stack. The frame header information includes fields such as preamble, synchronization word, source address, destination address, frame type identifier, and timestamp. All fields of the frame header information are completely extracted and saved to the buffer of the transmitting wireless communication device, while all bits of the frame header information remain unchanged.

[0051] The original payload data field of the historical transmitted data packets is entirely replaced with a compressed feature index sequence. The original payload data field is the valid data payload portion of the historical transmitted data packets, located after the frame header information and before the frame check sequence. The start and end positions of the original payload data field are determined from the bitstream of the historical transmitted data packets. All bits between the start and end positions are deleted, and the compressed feature index sequence is inserted into the original positions. The compressed feature index sequence is a sequence of consecutively arranged index codes, each index code being a fixed-length binary code. All index codes are concatenated sequentially to obtain the bitstream used for replacement.

[0052] A terminating identifier is added to the end of the replaced payload data field. The terminating identifier is placed immediately after the last bit of the last index code in the compressed feature index sequence. The terminating identifier is a specific bit sequence of a predetermined length that will not naturally appear in the compressed feature index sequence. The specific bit sequence is selected by counting all possible binary encoding values ​​of all index codes in the compressed feature index sequence, and reserving one binary encoding value that was not used for an index code as the specific bit sequence. For example, the number of binary encoding bits of the index code. Based on the number of quantization levels used when generating the compressed feature index sequence It is confirmed that the calculation method is as follows: The range of values ​​for the index code is: arrive The binary representation of an integer, converting binary values... corresponding A specific bit sequence of binary code serves as the termination identifier. The transmitting wireless communication device concatenates the frame header information, the compressed feature index sequence bit stream, and the termination identifier in sequence to generate a compressed data frame.

[0053] The compressed data frame is sent to the receiving wireless communication device via an industrial wireless communication protocol stack. The transmission process is as follows: the compressed data frame is passed to the medium access control layer of the industrial wireless communication protocol stack. The medium access control layer adds a necessary frame check sequence to the compressed data frame. The frame check sequence is generated using a cyclic redundancy check (CRC) algorithm, and the check range covers all bits of the frame header information, compressed feature index sequence, and stop identifier. Then, the compressed data frame with the frame check sequence is submitted to the physical layer. The physical layer performs channel coding and modulation, and transmits it as electromagnetic waves through the industrial wireless communication antenna. The industrial wireless communication protocol stack uses a protocol stack based on the IEEE 802.15.4 standard or the WirelessHART protocol stack.

[0054] At the receiving wireless communication device, after receiving a compressed data frame, the compressed data frame is parsed. The receiving physical layer demodulates and decodes the received electromagnetic wave signal to recover the compressed data frame bitstream with a frame check sequence, and delivers the bitstream to the media access control layer. The media access control layer uses the frame check sequence for cyclic redundancy check. After the check passes, the frame check sequence is removed, and the compressed data frame bitstream without the frame check sequence is delivered to the upper layer for processing. In the upper layer processing, the receiving wireless communication device locates the payload data field from the compressed data frame. The payload data field contains a concatenation of a compressed feature index sequence and a stop identifier. Starting from the beginning of the payload data field, the index code is read, and each... Each bit is parsed into an index code, and parsing continues until a specific bit sequence that perfectly matches the stop identifier is encountered. This process continues up to the last bit sequence. All index codes preceding the matching position are extracted to obtain the compressed feature index sequence.

[0055] Based on each index code in the compressed feature index sequence, the corresponding quantization level center value is searched in the pre-constructed local codebook table to reconstruct the offset reconstruction sequence. The construction of the local codebook table is performed at the receiving wireless communication device before receiving the compressed data frame. The receiving and transmitting wireless communication devices negotiate to obtain the same numerical fluctuation range and quantization level division method. The negotiation process is as follows: after generating the compressed feature index sequence, the transmitting wireless communication device sets the minimum value of the numerical fluctuation range... Maximum value and the number of quantification levels As parameter information, this information is sent to the receiving wireless communication device via independent control signaling frames. The receiving wireless communication device receives and stores this parameter information. Based on the stored parameter information, the receiving wireless communication device determines the range of numerical fluctuations. Evenly divided into There are several quantization levels, and the center value of each quantization level is calculated. The center value of each quantization level The calculation formula is:

[0056] in, Indicates the first The center value of each quantization level This is the sequence number of the quantification level. The range of values ​​is , This represents the minimum value within the range of numerical fluctuation. This represents the maximum value within the range of numerical fluctuation. This indicates the number of quantization levels. A mapping is established between each index code and its corresponding quantization level center value; index code "0" corresponds to the [number of quantization levels]. The quantization level, with index code "1" corresponding to the quantization level. Each quantization level, and so on, records all mapping relationships in a table, which is the local codebook table.

[0057] The receiving wireless communication device sequentially reads each index code in the compressed feature index sequence, searches the local codebook table using the index code as the key, and finds the quantization level center value corresponding to the index code in the local codebook table. All quantization level center values ​​obtained in the order of index code reading are arranged sequentially to form an offset reconstruction sequence. Each reconstruction offset in the offset reconstruction sequence is a quantization level center value.

[0058] Based on the offset reconstruction sequence and the pre-negotiated baseline feature values, the original feature value reconstruction sequence is recovered through inverse offset operations. The pre-negotiated baseline feature value is the median of the original feature value sequence. After generating the baseline feature value, the transmitting wireless communication device sends it to the receiving wireless communication device via control signaling frames, and the receiving wireless communication device stores the baseline feature value. Each reconstructed offset in the offset reconstruction sequence also carries symbol encoding information. The symbol encoding information is determined when the transmitting end generates the offset sequence. The transmitting wireless communication device needs to transmit the symbol encoding information corresponding to each offset element along with the compressed feature index sequence to the receiving wireless communication device. The transmission method is as follows: In the payload data field of the compressed data frame, a symbol encoding mapping field is appended immediately after the stop identifier. The symbol encoding mapping field records the symbol encoding contained in the offset element corresponding to each index code in the compressed feature index sequence, arranged in the order of the index codes. After parsing the compressed feature index sequence and the stop identifier, the receiving end continues to read the symbol encoding mapping field after the stop identifier to parse the symbol encoding corresponding to each index code.

[0059] For each reconstructed offset in the offset reconstruction sequence, obtain its corresponding symbol code. Determine the operation direction based on the symbol code: the first symbol corresponds to the addition direction, the second symbol to the subtraction direction, and the third symbol to the addition of zero direction. Calculate the base feature value with each reconstructed offset according to the determined operation direction. If it's an addition direction, the reconstructed feature value equals the base feature value plus the reconstructed offset; if it's a subtraction direction, the reconstructed feature value equals the base feature value minus the reconstructed offset; if it's the addition of zero direction, the reconstructed feature value equals the base feature value plus zero. Arrange all the reconstructed feature values ​​obtained sequentially to obtain the original feature value reconstruction sequence. The order of the reconstructed feature values ​​in the original feature value reconstruction sequence is consistent with the order of the feature values ​​in the original feature value sequence.

[0060] After reconstructing the original eigenvalue sequence, an inverse time-frequency domain transform is performed on it. This transforms each two-dimensional eigenvector in the original eigenvalue sequence back to its corresponding energy peak point information. The original eigenvalue sequence is composed of two-dimensional eigenvectors arranged sequentially, each represented as... ,in To reconstruct the center frequency, This is the moment when the reconstruction occurs. As the first The center frequency of each energy peak point will As the first The occurrence times of each energy peak point are used to reconstruct the information of each energy peak point. Based on all the energy peak point information and the timestamps in the frame header information of historical transmitted data packets, a reconstructed time-frequency spectrum is synthesized. The synthesis method is as follows: a two-dimensional real number matrix of the same size as the time-frequency spectrum generated by the sending end is created, with the rows of the matrix corresponding to the frequency axis and the columns corresponding to the time axis. The initial value of all elements in the matrix is ​​set to zero. The timestamps in the frame header information provide the time reference for the time-frequency spectrum. Using the timestamps as the zero point of the time axis, the row and column indices in the two-dimensional real number matrix are determined according to the occurrence time and center frequency of each energy peak point. The value of the corresponding matrix element is set to the amplitude value of the energy peak point. If there is no amplitude value, it is set to 1 to indicate that there is an energy peak at that time-frequency point. After setting the values ​​for all energy peak points, the reconstructed time-frequency spectrum is obtained.

[0061] An inverse short-time Fourier transform (ISFT) is performed on the reconstructed time-frequency spectrogram to recover the original payload data. The IFT uses the exact same window function, window length, and overlap length parameters as the IFT performed at the transmitter. The window function also uses a Hamming window, and the window length... The same preset time window length as the transmitter is used, and the overlap length uses the same half-window length sliding step as the transmitter. The inverse short-time Fourier transform overlap and superposition synthesis method adopts the weighted superposition averaging method, which performs equal weight averaging on the reconstructed signals of multiple windows at each time domain sampling point, removes the amplitude modulation introduced by window overlap, and finally recovers the complete original payload data time domain signal, thus completing the data decompression process.

[0062] See Figure 4 In the graph, the horizontal axis represents the reconstruction error magnitude in Hertz, ranging from approximately -15Hz to 15Hz; the vertical axis represents the probability density. In the legend, the solid line with dots represents data packet A, and the dashed line with squares represents data packet B. The curves illustrate the probability distribution characteristics of the reconstruction error magnitudes corresponding to the two data packets.

[0063] As can be seen from the figure, the reconstruction error of data packet A is concentrated in the range of -5Hz to 5Hz, and reaches a peak near 0Hz with a peak probability density of about 0.21, indicating that most reconstruction errors are small and the reconstruction accuracy is high. The probability density of data packet A drops rapidly to near zero outside of ±5Hz, indicating that cases with large reconstruction errors are extremely rare.

[0064] The reconstruction error amplitude distribution of data packet B is relatively wide, mainly concentrated in the range of -10Hz to 10Hz. Furthermore, the peak probability density near 0Hz of the reconstruction error is approximately 0.10, lower than that of data packet A, indicating that the reconstruction error of data packet B is more dispersed. The probability density of data packet B still exhibits a certain amplitude within the ±10Hz range, suggesting the existence of more reconstructions with larger error amplitudes.

[0065] Overall, both sets of data packets exhibit a symmetrical distribution of reconstruction errors centered at zero. However, the error distribution of data packet A is more concentrated, with a smaller error amplitude, consistent with the technical effect in Example 4 where steps such as frame header information parsing, replacing the original payload data with compressed feature index sequences, and adding termination identifiers effectively ensure the accuracy of data packet reconstruction. The error distribution of data packet B is more dispersed, possibly reflecting differences in the time-frequency domain feature extraction and quantization encoding processes of different historical transmitted data packets. This error distribution trend verifies the applicability and effectiveness of the reconstruction method based on compressed feature index sequences and local codebook tables in Example 4 for different data packets.

[0066] Example 5:

[0067] In practical implementation, after obtaining the compressed feature index sequence, the transmitting wireless communication device detects the channel quality indicator (CQI) of the current industrial communication network. The physical layer of the transmitting wireless communication device measures the received wireless signal in real time, obtaining the received signal strength indicator (RSS) value, measured in decibels (dB) and milliwatts (mW). The physical layer reports the RSS value to the media access control (MAC) layer, which uses the RSS value as the CQI. The CQI value is represented by the symbol... express.

[0068] The transmitting wireless communication device will display the channel quality indicator value. Compare with a preset robustness threshold. The preset robustness threshold is indicated by a symbol. express, The setting is -85 dBmW. The basis for setting -85 dBmW is that the typical receiving sensitivity of receivers used in industrial wireless communication under the lowest modulation and coding scheme is -95 dBmW to -90 dBmW. Adding a protection margin of 5 dBmW to 10 dBmW to the typical receiving sensitivity value results in -85 dBmW. Using -85 dBmW can trigger the enhanced protection mechanism in advance when shallow fading occurs in the channel, avoiding frequent triggering that leads to a decrease in transmission efficiency.

[0069] When the channel quality indicator value Below the preset robustness threshold At this time, the transmitting wireless communication device performs forward error correction coding on the compressed feature index sequence. The forward error correction coding uses systematic convolutional codes. The encoder of the systematic convolutional code consists of a set of shift registers and a modulo-2 adder. The number of shift registers is 6, the constraint length is 7, and the generator polynomial is represented in octal as (171, 133), corresponding to the first generator sequence in binary. Second generated sequence The encoder input is a bit stream of compressed feature index sequence. For each information bit input, the encoder outputs two parity bits. The information bits and parity bits together constitute the encoded output.

[0070] The coding rate of forward error correction coding is based on the channel quality indicator value. The fading range is adaptively selected. The fading range is divided using a preset robustness threshold. and receiver sensitivity level Using these as boundaries, the range of channel quality indicator values ​​is divided into three intervals. Receiver sensitivity level. Set to -95 dB / mW, consistent with the minimum receivable signal power for industrial wireless communication equipment. The three intervals are: Interval 1, Channel Quality Indicator Value. Greater than or equal to Second interval, channel quality indicator value Less than and greater than or equal to The third interval, channel quality indicator value. Less than and greater than or equal to ,in Set to -90 dBmW, for and The median value. When When in the second interval, the selected coding rate It is 1 / 2; when When located in the third interval, the selected coding rate It is 1 / 3; when Below At that time, the selected coding rate It is 1 / 4.

[0071] Based on the selected coding rate and the original number of bits in the compressed feature index sequence Calculate the number of redundant parity bits that need to be added. The calculation formula is:

[0072] in, Indicates the number of redundancy check bits. This represents the floor function. This represents the total number of bits in the compressed feature index sequence. The value is the sum of the number of bits of all index codes in the compressed feature index sequence. Indicates the selected forward error correction coding rate. The value can be 1 / 2, 1 / 3 or 1 / 4.

[0073] When performing systematic convolutional code encoding on the compressed feature index sequence, the compressed feature index sequence's... All information bits are input into the encoder, and all check bits output by the encoder are collected. The number of bits retained according to the punching pattern is... The parity bits are then appended to the information bits of the compressed feature index sequence to generate an enhanced index sequence. The enhanced index sequence is constructed by sequentially concatenating the original information bits and the appended redundant parity bits.

[0074] An enhanced index sequence is used to replace the compressed feature index sequence to generate compressed data frames. The replacement process is as follows: In the processing flow of the transmitting wireless communication device generating compressed data frames, the bit stream of the enhanced index sequence, which was originally intended to be placed in the payload data field, is replaced with the bit stream of the enhanced index sequence. The enhanced index sequence bit stream is then filled into the original payload data field positions of the historical transmitted data packets, and a termination identifier is added to the end of the enhanced index sequence bit stream. The termination identifier is generated in the same way as when forward error correction coding is not performed. Finally, frame header information is added to form the compressed data frame. The compressed data frame is sent to the receiving wireless communication device via the industrial wireless communication protocol stack.

[0075] When the channel quality indicator value Not lower than the preset robustness threshold In this case, forward error correction coding is not performed; the compressed feature index sequence is directly used to generate the compressed data frame. At this time, the payload data field of the compressed data frame consists directly of the compressed feature index sequence bitstream and a termination identifier, and the processing flow is completely consistent with that when enhanced protection is not triggered.

[0076] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A data transmission method for wireless communication devices in an industrial setting, characterized in that, include: Acquire historical transmission data packets from multiple wireless communication devices within an industrial communication network, perform time-frequency domain decomposition on the historical transmission data packets, extract multiple time-frequency domain sparse feature values ​​corresponding to each historical transmission data packet, and form an original feature value sequence. Based on the statistical distribution characteristics of the original feature value sequence, a benchmark feature value is selected, and the offset of each feature value in the original feature value sequence from the benchmark feature value is calculated to generate an offset sequence. Based on the numerical fluctuation range of the offset sequence, an adaptive quantization interval is constructed, and the offset sequence is mapped to an index code to obtain a compressed feature index sequence. The original payload data in the historical transmission data packet is replaced with the compressed feature index sequence to generate a compressed data frame, and data transmission between wireless communication devices is performed based on the compressed data frame.

2. The data transmission method for a wireless communication device in an industrial setting according to claim 1, characterized in that, The steps of acquiring historical transmission data packets from multiple wireless communication devices within an industrial communication network, performing time-frequency domain decomposition on the historical transmission data packets, and extracting multiple time-frequency domain sparse feature values ​​corresponding to each historical transmission data packet to form an original feature value sequence include: Perform a short-time Fourier transform on the historical transmitted data packets to obtain the time-spectrum diagram of each data packet within a preset time window; Energy peak points are detected on the time-spectrum graph, and the center frequency and occurrence time corresponding to each energy peak point are recorded. The center frequency and occurrence time are combined into a two-dimensional feature vector, and the two-dimensional feature vectors corresponding to all energy peak points in a data packet are sorted according to their occurrence time and concatenated to form the original feature value sequence.

3. The data transmission method for a wireless communication device in an industrial setting according to claim 2, characterized in that, The length of the preset time window is dynamically adjusted according to the channel coherence time of the industrial communication network.

4. The data transmission method for a wireless communication device in an industrial setting according to claim 1, characterized in that, The step of selecting a benchmark feature value based on the statistical distribution characteristics of the original feature value sequence, and calculating the offset of each feature value in the original feature value sequence relative to the benchmark feature value to generate an offset sequence includes: The median of the original feature value sequence is calculated as the baseline feature value; Subtract the baseline feature value from each feature value in the original feature value sequence to obtain the original offset corresponding to each feature value; The original offsets are symbolically encoded: positive offsets are encoded as the first symbol, negative offsets as the second symbol, and zero offsets as the third symbol. The encoded symbols are then combined with the absolute values ​​of the original offsets to generate the offset sequence.

5. The data transmission method for a wireless communication device in an industrial setting according to claim 1, characterized in that, The steps of constructing an adaptive quantization interval based on the numerical fluctuation range of the offset sequence, mapping the offset sequence to an index code, and obtaining a compressed feature index sequence include: The maximum and minimum values ​​of all offsets in the offset sequence are statistically analyzed, and the range of numerical fluctuation is determined using the maximum and minimum values ​​as boundaries. The numerical fluctuation range is evenly divided into multiple quantization levels, and each quantization level corresponds to a unique index code. Traverse the offset sequence, determine the quantization level to which each offset belongs, extract the index code corresponding to the quantization level, arrange all the extracted index codes in the traversal order, and obtain the compressed feature index sequence.

6. The data transmission method for a wireless communication device in an industrial setting according to claim 5, characterized in that, The number of quantization levels is adaptively determined based on the ratio of the standard deviation of the offset sequence to the preset quantization step size.

7. The data transmission method for a wireless communication device in an industrial setting according to claim 1, characterized in that, The steps of replacing the original payload data in the historical transmitted data packets with the compressed feature index sequence to generate compressed data frames, and performing data transmission between wireless communication devices based on the compressed data frames, include: In the wireless communication device at the transmitting end, the frame header information of the historical transmitted data packets is parsed, and the frame header information is retained unchanged; The original payload data field of the historical transmitted data packet is replaced entirely with the compressed feature index sequence, and a termination identifier is added to the end of the replaced payload data field to generate the compressed data frame. The compressed data frame is sent to the receiving wireless communication device via an industrial wireless communication protocol stack.

8. The data transmission method for a wireless communication device in an industrial setting according to claim 7, characterized in that, The termination identifier is a specific bit sequence of a preset length, which will not appear naturally in the compressed feature index sequence, and is used to identify the end position of the payload data.

9. The data transmission method for a wireless communication device in an industrial setting according to claim 7, characterized in that, After the step of sending the compressed data frame to the receiving wireless communication device via the industrial wireless communication protocol stack, the method further includes: At the receiving wireless communication device, after receiving the compressed data frame, the payload data field of the compressed data frame is parsed, and the compressed feature index sequence and the termination identifier are extracted. Based on each index code in the compressed feature index sequence, the corresponding quantization level center value is found in the pre-constructed local codebook table to reconstruct the offset reconstruction sequence. Based on the offset reconstruction sequence and the pre-negotiated baseline feature values, the original feature value reconstruction sequence is recovered through inverse offset operation.

10. The data transmission method for a wireless communication device in an industrial setting according to claim 9, characterized in that, The step of reconstructing the offset reconstruction sequence by finding the corresponding quantization level center value in the pre-constructed local codebook table based on each index code in the compressed feature index sequence includes: Obtain the same numerical fluctuation range and quantization level division method as the sending end, construct the local codebook table, and the local codebook table records the mapping relationship between each index code and a quantization level center value; Each index code in the compressed feature index sequence is read sequentially, and the local codebook table is retrieved using the index code as the key to obtain the corresponding quantization level center value; All the quantization level center values ​​obtained in sequence are arranged to form the offset reconstruction sequence.

Images