A method and apparatus for gaseous detection for human vital signs search

Abstract

The application discloses a gas detection method and instrument for human vital sign search, and particularly relates to the technical field of gaseous biological substance detection, and comprises a first short-time suction section resistance sampling sequence and a channel calibration table output by an intelligent sensing system at the same air inlet probe position, the resistance sampling sequence is from a solid sensitive layer whose resistance changes after absorbing or reacting with a fluid, and the channel calibration table records resistance channel numbers and gas source groups; the difference between the latter value and the former value of adjacent sampling positions of the same resistance channel is calculated, and a first suction response word is generated according to the difference value symbol; after the first short-time suction section ends, the air pump is stopped and the air inlet probe position is kept, and a pump-stopped waiting section resistance sampling sequence is acquired; through the execution of the first short-time suction, the pump-stopped waiting and the second short-time suction at the same air inlet probe position, the release sequence of residual gas is recorded, and Vitbi alignment and particle filtering are combined to distinguish gas path residual continuation and exhaled gas re-supply.

Description

Technical Field

[0001] This invention relates to the field of gaseous biological substance detection technology, and more specifically, to a gas detection method and instrument for searching for human vital signs. Background Technology

[0002] The current search and rescue of people buried in rubble mainly relies on audio and video detection, infrared thermal imaging, radar detection, search dogs, and a small number of life detection devices based on gas sensing. Gas detection equipment typically inserts an air inlet probe into rubble gaps, collapsed cavities, or debris flow accumulation layers. A local gas is extracted by an air pump, and a sensor array is used to detect CO2, NH3, as well as trace gases related to human exhalation or metabolism, such as acetone, ethylbenzene, acetaldehyde, pentanal, and heptal, and outputs alarm prompts accordingly. In actual rescue operations, trapped individuals are often surrounded by small, poorly ventilated cavities with limited access to the outside world. Trace amounts of gas produced by human exhalation slowly accumulate within these cavities. When the air intake probe continuously draws air in, the gas already accumulated near the probe is carried away first, while new gas from the depths of the cavity cannot immediately replenish the probe inlet. Consequently, the sensor array's subsequent response is prone to weakened signals for CO2, NH3, acetone, acetaldehyde, etc., inconsistent sampling before and after, and failure to maintain suspected alarms. This phenomenon does not directly indicate the absence of living individuals. Rather, it may be caused by the aspiration action altering the original gas distribution, leading to a single concentration assessment or continuous aspiration assessment misinterpreting the result of diluting the limited exhaled air mass as the absence of vital signs. The technical problem this application aims to solve is: how to determine the presence of a living human being based on the characteristics of gas re-injection, even when trace amounts of exhaled gas in a buried space are carried away by the suction action and replenishment is delayed. Summary of the Invention

[0003] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide a gas detection method and instrument for searching human vital signs. By performing a first short-term suction, a pump stop waiting period, and a second short-term suction at the same air inlet probe position, the release sequence of residual gas is recorded. Viterbi alignment and particle filtering are combined to distinguish between the continuation of residual gas path and the re-injection of exhaled gas, thereby solving the problems mentioned in the background art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: a gas detection method for searching human vital signs, comprising: S1. Obtain the first short-time suction segment resistance sampling sequence and channel calibration table output by the intelligent sensing system at the same air intake probe position. The resistance sampling sequence comes from the absorbing fluid or the solid sensitive layer whose resistance changes after reacting with the fluid. The channel calibration table records the resistance channel number and the gas source group. Subtract the previous value from the subsequent value of adjacent sampling positions of the same resistance channel and generate the first suction response word according to the difference sign. S2. After the first short-term suction phase ends, stop the air pump and maintain the position of the air inlet probe. Obtain the resistance sampling sequence of the pump stop waiting phase. Generate a pump stop difference symbol string by subtracting the previous value from the subsequent value. Write the first sampling bit that converts the positive sign to a negative sign into the residual release sequence table and output the pump stop recovery record. S3. Obtain the resistance sampling sequence of the second short-time suction segment according to the same air pump speed and suction duration of the first short-time suction segment, generate the second difference symbol string, take the channel order of the residual release order table as the residual state chain, take the positive sign sampling bit order of the second difference symbol string as the supplementary state chain, and output the second suction source mark sequence through the Viterbi alignment algorithm. S4. Generate particle status words in the edge computing node according to the second suction source marker sequence, predict the source segment according to the air pump status, update the particle weight according to the number of times the source segment is consistent with the second suction source marker sequence, and output the exhaled gas replenishment estimation record through the particle filtering algorithm. S5. Read the exhaled gas replenishment estimation record, and write the general exhaled gas replenishment position and the human trace replenishment position according to the gas source group and positive sign within the same source segment. After both types of replenishment positions are written, generate the live human existence result and output it to the display screen and buzzer.

[0005] In a preferred embodiment, S1 includes: S11. Obtain the channel registration record, sensitive layer reaction record, and exhalation source record within the intelligent sensing system; for any resistance channel number, read the solid sensitive layer number from the channel registration record, then read the reaction gas number from the sensitive layer reaction record using the solid sensitive layer number, and subsequently read the gas source group from the exhalation source record using the reaction gas number. Write the resistance channel number and gas source group as channel calibration items to generate a channel calibration table. S12. Obtain the resistance sampling sequence before the air pump starts and the resistance sampling sequence of the first short-time suction section at the same air intake probe position; for any resistance channel number, add the resistance sampling values ​​before the air pump starts bit by bit and divide by the number of sampling bits to obtain the reference resistance value, and then subtract the reference resistance value from the resistance value of each sampling bit in the first short-time suction section to generate the reference subtraction sequence. S13. For adjacent sampling bits under the same resistance channel number in the reference subtraction sequence, subtract the reference subtraction value of the current sampling bit from the reference subtraction value of the next sampling bit to generate the channel difference; write a positive sign if the channel difference is greater than zero, write a zero sign if the channel difference is equal to zero, and write a negative sign if the channel difference is less than zero. Write the difference sign into the corresponding gas source group according to the channel calibration table to generate the first suction response word.

[0006] In a preferred embodiment, S2 includes: S21. Read the sampling clock and air intake probe position number of the intelligent sensing system at the moment the air pump stops, take the moment the air pump stops as the starting position of the pump stop waiting segment, and copy the number of sampling positions of the first short-term suction segment to generate the sampling position number of the pump stop waiting segment. S22. Obtain the sampling sequence of the pump stop waiting section resistor. For any adjacent sampling bit under any resistor channel number, subtract the previous value from the next value. If the difference is greater than zero, write a positive sign; if the difference is equal to zero, write a zero sign; if the difference is less than zero, write a negative sign. Generate the pump stop difference symbol string.

[0007] In a preferred embodiment, S2 further includes: S23. Based on the pump stop difference symbol string, sequentially scan the sampling bits under the same resistor channel number. When a positive symbol is read for the first time, write it to the rising memory bit. When a negative symbol is read for the first time, write the negative symbol sampling bit number as the residual release bit. When no negative symbol is read, write the residual release bit to an empty bit. S24. Read the gas source group corresponding to the residual release position according to the channel calibration table, use the resistance channel number as the table key, and use the gas source group and residual release position as the table value. Write the residual release sequence table from front to back according to the residual release position to generate a pump stop recovery record.

[0008] In a preferred embodiment, S3 includes: S31. Obtain the resistance sampling sequence of the second short-time suction segment at the same air intake probe position according to the air pump speed and suction duration of the first short-time suction segment. Perform the subtraction of the previous value from the subsequent value on adjacent sampling positions under the same resistance channel number. Write a positive sign if the difference is positive, write a zero sign if the difference is zero, and write a negative sign if the difference is negative to generate a second difference symbol string. S32. Read the residual release bit according to the residual release sequence table. For the positive sign sampling bit under the same resistance channel number in the second difference symbol string, subtract the residual release bit number from the sampling bit number. If the residual release bit is empty or the calculation result is negative, write it into the supplementary observation item. If the calculation result is zero or positive, write it into the residual observation item and generate the source observation table. S33. Establish a status cell in the source observation table that corresponds to the sampling number and the source status. The source status adopts a binary code. The zero code corresponds to the residual source and the one code corresponds to the supplementary source. For the status cell of sampling number one, write the cumulative value to zero when the corresponding observation item is hit, write the cumulative value to one when the corresponding observation item is not hit, and write the empty backtracking pointer.

[0009] In a preferred embodiment, S3 further includes: S34. When executing the Viterbi alignment algorithm, for each state cell after the sampling bit number one, read the zero code state cell and the one code state cell of the previous sampling bit respectively. If the state codes are the same, write the transition value zero; if the state codes are different, write the transition value one. If the current sampling bit hits the corresponding observation, write the observation value zero; if it misses the corresponding observation, write the observation value one. Add the previous cumulative value, transition value, and observation value to generate a candidate value. Take the sequence number one according to the ascending order of the candidate value and the ascending order of the previous state code and write it into the current state cell. S35. Read the status grid of the last sampled bit, take the first number as the last source status according to the ascending order of cumulative cost value and status code, read the backtracking pointer in reverse to the first sampled bit, and then rewrite the source status as the residual source mark or the supplemented source mark according to the ascending order of the sampled bit, to generate the second suction source mark sequence.

[0010] In a preferred embodiment, S4 includes: S41. Within the edge computing node, the particle number is generated by concatenating the resistance channel number of the second short-time suction segment with the sampling bit number. The alignment source code in the second suction source mark sequence, the symbol code in the second difference symbol string, the group code in the channel calibration table, and the air pump status code are read according to the particle number. The initial particle status word is written and the weight code is written as one. S42. In the prediction operation of the particle filter algorithm, read the air pump status code in the initial particle state word. When the air pump status code is the start code, write the prediction source code as the fill code. When the air pump status code is the stop code, write the prediction source code as the residual code. Replace the source code in the initial particle state word with the prediction source code to generate the prediction particle state word.

[0011] In a preferred embodiment, S4 further includes: S43. Perform an XOR operation between the prediction source code in the predicted particle state word and the aligned source code of the same particle number. When the XOR result is zero, add two to the weight code. When the XOR result is one, subtract one from the weight code. When the weight code is negative, rewrite it to zero to generate the updated particle state word. S44. First, sort the updated particle status words in descending order by weight code value, then sort them in ascending order by particle number, and then copy the updated particle status words in a cyclic manner according to the sorting order until the number of copies is equal to the product of the number of resistor channels and the number of sampling bits in the second short-time suction segment, and generate a resampled particle table. S45. Read the particle status words in the resampled particle table whose predicted source code is the supplement code and whose symbol code is a positive sign, and write them into the exhaled gas supplement estimation record according to the group code corresponding to the particle number.

[0012] In a preferred embodiment, S5 includes: S51. Group the particle status words in the exhaled gas replenishment estimation record according to the source segment. Under each source segment, establish a general exhalation replenishment position and a human trace replenishment position. When the symbol code of the particle status word is a positive sign and the group code is the general exhalation code, write the particle number into the general exhalation replenishment position. When the symbol code of the particle status word is a positive sign and the group code is the human trace code, write the particle number into the human trace replenishment position to generate a replenishment position record. S52. Based on the supplementary position record, read the general exhalation supplementary position and the human trace supplementary position for each source segment. When the general exhalation supplementary position has a particle number, write the general exhalation confirmation position. When the human trace supplementary position has a particle number, write the human trace confirmation position. Perform a bitwise AND operation on the two confirmation positions to generate the live human existence word. S53. When the presence of a living human is 1, write the alarm status to the display status and write the sound status to the buzzer status; when the presence of a living human is zero, write the retest status to the display status and write the mute status to the buzzer status.

[0013] A gas detection instrument for searching for human vital signs, the instrument comprising a sampling calibration module, a pump stop and residual gas release module, an alignment and source determination module, a filtering and estimation module, and a same-source alarm module: The sampling and calibration module is used to acquire the first short-time suction segment resistance sampling sequence and channel calibration table output by the intelligent sensing system at the same air inlet probe position. The resistance sampling sequence comes from the absorbing fluid or the solid sensitive layer whose resistance changes after reacting with the fluid. The channel calibration table records the resistance channel number and the gas source group. The module subtracts the previous value from the subsequent value of adjacent sampling positions of the same resistance channel and generates the first suction response word according to the difference sign. The pump stop and residual release module is used to stop the air pump and maintain the position of the air inlet probe after the first short-term suction section ends. It acquires the resistance sampling sequence of the pump stop waiting section, generates a pump stop difference symbol string by subtracting the previous value from the later value, writes the first sampling bit that converts the positive sign to the negative sign into the residual release sequence table, and outputs the pump stop recovery record. The alignment and source determination module is used to obtain the resistance sampling sequence of the second short-time suction segment according to the same air pump speed and suction duration of the first short-time suction segment, generate the second difference symbol string, take the channel order of the residual release sequence table as the residual state chain, take the positive sign sampling bit order of the second difference symbol string as the supplementary state chain, and output the second suction source mark sequence through the Viterbi alignment algorithm. The filtering estimation module is used to generate particle state words based on the second suction source marker sequence within the edge computing node, predict the source segment according to the air pump status, update the particle weight according to the number of times the source segment matches the second suction source marker sequence, and output the exhaled gas to replenish the estimation record through the particle filtering algorithm. The same-source alarm module is used to read the estimated record of exhaled gas replenishment. Within the same source segment, it writes the general exhaled gas replenishment position and the human trace replenishment position according to the gas source group and positive sign. After both types of replenishment positions are written, the result of the presence of a living human is generated and output to the display screen and buzzer.

[0014] The technical effects and advantages of this invention are as follows: 1. This solution uses the same-site sampling of the first short-term suction, pump stop waiting and the second short-term suction to make the exhaled gas replenishment process a basis for judgment, reducing the risk of misjudging the limited exhaled air mass as no vital signs after it is thinned out. 2. Write the fallback sampling position in the pump stop waiting period into the residual release priority table so that the residual response in the pipeline and sensor array is recorded first, and then used for identification and reading by the second short-term suction, which relatively reduces false alarms caused by residual gas. 3. Viterbi alignment is performed on the second short-time suction segment so that the positive response first distinguishes between residual sources and replenishment sources before entering the subsequent estimation, which relatively improves the stability of the judgment when the sampling responses before and after are inconsistent; 4. Perform particle filtering within the edge computing node, and use the pump status, source marker, and resistance sign together for the supplementary estimation to reduce the reliance of on-site judgment on remote analysis; 5. By using both general expiratory replenishment sites and human trace replenishment sites within the same source segment to confirm the live results, the alarm is simultaneously constrained by two types of expiratory characteristics, thus relatively reducing the probability of a single response triggering an alarm. Attached Figure Description

[0015] Figure 1 This is a flowchart outlining the method steps of the present invention; Figure 2 This is a schematic diagram of the instrument module structure of the present invention. Detailed Implementation

[0016] 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, and 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.

[0017] Refer to the instruction manual appendix Figure 1-2 The present invention provides a gas detection method for searching human vital signs, comprising: S1. Obtain the first short-time suction segment resistance sampling sequence and channel calibration table output by the intelligent sensing system at the same air intake probe position. The resistance sampling sequence comes from the absorbing fluid or the solid sensitive layer whose resistance changes after reacting with the fluid. The channel calibration table records the resistance channel number and the gas source group. Subtract the previous value from the subsequent value of adjacent sampling positions of the same resistance channel and generate the first suction response word according to the difference sign. In this embodiment, during the first short-term inhalation, the change in resistance of the solid sensitive layer output by the intelligent sensing system is converted into a first inhalation response word that can be read for subsequent residual release identification and replenishment source identification. The principle is to first classify each resistance channel into the exhaled gas source group, then use the resistance state before the air pump starts as the reference for the same air intake probe position, and finally write the direction of resistance change in the first short-term inhalation segment as a symbolic response, so that the initial sampling results formed by the air intake probe, pipeline, air pump and sensor array can be mapped to the exhaled gas analysis in gaseous biological substances. This implementation process includes the following steps: S11 is used to establish a traceable correspondence between the resistance channel and the exhaled gas source group; the input is the channel registration record, the sensitive layer reaction record and the exhaled gas source record in the intelligent sensing system, wherein the channel registration record contains the resistance channel number and the solid sensitive layer number, the sensitive layer reaction record contains the solid sensitive layer number and the reaction gas number, and the exhaled gas source record contains the reaction gas number and the gas source group; During processing, for any resistance channel number, first read the corresponding solid sensitive layer number from the channel registration record, then read the reaction gas number from the sensitive layer reaction record using the solid sensitive layer number, and then read the gas source group from the exhalation source record using the reaction gas number. The resistance channel number and gas source group are then written as channel calibration items. The output is a channel calibration table, which is used by S12 and S13 to read the gas source group according to the resistance channel number. If the resistance channel number is missing a solid sensitive layer number, then an uncalibrated channel mark is written; if the reactive gas number is missing a gas source group, then the resistance channel number will not participate in the generation of the first suction response word; if one solid sensitive layer number corresponds to multiple reactive gas numbers, then channel calibration items are generated separately in ascending order of reactive gas numbers, and are not merged into one channel calibration item. S12 is used to remove the resistance baseline formed at the same inlet probe position before the air pump starts, so that the response of the first short-time suction section comes from the gas entering after suction, rather than from the background state of the solid sensitive layer; the input quantities are the resistance sampling sequence before the air pump starts and the resistance sampling sequence of the first short-time suction section at the same inlet probe position. Both types of resistance sampling sequences include the resistance channel number, sampling position number, resistance value, sampling clock, inlet probe position number and air pump status code; During processing, for any resistance channel number, the resistance sampling values ​​before the air pump starts are added bit by bit, and the sum is divided by the number of sampling positions before the air pump starts to obtain the reference resistance value. Then, the reference resistance value is subtracted from the resistance value of each sampling position in the first short-time suction segment to obtain the reference deduction value of that sampling position. The output is the reference deduction sequence, which is used by S13 to calculate the change direction of adjacent sampling positions. If the number of sampling positions is zero before the air pump starts, the resistor channel number is written into the reference missing mark; if the resistance value of a certain sampling position is missing in the first short-time suction segment, the sampling position is written into the empty sampling mark, and the empty sampling mark is not included in the calculation of the difference between adjacent sampling positions. S13 is used to convert the resistance change direction in the first short-time suction section into a sign response under the source group, serving as the starting reference for subsequent pump stop recovery and second short-time suction alignment; the input quantities are the reference subtraction sequence and the channel calibration table; During processing, for adjacent sampling bits under the same resistor channel number in the reference subtraction sequence, the reference subtraction value of the current sampling bit and the reference subtraction value of the next sampling bit are read, and the reference subtraction value of the current sampling bit is subtracted from the reference subtraction value of the next sampling bit to generate the channel difference. Before writing the difference sign, first read the baseline fluctuation range of the same resistance channel number in the resistance sampling sequence before the air pump starts. The baseline fluctuation range is generated by the deviation of the resistance value of each sampling bit from the reference resistance value before the air pump starts. Write a zero sign when the channel difference falls within the baseline fluctuation range, write a positive sign when the channel difference is higher than the upper boundary of the baseline fluctuation range, and write a negative sign when the channel difference is lower than the lower boundary of the baseline fluctuation range. If the channel difference is greater than zero, write a positive sign; if the channel difference is equal to zero, write a zero sign; if the channel difference is less than zero, write a negative sign. Then, read the gas source group corresponding to the resistor channel number according to the channel calibration table, and write the difference sign into the sampling bit of the corresponding gas source group. The output is the first suction response word, which contains at least the resistor channel number, gas source group, sampling bit number, and difference sign, for the pump stop waiting section to generate the residual release sequence table. If the current sampling bit or the next sampling bit is marked as an empty sampling bit, then no channel difference will be generated for that adjacent sampling bit; if the resistor channel number in the channel calibration table is marked as an uncalibrated channel, then the difference sign for that resistor channel number will not be written into the gas source group. Through the above processing, the first short-term suction segment is no longer just a mixed gas reading, but is converted into a first suction response word with a resistance channel source, a reference subtraction result and a sampling position change direction, which can provide the same data caliber for subsequent judgment of "residual release after pump stoppage" and "replenishment of exhaled gas during the second short-term suction". In practical applications: After rescuers insert the air intake probe into the same position in the collapsed cavity, the sensor array records the resistance baseline of the solid sensitive layer before the air pump is started. After the air pump is started, the gas in the cavity is extracted for a short time. The intelligent sensing system calibrates the resistance channels corresponding to the general exhalation source and the trace source of human body, and writes the resistance rise, hold or fall caused by suction into the first suction response word, so that subsequent steps can identify whether there is human exhaled gas that can still be recovered after being taken away by the suction action at this position.

[0018] S2. After the first short-term suction phase ends, stop the air pump and maintain the position of the air inlet probe. Obtain the resistance sampling sequence of the pump stop waiting phase. Generate a pump stop difference symbol string by subtracting the previous value from the subsequent value. Write the first sampling bit that converts the positive sign to a negative sign into the residual release sequence table and output the pump stop recovery record. In this embodiment, S2 is used to record the resistance drop process during the pump stop waiting period after the first short-term suction phase ends, and to write the release sequence caused by residual gas in the inlet probe, pipeline, air pump, and sensor array into a residual release order table. During processing, the pump stop waiting period is first determined using the air pump stop time and the number of sampling points in the first short-term suction phase. Then, the resistance values ​​at the same inlet probe position and under the same resistance channel number are processed by difference symbolization. Subsequently, the residual release points are extracted from the difference symbol changes. Finally, the residual release points are bound to the gas source group and written into the pump stop recovery record for the second short-term suction phase to establish a residual state chain. This implementation process includes the following steps: S21 is used to determine the sampling range of the pump stop waiting section and ensure that the pump stop waiting section and the first short-term suction section come from the same inlet probe position; the input quantities are the pump stop time, the sampling clock of the intelligent sensing system, the current inlet probe position number, the first short-term suction section record and the first short-term suction section suction parameter record, wherein the first short-term suction section record includes the inlet probe position number, and the first short-term suction section suction parameter record includes the number of sampling positions; During processing, the current intake probe position number is read at the moment the air pump stops and compared with the intake probe position number recorded in the first short-term suction segment. If they are the same, the sampling clock that appears first after the moment the air pump stops is taken as the sampling position number of the pump stop waiting segment. Then, the sampling position numbers of the pump stop waiting segment are generated sequentially according to the sampling clock order until the number of sampling positions of the pump stop waiting segment is equal to the number of sampling positions of the first short-term suction segment. The output includes the sampling position number of the pump stop waiting segment, the starting position of the pump stop waiting segment, and the position record of the pump stop waiting segment. These are written into the current sampling record of the intelligent sensing system and read by S22. If the current intake probe position number is different from the intake probe position number in the first short-term suction segment record, a position change flag is written to the pump stop waiting segment position record, the pump stop recovery record for this round stops generating, and the resistance sampling sequence of the first short-term suction segment at the same intake probe position is re-acquired; if the number of sampling bits in the first short-term suction segment is zero, a sampling quantity missing flag is written to the pump stop waiting segment position record, and the pump stop difference symbol string for this round is not generated; if no sampling clock is read after the pump stops, a clock missing flag is written to the pump stop waiting segment position record, and the pump stop recovery record for this round stops generating. S22 is used to convert the pump stop waiting section resistance sampling sequence into a pump stop difference symbol string arranged by resistance channel number and sampling position number; the input is the pump stop waiting section sampling position number, the pump stop waiting section position record and the pump stop waiting section resistance sampling sequence. The pump stop waiting section resistance sampling sequence includes the resistance channel number, sampling position number, resistance value, sampling clock, air inlet probe position number and air pump status code. During processing, first, select the resistance sampling values ​​whose intake probe position number is equal to the intake probe position number and whose sampling position number belongs to the sampling position number of the pump stop waiting section in the position record. Then, for any resistance channel number, read the resistance value of the current sampling position and the resistance value of the next sampling position in ascending order of sampling position number. Subtract the resistance value of the current sampling position from the resistance value of the next sampling position to generate the pump stop channel difference. Write a positive sign if the pump stop channel difference is greater than zero, write a zero sign if the pump stop channel difference is equal to zero, and write a negative sign if the pump stop channel difference is less than zero. The output is a pump stop difference symbol string, which contains at least the resistor channel number, the current sampling bit number, the next sampling bit number, and the difference symbol, for S23 to scan the residual release bit; If the resistance value of the current sampling bit or the resistance value of the next sampling bit is missing, then an empty symbol is written to the corresponding current sampling bit. The empty symbol does not participate in the generation of the residual release bit. If there is a sampling bit in the pump status code of the pump waiting segment resistance sampling sequence that is the start code, then a status conflict flag is written to the corresponding sampling bit. The sampling bit with the status conflict flag does not participate in the generation of the pump stop difference symbol string. If the same resistance channel number has no valid resistance value in the sampling bit number of the pump stop waiting segment, then an empty channel flag is written to the resistance channel number. S23 is used to extract the sampling bit of the residual gas release from the pump stop difference symbol string, where the gas release changes from rising to falling; the input is the pump stop difference symbol string. During processing, the sampling bits under the same resistor channel number are scanned in ascending order of the current sampling bit number. If a negative or zero sign is read before the rising memory bit is written, the residual release bit is not written. When a positive sign is read for the first time, the current sampling bit number corresponding to the positive sign is written to the rising memory bit; after the rising memory bit is written, scanning continues. When a zero sign is read, scanning continues. When a negative sign is read for the first time, the current sampling bit number corresponding to the negative sign is written as the residual release bit, and the scanning of the resistor channel number ends; the output is the residual release bit corresponding to the resistor channel number, which is used by S24 to write the residual release sequence table. If no negative sign is read after the rising memory bit is written, the residual release bit is written empty; if the stop difference symbol string under the same resistor channel number is all empty, the residual release bit is written empty; if the same resistor channel number has an empty channel mark, the residual release bit is written empty; if a negative sign appears before the rising memory bit is written, the negative sign is only saved as the stop phase fallback record and is not used as the residual release bit. S24 is used to bind the residual release position to the gas source group and form a residual release sequence table for the second short-time suction section to read; the input quantities are the residual release position and the channel calibration table; During processing, for resistor channels with non-empty residual release positions, the corresponding gas source group is read from the channel calibration table, the resistor channel number is used as the table entry key, and the gas source group and residual release position are concatenated to form the table entry value. Then, all entries are arranged in ascending order of residual release position. If the residual release positions are the same, they are arranged in ascending order of resistance channel number. The arranged entries are written into the residual release sequence table. The output is the residual release sequence table and the pump stop recovery record, which are used by the second short-time suction section to read the residual status chain. Among them, the residual release position is only used as a residual reference when the source is aligned in the second short-time suction segment. It does not generate a live human presence word separately, nor does it directly trigger the display screen or buzzer. The resistor channel number with position change mark, state conflict mark, uncalibrated residual item, empty channel mark or empty position record is not written into the residual state chain. If the gas source group is not read in the channel calibration table, the corresponding resistance channel number is written into the uncalibrated residual item, and the uncalibrated residual item does not enter the residual state chain; if the residual release position is empty, the corresponding resistance channel number is not written into the residual release sequence table, but the empty position record is retained in the pump stop recovery record; if the residual release sequence table is empty, after reading the empty table in the second short-time suction stage, the positive sign sampling bit of the corresponding resistance channel number is written into the supplementary observation item. Through the above processing, the pump stop waiting period is converted into a pump stop recovery record with resistance channel number, gas source group and residual release position, so that the exhaled gas response remaining in the inlet probe, pipeline and sensor array after the first short-term suction period is formed in sequence; after the second short-term suction period, this sequence is used to distinguish between residual release in pipeline and new exhaled gas replenishment in the buried space. In practical applications: After rescuers insert the air intake probe into the collapsed cavity, the first short-term suction phase removes the exhaled gas that has accumulated near the cavity entrance. After the air pump stops, the intelligent sensing system continues to record the resistance value of the solid sensitive layer. When the exhaled gas-related resistance channel first shows a positive sign and then a negative sign, the negative sign sampling bit is written as the residual release bit. The subsequent second short-term suction phase uses the residual release sequence table to determine whether the new positive sign response comes from the continuation of the gas path residue or from the exhaled gas that has been re-injected in the same buried space.

[0019] S3. Obtain the resistance sampling sequence of the second short-time suction segment according to the same air pump speed and suction duration of the first short-time suction segment, generate the second difference symbol string, take the channel order of the residual release order table as the residual state chain, take the positive sign sampling bit order of the second difference symbol string as the supplementary state chain, and output the second suction source mark sequence through the Viterbi alignment algorithm. In this embodiment, S3 is used to distinguish between residual gas path continuation and re-injection of exhaled gas within the second short-term suction segment. During processing, the second short-term suction segment uses the same pump speed and suction duration as the first short-term suction segment, and maintains the same inlet probe position number. Then, the second difference symbol string is mapped to the residual release position in the pump stop recovery record to form a source observation table. The path cost of residual source and re-injection source is then calculated bit by bit using the Viterbi alignment algorithm, and finally, a second suction source marker sequence is generated for the particle filtering step within the edge computing node to read. This implementation process includes the following steps: S31 is used to obtain the direction of resistance change in the second short-time suction segment; the input quantities are the suction parameter record of the first short-time suction segment, the record of the first short-time suction segment, the current air intake probe position number, and the resistance sampling sequence of the second short-time suction segment, wherein the suction parameter record of the first short-time suction segment includes the air pump speed code, the suction duration code, and the number of sampling bits; During processing, first read the intake probe position number recorded in the first short-term suction segment and compare it with the current intake probe position number. If they are the same, start the air pump according to the suction parameter record of the first short-term suction segment and collect the resistance sampling sequence of the second short-term suction segment under the same intake probe position number. Then, for adjacent sampling bits under the same resistor channel number, subtract the current sampling bit's resistor value from the next sampling bit's resistor value. If the difference is positive, write a positive sign; if the difference is zero, write a zero sign; if the difference is negative, write a negative sign. The output is a second difference sign string, which is written into the resistor channel number, the current sampling bit number, the next sampling bit number, and the difference sign, for S32 to read the positive sign sampling bit. If the current intake probe position number is different from the intake probe position number recorded in the first short-term suction segment, the position change mark is written into the resistance sampling sequence of the second short-term suction segment, and the identification of the source of the second suction segment stops; if there is a missing resistance value in an adjacent sampling position, an empty symbol is written into the corresponding sampling position, and the empty symbol is not entered into the source observation table. S32 is used to allocate the positive sign sampling bits in the second short-time suction phase as residual observations or supplementary observations; the inputs are the second difference symbol string and the residual release order table; During processing, for the same resistor channel number, read the residual release bit in the residual release sequence table, and then read the positive sign sampling bit in the second difference symbol string; When the residual release bit is empty, the sampling bit number subtraction is not performed, and the positive sign sampling bit under the resistor channel number is directly written into the supplementary observation item; when the residual release bit is not empty, the relative bit number difference is generated by subtracting the residual release bit number from the positive sign sampling bit number. When the relative position difference is negative, the supplementary observation item is written; when the relative position difference is zero or positive, the residual observation item is written. A relative position difference of zero indicates that the inlet of the second short-term suction section still receives the residual release from the pump stop waiting section. The output is the source observation table, which is used by S33 to establish the status grid. If the resistor channel number in the residual release priority table is an unlabeled residual item, then no residual observation item will be generated for that resistor channel number; if there is no positive sign sampling bit in the second difference symbol string, then an empty observation item will be written for that resistor channel number in the source observation table. S33 is used to establish the initial state grid for the Viterbi alignment algorithm; the input is the source observation table; during processing, the state grid is established according to the sampling position number. The state grid includes the sampling position number, state code, observation cost, transition cost, cumulative cost, and backtracking pointer. The state code adopts a binary code. Zero code is written as the residual source, and one code is written as the supplementary source. For sampling bit number 1, establish a zero-code state cell and a one-code state cell respectively. When sampling bit number 1 hits the observation item corresponding to the state code, the cumulative cost value of the state cell is written to zero. When sampling bit number 1 does not hit the observation item corresponding to the state code, the cumulative cost value of the state cell is written to one, and the backtracking pointer is written to null. The output is the initial state cell, which is used by S34 to accumulate the path cost bit by bit. If the sampling bit number is an empty observation, the cumulative value of both the zero-code state cell and the one-code state cell is written as 1; the value of zero and the value of one are given directly by the state hit rule and are not read from external empirical values. S34 is used to calculate the source path for each sample bit using the Viterbi alignment algorithm; the inputs are the initial state grid and the source observation table. During processing, a zero-code state cell and a one-code state cell are established for each sampled bit after the first sampled bit, and the zero-code state cell and one-code state cell of the previous sampled bit are read for the current state cell. When the previous state code is the same as the current state code, the transfer value is written to zero; when the previous state code is different from the current state code, the transfer value is written to one. When the current sampling bit matches the observation item corresponding to the current status code, the observation value is written to zero; when the current sampling bit does not match the observation item corresponding to the current status code, the observation value is written to one. Then, the previous cumulative value, the transfer value, and the observation value are added together to obtain two candidate values, which are then sorted in ascending order of candidate values. The source of the observation cost value is the source observation table formed by S32. The source observation table divides the sampled bits into residual observation items or supplementary observation items according to the relative position difference between the positive sign sampling position number and the residual release position number of the second short-time suction segment. The relative position difference corresponds to the sampling timing difference between the second short-time suction response and the pump stop residual release. The positive sign in the second difference symbol string corresponds to the direction of resistance change of adjacent sampling positions. The residual release position comes from the recovery timing of the first rise and then fall in the pump stop waiting segment. The second short-time suction segment uses the air pump speed and suction duration of the first short-time suction segment, so that the source observation table is generated under the same suction conditions. Therefore, a hit writes zero and a miss writes one, indicating the matching relationship between the current source state and the pump stop residual release timing, the second short-time suction supplementary timing, and the same air pump suction conditions. When candidate values ​​are the same, they are sorted in ascending order according to the previous state code. The candidate value with the first index is written into the cumulative value of the current state cell, and the corresponding previous state code is written into the backtracking pointer of the current state cell. The output is a Viterbi path table, which is used by S35 to read the source state in reverse. If a state cell in the previous sampled bit is empty, then that state cell will not participate in the candidate value calculation; if both candidate values ​​are empty, then the current state cell will be marked with an empty path. S35 is used to convert the Viterbi path table into the second suction source marker sequence; the input is the Viterbi path table; during processing, the zero code status cell and one code status cell of the last sample bit are read and arranged in ascending order according to the cumulative cost value. If the cumulative cost values ​​are the same, they are arranged in ascending order according to the status code. The first number is taken as the last source state, and then the backtracking pointer corresponding to the last source state is read bit by bit in reverse to the first sample bit. After the reverse reading is completed, the source states read are arranged in ascending order of sampling bit number, and the zero code is rewritten as the residual source marker, and the one code is rewritten as the supplementary source marker; The output is the second suction source marker sequence. The second suction source marker sequence is written with the resistor channel number, sampling bit number and source marker, which is used by S4 to generate the particle state word. If the last sample bit in the Viterbi path table is empty, the second suction source marker sequence is written with an empty source marker, and the empty source marker does not enter the particle state word; Through the above processing, the positive sign response of the second short-term suction segment does not directly enter the liveness determination. Instead, it first calculates the positional relationship with the residual release position formed in the pump stop waiting segment, and then obtains the source mark of each sampling position through Viterbi path backtracking, so that the source division of the gas path residual continuation and the re-injection of exhaled gas is completed before entering the particle filter. In practical applications: Rescuers keep the air intake probe in the same position in the collapsed cavity. The first short suction phase removes the exhaled gas that has accumulated near the cavity inlet. The pump stop waiting phase records the residual release position of the pipeline and sensor array. The second short suction phase samples again at the same pump speed and suction duration. When the positive sign sampling position of the exhaled gas related resistance channel is earlier than the residual release position, the Viterbi path tendency writes the replenishment source mark. When the positive sign sampling position appears after the residual release position, the Viterbi path tendency writes the residual source mark. Subsequent edge computing nodes use this to determine whether the second response comes from the replenishment of new exhaled gas in the buried space or from the continuation of residual gas in the gas path.

[0020] S4. Generate particle status words in the edge computing node according to the second suction source marker sequence, predict the source segment according to the air pump status, update the particle weight according to the number of times the source segment is consistent with the second suction source marker sequence, and output the exhaled gas replenishment estimation record through the particle filtering algorithm. In this embodiment, S4 is used to recursively correct the source determination of the second short-term suction segment within the edge computing node, so that the aligned source code obtained by the Viterbi alignment algorithm, the symbol code obtained by the second difference symbol string, the group code obtained by the channel calibration table, and the air pump status code are all fed into the particle filtering algorithm. During processing, particle numbers are first generated according to the resistance channel number and sampling bit number, then the initial particle status word is generated, then the source code is predicted according to the air pump status, and the weight code is corrected with the aligned source code. Finally, the exhaled gas replenishment estimation record is obtained through resampling, which is then written to the general exhaled gas replenishment bit and the human trace replenishment bit by S5. This implementation process includes the following steps: S41 is used to convert the resistance channel number and sampling position number of each resistor channel in the second short-time suction section into a particle state that can be processed by the particle filter algorithm; the input quantities are the second suction source mark sequence, the second difference symbol string, the channel calibration table, the air pump state sequence, and the second short-time suction section sampling configuration record. During processing, the length of the resistance channel number field and the length of the sampling position number field in the sampling configuration record are read in the edge computing node. If the field length is insufficient, zeros are added to the left. Then, the resistance channel number field is concatenated with the sampling position number field to generate the particle number. The alignment source code, symbol code, group code and air pump status code are read according to the particle number and written into the initial particle status word. The output is the initial particle state word, which is read by S42. If the alignment source code is not read for the particle number, the alignment source code is written as an empty source code. If the group code is not read for the particle number, the group code is written as an unlabeled group code. The weight code uses a signed integer, and the initial weight code is written as one according to the rules. S42 is used to predict source changes during the second short-term suction phase based on the air pump's operating status; the input is the initial particle status word. During processing, the air pump status code in the initial particle status word is read in the prediction operation of the particle filter algorithm. When the air pump status code is the start code, it indicates that the second short-time suction stage is pumping gas from the buried cavity into the sensor array. The prediction source code is written as the supplement code. When the air pump status code is a stop code, it indicates that the current sampling position is in the process of suction interruption or cross-segment residual reading. The predicted source code is written as the residual code. Then, the predicted source code is used to replace the source code in the initial particle status word to generate the predicted particle status word. The output is the predicted particle state word, which is used by S43 to perform weight correction. If the air pump state code is an empty state code, the predicted source code of the previous sample bit of the same particle number is read and used. If the previous sample bit does not exist, the predicted source code is written as an empty source code. The empty source code does not participate in the XOR correction in S43. When the air pump status code is the on code, the supplement code is written, which only represents the prediction source code in the particle filter algorithm prediction operation. The prediction source code is not directly written into the live human existence word. The prediction source code needs to be XORed with the aligned source code of the same particle number in S43, and after resampling in S44 and filtering by supplement code, positive sign and group code in S45, it will be entered into the exhaled gas supplementation estimation record. S43 is used to correct the prediction results of the particle filter algorithm using the Viterbi alignment result; the input is the predicted particle state word and the alignment source code corresponding to the same particle number; During processing, the prediction source code and alignment source code in the predicted particle state word are XORed. When the XOR result is zero, it means that the particle prediction result is the same as the Viterbi alignment result. The weight code is incremented by two before entering S43. When the XOR result is one, it indicates that the particle prediction result is different from the Viterbi alignment result, and the weight code is decremented by one from the value before entering S43; when the weight code calculation result is negative, it is rewritten to zero; the output is to update the particle state word for S44 sorting and copying. If the predicted source code or the aligned source code is an empty source code, the weight code retains the value before entering S43, and a source missing flag is written in the updated particle state word; S44 is used to retain the particle state word with a high weight code and a clear particle numbering order through resampling; the input is the updated particle state word, the number of resistor channels, and the number of sampling bits in the second short-time suction segment. During processing, the particle status words are first updated in descending order of weight code value. If the weight code values ​​are the same, they are updated in ascending order of particle number. Then, starting from the first updated particle status word after the arrangement, the order is copied. After copying to the last updated particle status word, the process returns to the first updated particle status word and continues to copy until the number of copies is equal to the product of the number of resistor channels and the number of sampling bits in the second short-time suction segment. The output is a resampled particle table, which is used by S45 to extract and supplement the source. If the weight codes of all updated particle status words are zero, the cyclic copying is still performed in ascending order of particle number, so that the number of records in the resampled particle table is equal to the product of the number of resistor channels and the number of sampling bits in the second short-time suction segment. S45 is used to extract exhaled gas from the resampled particle table and add it to the estimation record; the input is the resampled particle table. During processing, the particle state word whose prediction source code is the supplement code and whose symbol code is a positive sign is read. The corresponding group code is read according to the particle number, and the particle number, supplement code, positive sign code, group code, weight code and sampling position number are written into the exhaled gas supplement estimation record. The output is the exhaled gas replenishment estimation record, which is read and written by S5 to the general exhaled gas replenishment position and the human trace replenishment position; if the group code of the particle status word is an unlabeled group code, then the particle status word is not written to the exhaled gas replenishment estimation record. If there is no particle state word in the resampled particle table whose predicted source code is the fill code and whose symbol code is a positive sign, then the exhaled gas fill estimation record is written to an empty fill flag. Through the above processing, the positive sign response in the second short-time suction segment is first aligned to form an aligned source code through the Viterbi alignment algorithm, and then the source prediction, weight correction and resampling are performed in the edge computing node through the particle filter algorithm, so that the exhaled gas replenishment estimation record can reflect the new exhaled gas replenishment under the same air intake probe position, rather than just reflecting the continuation of the gas path residue. In practical applications: rescuers keep the air intake probe in the same position in the collapsed cavity. During the second short-term suction phase, local gas is re-extracted. The edge computing node performs an XOR comparison between the replenishment code obtained by Viterbi alignment and the replenishment code predicted under the air pump on state. When the exhaled gas related resistance channel shows a positive sign during the second short-term suction phase, and the resampled particle table retains the replenishment code, the corresponding particle status word is written into the exhaled gas replenishment estimation record. Subsequent steps use this to determine whether there is both general exhaled gas replenishment and trace human replenishment at the same time.

[0021] S5. Read the exhaled gas replenishment estimation record, and write the general exhaled gas replenishment position and human trace replenishment position according to the gas source group and positive sign within the same source segment. After both types of replenishment positions are written, generate the live human existence result and output it to the display screen and buzzer. In this embodiment, S5 is used to convert the exhaled gas replenishment estimation record into an actionable result of the presence of a living human on-site. During processing, the edge computing node first reads the particle status word in groups according to the source fragment, then writes the general exhaled gas replenishment position and the trace human replenishment position into the same source fragment. Subsequently, it generates a living human presence word through a confirmation bit AND operation, and writes the living human presence word into the display status word and the buzzer status word, so that the exhaled gas replenishment evidence obtained by the air intake probe at the same position is converted into an on-site search and rescue prompt. This implementation process includes the following steps: S51 is used to decompose the exhaled gas replenishment estimation record into two types of replenishment positions; the input is the exhaled gas replenishment estimation record, which at least includes particle number, source segment, symbol code, group code, weight code and sampling position number; Among them, the source segment represents the set of sampling positions in the same group that are marked as replenishment sources and retained in the resampled particle table within the same inlet probe position and the same second short-time suction segment, after being marked as replenishment sources by the second suction source marker sequence; the particle status words within the same source segment have the same inlet probe position number, the same second short-time suction segment number and the same replenishment source marker, which are used to distinguish between replenishment events in the same round and isolated ambient gas responses or pipeline residual responses. During processing, the particle state words in the exhaled gas replenishment estimation record are grouped according to the source segment, and a general exhaled gas replenishment position and a human trace replenishment position are established under each source segment; When the symbol code of the particle status word is a positive sign and the group code is the general exhalation code, the particle number is written into the general exhalation fill-in position; when the symbol code of the particle status word is a positive sign and the group code is the human trace code, the particle number is written into the human trace fill-in position. When writing multiple particle numbers to the same fill bit, they are first written in ascending order of sampling bit number. If the sampling bit numbers are the same, they are written in ascending order of resistor channel number. The output is the fill bit record, which is read by S52. If the source segment of the particle status word is an empty source segment, the particle status word will not be entered into the fill-in record; if the group code of the particle status word is an unlabeled group code, the particle status word will not be written into the general exhalation fill-in or human trace fill-in; if the exhaled gas fill-in estimation record is an empty fill-in marker, the fill-in record will be written into an empty record. S52 is used to generate a live human presence word based on the simultaneous writing of two types of padding bits under the same source segment; the input is the padding bit record. During processing, the general exhalation replenishment position and the human trace replenishment position are read from each source segment. When a particle number exists in the general exhalation replenishment position, the general exhalation confirmation position is written as 1. When the general exhalation replenishment position is empty, the general exhalation confirmation position is written as 0. When a particle number exists in the human trace fill-in position, the human trace confirmation position is written as 1; when the human trace fill-in position is empty, the human trace confirmation position is written as 0. Then, a bitwise AND operation is performed on the general exhalation confirmation position and the human trace confirmation position under the same source segment, and the result is written as the source confirmation word. When the source confirmation word of any source segment is 1, the live human existence word is written as 1; when the source confirmation words of all source segments are zero, the live human existence word is written as zero; the output is the live human existence word, which is used by S53 to write the status word of the display screen and the status word of the buzzer. Among them, the "Live Human Presence" indicates the gas detection prompt result formed after the source, positive sign and corresponding group code are written together at the same air intake probe position and the same source segment for both general expiratory replenishment and human trace replenishment. The "Live Human Presence" is used for on-site search and rescue prompts and does not indicate a medical confirmation conclusion of leaving the aspiration position, aspiration sequence and source segment conditions. If the entry record is empty, then the general exhalation confirmation position, human trace confirmation position, source confirmation word, and living human presence word should all be written as zero; S53 is used to write the presence of a living human to the on-site notification component; the input values ​​are the presence of a living human, the status word on the display screen, and the status word on the buzzer. During processing, when the presence of a living human is 1, the display status word is written to the alarm status code, and the buzzer status word is written to the sound status code; when the presence of a living human is zero, the display status word is written to the retest status code, and the buzzer status word is written to the mute status code; the output is the display execution record and the buzzer execution record, which are available for rescue personnel to read on-site. If the display status word writing fails, the buzzer status word writing is retained and a non-response mark is written to the display execution record; if the buzzer status word writing fails, the display status word writing is retained and a non-response mark is written to the buzzer execution record. Through the above processing, the exhaled gas replenishment estimation record only generates a live human presence word when both a general exhaled gas replenishment position and a human trace replenishment position are formed simultaneously under the same source segment, thereby separating the new exhaled gas replenishment from the continuation of gas path residue and entering the alarm result. The result of the presence of living humans output by this method refers to the gas detection prompt result that meets the conditions of distinguishing residual release after pump stoppage, identifying the source of the second short-term suction replenishment, and the dual-group replenishment of the same source segment within the confined search and rescue space. If the response of environmental putrefactive gas, combustion residue, external personnel exhalation, animal-derived gas or pipeline residue does not simultaneously form a general exhalation replenishment position and a human trace replenishment position within the same source segment, no alarm status code will be generated. In practical applications: After rescuers complete the second short-term suction while holding the air intake probe in the same position in the collapsed cavity, if the edge computing node simultaneously records positive sign particle status words for both general exhalation sources and trace human sources within the same source segment, the display screen writes an alarm status code and the buzzer sounds; if only a single source replenishment or residual source replenishment is recorded, the display screen writes a retest status code and the buzzer remains silent, allowing rescuers to continue retesting at the same air intake probe position or move to an adjacent gap for sampling.

[0022] Furthermore, the present invention also includes a gas detection instrument for searching for human vital signs, the instrument comprising a sampling calibration module, a pump stop and residual release module, an alignment and source determination module, a filtering and estimation module, and a same-source alarm module: The sampling and calibration module is used to acquire the first short-time suction segment resistance sampling sequence and channel calibration table output by the intelligent sensing system at the same air inlet probe position. The resistance sampling sequence comes from the absorbing fluid or the solid sensitive layer whose resistance changes after reacting with the fluid. The channel calibration table records the resistance channel number and the gas source group. The module subtracts the previous value from the subsequent value of adjacent sampling positions of the same resistance channel and generates the first suction response word according to the difference sign. The pump stop and residual release module is used to stop the air pump and maintain the position of the air inlet probe after the first short-term suction section ends. It acquires the resistance sampling sequence of the pump stop waiting section, generates a pump stop difference symbol string by subtracting the previous value from the later value, writes the first sampling bit that converts the positive sign to the negative sign into the residual release sequence table, and outputs the pump stop recovery record. The alignment and source determination module is used to obtain the resistance sampling sequence of the second short-time suction segment according to the same air pump speed and suction duration of the first short-time suction segment, generate the second difference symbol string, take the channel order of the residual release sequence table as the residual state chain, take the positive sign sampling bit order of the second difference symbol string as the supplementary state chain, and output the second suction source mark sequence through the Viterbi alignment algorithm. The filtering estimation module is used to generate particle state words based on the second suction source marker sequence within the edge computing node, predict the source segment according to the air pump status, update the particle weight according to the number of times the source segment matches the second suction source marker sequence, and output the exhaled gas to replenish the estimation record through the particle filtering algorithm. The same-source alarm module is used to read the estimated record of exhaled gas replenishment. Within the same source segment, it writes the general exhaled gas replenishment position and the human trace replenishment position according to the gas source group and positive sign. After both types of replenishment positions are written, the result of the presence of a living human is generated and output to the display screen and buzzer.

[0023] Working principle: This scheme first uses the air intake probe to perform a short-term suction at the same position. The intelligent sensing system reads the resistance change of the solid sensitive layer and generates the first suction response word according to the gas source group corresponding to the resistance channel. Then, the air pump is stopped while keeping the air intake probe position unchanged, and the sampling position of each resistance channel from rising to falling during the pump stop waiting period is recorded to form a residual release priority table. Then, a second short-term suction is performed at the same air pump speed and suction duration. The Viterbi alignment algorithm is used to determine whether the second resistance response belongs to the continuation of gas path residual or the replenishment of new exhaled gas. The source judgment is corrected by particle filtering in the edge computing node. Finally, after confirming the general exhaled gas replenishment and the human trace replenishment in the same source segment, the presence of a living human is generated, and the display screen and buzzer alarm are activated. For example, in building collapse rescue, rescuers insert an air intake probe into the gaps in the rubble. The first short suction removes the exhaled gas that has accumulated near the gap entrance. After the pump stops, the equipment records the release sequence of the residual gas in the pipeline and sensor array. Then, a second short suction is performed at the same location. If the second positive resistance response is determined to be exhaled gas that has been re-introduced into the buried space, rather than the previous residual gas, and includes both general exhaled gas sources and trace human sources, the equipment will sound an alarm on site. If only residual continuation or a single source response occurs, the equipment remains in retesting mode, and rescuers continue to sample at the location or adjacent gaps.

[0024] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A gas detection method for searching human vital signs, characterized in that, include: S1. Obtain the first short-time suction segment resistance sampling sequence and channel calibration table output by the intelligent sensing system at the same air intake probe position. The resistance sampling sequence comes from the absorbing fluid or the solid sensitive layer whose resistance changes after reacting with the fluid. The channel calibration table records the resistance channel number and the gas source group. Subtract the previous value from the subsequent value of adjacent sampling positions of the same resistance channel and generate the first suction response word according to the difference sign. S2. After the first short-term suction phase ends, stop the air pump and maintain the position of the air inlet probe. Obtain the resistance sampling sequence of the pump stop waiting phase. Generate a pump stop difference symbol string by subtracting the previous value from the subsequent value. Write the first sampling bit that converts the positive sign to a negative sign into the residual release sequence table and output the pump stop recovery record. S3. Obtain the resistance sampling sequence of the second short-time suction segment according to the same air pump speed and suction duration of the first short-time suction segment, generate the second difference symbol string, take the channel order of the residual release order table as the residual state chain, take the positive sign sampling bit order of the second difference symbol string as the supplementary state chain, and output the second suction source mark sequence through the Viterbi alignment algorithm. S4. Generate particle status words in the edge computing node according to the second suction source marker sequence, predict the source segment according to the air pump status, update the particle weight according to the number of times the source segment is consistent with the second suction source marker sequence, and output the exhaled gas replenishment estimation record through the particle filtering algorithm. S5. Read the exhaled gas replenishment estimation record, and write the general exhaled gas replenishment position and the human trace replenishment position according to the gas source group and positive sign within the same source segment. After both types of replenishment positions are written, generate the live human existence result and output it to the display screen and buzzer.

2. The gas detection method for searching human vital signs according to claim 1, characterized in that: S1 includes: S11. Obtain the channel registration record, sensitive layer reaction record, and exhalation source record within the intelligent sensing system; for any resistance channel number, read the solid sensitive layer number from the channel registration record, then read the reaction gas number from the sensitive layer reaction record using the solid sensitive layer number, and subsequently read the gas source group from the exhalation source record using the reaction gas number. Write the resistance channel number and gas source group as channel calibration items to generate a channel calibration table. S12. Obtain the resistance sampling sequence before the air pump starts and the resistance sampling sequence of the first short-time suction section at the same air intake probe position; for any resistance channel number, add the resistance sampling values ​​before the air pump starts bit by bit and divide by the number of sampling bits to obtain the reference resistance value, and then subtract the reference resistance value from the resistance value of each sampling bit in the first short-time suction section to generate the reference subtraction sequence. S13. For adjacent sampling bits under the same resistance channel number in the reference subtraction sequence, subtract the reference subtraction value of the current sampling bit from the reference subtraction value of the next sampling bit to generate the channel difference; write a positive sign if the channel difference is greater than zero, write a zero sign if the channel difference is equal to zero, and write a negative sign if the channel difference is less than zero. Write the difference sign into the corresponding gas source group according to the channel calibration table to generate the first suction response word.

3. The gas detection method for searching human vital signs according to claim 2, characterized in that: S2 includes: S21. Read the sampling clock and air intake probe position number of the intelligent sensing system at the moment the air pump stops, take the moment the air pump stops as the starting position of the pump stop waiting segment, and copy the number of sampling positions of the first short-term suction segment to generate the sampling position number of the pump stop waiting segment. S22. Obtain the sampling sequence of the pump stop waiting section resistor. For any adjacent sampling bit under any resistor channel number, subtract the previous value from the next value. If the difference is greater than zero, write a positive sign; if the difference is equal to zero, write a zero sign; if the difference is less than zero, write a negative sign. Generate the pump stop difference symbol string.

4. A gas detection method for searching human vital signs according to claim 3, characterized in that: S2 also includes: S23. Based on the pump stop difference symbol string, sequentially scan the sampling bits under the same resistor channel number. When a positive symbol is read for the first time, write it to the rising memory bit. When a negative symbol is read for the first time, write the negative symbol sampling bit number as the residual release bit. When no negative symbol is read, write the residual release bit to an empty bit. S24. Read the gas source group corresponding to the residual release position according to the channel calibration table, use the resistance channel number as the table key, and use the gas source group and residual release position as the table value. Write the residual release sequence table from front to back according to the residual release position to generate a pump stop recovery record.

5. A gas detection method for searching human vital signs according to claim 4, characterized in that: S3 includes: S31. Obtain the resistance sampling sequence of the second short-time suction segment at the same air intake probe position according to the air pump speed and suction duration of the first short-time suction segment. Perform the subtraction of the previous value from the subsequent value on adjacent sampling positions under the same resistance channel number. Write a positive sign if the difference is positive, write a zero sign if the difference is zero, and write a negative sign if the difference is negative to generate a second difference symbol string. S32. Read the residual release bit according to the residual release sequence table. For the positive sign sampling bit under the same resistance channel number in the second difference symbol string, subtract the residual release bit number from the sampling bit number. If the residual release bit is empty or the calculation result is negative, write it into the supplementary observation item. If the calculation result is zero or positive, write it into the residual observation item and generate the source observation table. S33. Establish a status cell in the source observation table that corresponds to the sampling number and the source status. The source status adopts a binary code. The zero code corresponds to the residual source and the one code corresponds to the supplementary source. For the status cell of sampling number one, write the cumulative value to zero when the corresponding observation item is hit, write the cumulative value to one when the corresponding observation item is not hit, and write the empty backtracking pointer.

6. A gas detection method for searching human vital signs according to claim 5, characterized in that: S3 also includes: S34. When executing the Viterbi alignment algorithm, for each state cell after the sampling bit number one, read the zero code state cell and the one code state cell of the previous sampling bit respectively. If the state codes are the same, write the transition value zero; if the state codes are different, write the transition value one. If the current sampling bit hits the corresponding observation, write the observation value zero; if it misses the corresponding observation, write the observation value one. Add the previous cumulative value, transition value, and observation value to generate a candidate value. Take the sequence number one according to the ascending order of the candidate value and the ascending order of the previous state code and write it into the current state cell. S35. Read the status grid of the last sampled bit, take the first number as the last source status according to the ascending order of cumulative cost value and status code, read the backtracking pointer in reverse to the first sampled bit, and then rewrite the source status as the residual source mark or the supplemented source mark according to the ascending order of the sampled bit, to generate the second suction source mark sequence.

7. A gas detection method for searching human vital signs according to claim 6, characterized in that: S4 includes: S41. Within the edge computing node, the particle number is generated by concatenating the resistance channel number of the second short-time suction segment with the sampling bit number. The alignment source code in the second suction source mark sequence, the symbol code in the second difference symbol string, the group code in the channel calibration table, and the air pump status code are read according to the particle number. The initial particle status word is written and the weight code is written as one. S42. In the prediction operation of the particle filter algorithm, read the air pump status code in the initial particle state word. When the air pump status code is the start code, write the prediction source code as the fill code. When the air pump status code is the stop code, write the prediction source code as the residual code. Replace the source code in the initial particle state word with the prediction source code to generate the prediction particle state word.

8. A gas detection method for searching human vital signs according to claim 7, characterized in that: S4 also includes: S43. Perform an XOR operation between the prediction source code in the predicted particle state word and the aligned source code of the same particle number. When the XOR result is zero, add two to the weight code. When the XOR result is one, subtract one from the weight code. When the weight code is negative, rewrite it to zero to generate the updated particle state word. S44. First, sort the updated particle status words in descending order by weight code value, then sort them in ascending order by particle number, and then copy the updated particle status words in a cyclic manner according to the sorting order until the number of copies is equal to the product of the number of resistor channels and the number of sampling bits in the second short-time suction segment, and generate a resampled particle table. S45. Read the particle status words in the resampled particle table whose predicted source code is the supplement code and whose symbol code is a positive sign, and write them into the exhaled gas supplement estimation record according to the group code corresponding to the particle number.

9. A gas detection method for searching human vital signs according to claim 8, characterized in that: S5 includes: S51. Group the particle status words in the exhaled gas replenishment estimation record according to the source segment. Under each source segment, establish a general exhalation replenishment position and a human trace replenishment position. When the symbol code of the particle status word is a positive sign and the group code is the general exhalation code, write the particle number into the general exhalation replenishment position. When the symbol code of the particle status word is a positive sign and the group code is the human trace code, write the particle number into the human trace replenishment position to generate a replenishment position record. S52. Based on the supplementary position record, read the general exhalation supplementary position and the human trace supplementary position for each source segment. When the general exhalation supplementary position has a particle number, write the general exhalation confirmation position. When the human trace supplementary position has a particle number, write the human trace confirmation position. Perform a bitwise AND operation on the two confirmation positions to generate the live human existence word. S53. When the presence of a living human is 1, write the alarm status to the display status and write the sound status to the buzzer status; when the presence of a living human is zero, write the retest status to the display status and write the mute status to the buzzer status.

10. A gas detection instrument for searching human vital signs, used to implement the gas detection method for searching human vital signs as described in any one of claims 1-9, the instrument comprising a sampling calibration module, a pump stop and residual release module, an alignment and source determination module, a filtering and estimation module, and a same-source alarm module, characterized in that: The sampling and calibration module is used to acquire the first short-time suction segment resistance sampling sequence and channel calibration table output by the intelligent sensing system at the same air inlet probe position. The resistance sampling sequence comes from the absorbing fluid or the solid sensitive layer whose resistance changes after reacting with the fluid. The channel calibration table records the resistance channel number and the gas source group. The module subtracts the previous value from the subsequent value of adjacent sampling positions of the same resistance channel and generates the first suction response word according to the difference sign. The pump stop and residual release module is used to stop the air pump and maintain the position of the air inlet probe after the first short-term suction section ends. It acquires the resistance sampling sequence of the pump stop waiting section, generates a pump stop difference symbol string by subtracting the previous value from the later value, writes the first sampling bit that converts the positive sign to the negative sign into the residual release sequence table, and outputs the pump stop recovery record. The alignment and source determination module is used to obtain the resistance sampling sequence of the second short-time suction segment according to the same air pump speed and suction duration of the first short-time suction segment, generate the second difference symbol string, take the channel order of the residual release sequence table as the residual state chain, take the positive sign sampling bit order of the second difference symbol string as the supplementary state chain, and output the second suction source mark sequence through the Viterbi alignment algorithm. The filtering estimation module is used to generate particle state words based on the second suction source marker sequence within the edge computing node, predict the source segment according to the air pump status, update the particle weight according to the number of times the source segment matches the second suction source marker sequence, and output the exhaled gas to replenish the estimation record through the particle filtering algorithm. The same-source alarm module is used to read the estimated record of exhaled gas replenishment. Within the same source segment, it writes the general exhaled gas replenishment position and the human trace replenishment position according to the gas source group and positive sign. After both types of replenishment positions are written, the result of the presence of a living human is generated and output to the display screen and buzzer.

Images