Experimental process data recording method and system based on biometrics

Abstract

The present application relates to the field of data processing technology, and discloses a method, system and storage medium for recording experimental process data based on biometrics. The method comprises: performing voiceprint recognition and natural language processing on the real-time voice instructions of the detection operator to obtain experimental operation records; performing real-time collection and abnormal detection processing on the detection instrument data to obtain safety indicator data and potential risk alarms; performing step optimization analysis on the experimental operation records, safety indicator data and potential risk alarms to obtain detection step optimization suggestions; performing time series integration processing on the experimental operation records, safety indicator data, potential risk alarms and detection step optimization suggestions to obtain traceable experimental reports and detection safety level assessment results. The present application improves the efficiency and accuracy of experimental process data recording based on biometrics.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a method and system for recording experimental process data based on biometrics. Background Art

[0002] Existing scenario-specific detection and operation technologies cover a number of important areas, including daily medical testing, clinical bedside testing, patient-initiated POCT testing, food safety testing, and operation scenarios with preset overall processes such as surgery. These technologies play a key role in their respective fields, such as daily medical testing provides rapid assessment of health status, clinical bedside testing enables instant diagnosis, POCT enables patients to self-monitor at home, food safety testing ensures food quality, and surgical procedures ensure the accuracy and safety of medical interventions. The common feature of these technologies is that they need to follow preset operating steps, and the results have an important impact on safety and health.

[0003] However, the detection and operation technologies in these specific scenarios still face some common challenges. First, the operation process is easily affected by human factors, which may lead to missed steps, wrong sequence or improper operation, affecting the accuracy of the results. Secondly, data recording and management often rely on manual operations, which are prone to information loss or incorrect recording. Furthermore, these processes usually lack real-time monitoring and instant feedback mechanisms, making it difficult to detect and correct potential problems in a timely manner. In addition, differences in skill levels between different operators may lead to inconsistent results. Finally, these technologies usually lack systematic quality control and continuous optimization mechanisms, making it difficult to achieve long-term efficiency improvements and risk management. Summary of the invention

[0004] The present application provides a method and system for recording experimental process data based on biometrics, which are used to improve the efficiency and accuracy of experimental process data recording based on biometrics.

[0005] In the first aspect, the present application provides a method for recording experimental process data based on biometrics, and the method for recording experimental process data based on biometrics includes: performing operation record analysis on the real-time voice instructions of the detection operator to obtain an experimental operation record; based on the experimental operation record, real-time collection and abnormal detection processing of the detection instrument data to obtain safety indicator data and potential risk alarms; performing step optimization analysis on the experimental operation record, the safety indicator data and the potential risk alarm to obtain detection step optimization suggestions; performing time serialization integration processing on the experimental operation record, the safety indicator data, the potential risk alarm and the detection step optimization suggestions to obtain a traceable experimental report and detection safety level assessment results.

[0006] In combination with the first aspect, in a first implementation method of the first aspect of the present application, the operation record analysis of the real-time voice instructions of the detection operator to obtain the experimental operation record includes: based on a preset biometric library, biometric collection and identity authentication processing of the detection operator to obtain an operator identity confirmation result and an experiment authorization mark; based on the operator identity confirmation result and the experiment authorization mark, voiceprint recognition and natural language processing are performed on the real-time voice instructions of the detection operator to obtain an experimental operation record that matches the preset experimental step library.

[0007] In combination with the first aspect, in the second implementation method of the first aspect of the present application, the detection operator is subjected to biometric collection and identity authentication processing based on a preset biometric library to obtain an operator identity confirmation result and an experiment authorization mark, including: collecting and processing the experimental instruction voice of the detection operator to obtain original voiceprint data related to the experiment, and performing laboratory background noise elimination processing on the original voiceprint data related to the experiment to obtain target detection voiceprint data; performing experimental operation-related feature extraction processing on the target detection voiceprint data to obtain a detection voiceprint feature vector, and performing laboratory environment adaptability normalization processing on the detection voiceprint feature vector to obtain a standardized detection voiceprint feature; performing similarity comparison between the standardized detection voiceprint feature and the pre-stored detector voiceprint template in the biometric library. Calculate and process to obtain a detection voiceprint matching score; collect and process the facial image of the detection operator after wearing the protective equipment to obtain the detection protection state original face image; perform feature extraction on the detection protection state original face image to obtain a protection state face feature vector; perform similarity calculation on the protection state face feature vector and the pre-stored detector protection state face template in the biometric library to obtain the detection protection state face matching score; perform weighted fusion processing on the detection voiceprint matching score and the detection protection state face matching score based on the detection security level to obtain the detection comprehensive identity matching score; based on the preset detection authority threshold, perform graded comparison processing on the detection comprehensive identity matching score to obtain the operator identity confirmation result and the experiment authorization mark.

[0008] In combination with the first aspect, in a third implementation of the first aspect of the present application, based on the operator identity confirmation result and the experimental authorization mark, the real-time voice instructions of the detection operator are subjected to voiceprint recognition and natural language processing to obtain an experimental operation record that matches a preset experimental step library, including: collecting and processing the real-time voice instructions of the detection operator to obtain original voice data, and filtering the original voice data with laboratory environmental noise to obtain purified voice data; framing and feature extraction processing are performed on the purified voice data to obtain a voice feature vector, and matching processing is performed on the voice feature vector with a voiceprint model associated with the operator identity confirmation result to obtain a voiceprint consistency score; based on the voiceprint The purified voice data is decoded based on the consistency score and the experimental authorization mark to obtain a voice instruction text; the voice instruction text is processed for security terminology recognition to obtain a key experimental step phrase; the key experimental step phrase is processed for context association to obtain structured experimental step description data; the structured experimental step description data is matched with the preset experimental step library to obtain step matching probability distribution data; based on the step matching probability distribution data, the structured experimental step description data is selected for the best match to obtain the current experimental operation step; the current experimental operation step, the operator identity confirmation result and the current timestamp are associated to obtain the experimental operation record.

[0009] In combination with the first aspect, in a fourth implementation of the first aspect of the present application, based on the experimental operation record, the detection instrument data is collected and processed for abnormality detection in real time to obtain safety indicator data and potential risk alarms, including: performing POCT detection step parsing processing on the experimental operation record to obtain a rapid detection step identifier, and based on the rapid detection step identifier, performing data collection trigger processing on the corresponding POCT device to obtain an original safety indicator data stream; performing interference elimination processing on the original safety indicator data stream to obtain pure safety indicator data; performing standard unit conversion and POCT device sensitivity calibration processing on the pure safety indicator data to obtain standardized Indicator value; performing indicator category correlation analysis on the standardized indicator value to obtain indicator distribution characteristics; based on a preset indicator threshold data set, performing over-standard risk determination on the indicator distribution characteristics to obtain indicator abnormality marks; performing indicator abnormality factor tracing analysis on the indicator abnormality marks to obtain abnormal trend characteristics; performing evaluation on the abnormal trend characteristics to obtain potential risk levels; performing detection safety level association processing on the standardized indicator value and the potential risk level to obtain the safety indicator data, and based on the potential risk level and preset indicator early warning rules, performing detection safety level classification processing on the potential risks to obtain the potential risk alarms.

[0010] In combination with the first aspect, in the fifth implementation method of the first aspect of the present application, the step optimization analysis of the experimental operation records, the safety indicator data and the potential risk alarm is performed to obtain detection step optimization suggestions, including: performing detection process decomposition processing on the experimental operation records to obtain a fine-grained operation step sequence; performing correlation matching processing on the safety indicator data and the potential risk alarm to obtain a step-level risk indicator mapping; based on the fine-grained operation step sequence and the step-level risk indicator mapping, performing time efficiency evaluation processing on the experimental process to obtain step time distribution; performing comprehensive analysis processing on the step time distribution and the step-level risk indicator mapping to obtain an initial optimization plan; performing compliance verification processing on the initial optimization plan and the preset standard operating procedures to obtain compliance optimization suggestions; based on the compliance optimization suggestions and the current laboratory resource status collected in real time, performing feasibility evaluation processing on the initial optimization plan to obtain the detection step optimization suggestions.

[0011] In combination with the first aspect, in a sixth implementation method of the first aspect of the present application, the experimental operation records, the safety indicator data, the potential risk alarms and the detection step optimization suggestions are time-serialized and integrated to obtain a traceable experimental report and a detection safety level assessment result, including: extracting and sorting the experimental operation records with timestamps to obtain an experimental step time series; performing time correlation processing on the safety indicator data to obtain an indicator change trend chart; performing time node mapping processing on the potential risk alarms to obtain a risk event timeline; performing time correlation analysis processing on the detection step optimization suggestions to obtain an optimization suggestion implementation time point; based on the experimental step time series, the indicator change trend chart, the risk event timeline and the optimization suggestion implementation time point, performing multidimensional data integration processing through a time series data fusion algorithm to obtain a full-process detection time series data; performing key node extraction and causal relationship analysis processing on the full-process detection time series data to obtain a list of critical control points; based on the critical control point list and preset rating standards, performing a comprehensive evaluation processing on the entire experimental process to obtain the traceable experimental report and the detection safety level assessment result.

[0012] In combination with the first aspect, in the seventh implementation method of the first aspect of the present application, the key node extraction and causal analysis processing of the time series data of the entire detection process are performed to obtain a list of key control points, including: performing voiceprint recognition event marking processing on the time series data of the entire detection process to obtain a set of operator voice command time points; based on the preset experimental process verification nodes, the operator voice command time point set is classified and processed to obtain a sequential node set and a parallel node set; each node in the sequential node set is monitored and recorded in a predetermined order, and if a sequence error is detected, an alarm message is generated; the nodes in the parallel node set are recorded in a variable order. Processing, recording the actual execution order of the nodes, but not triggering an alarm for changes in the order; based on the sequential node set and the parallel node set, segmenting the safety indicator data to obtain a critical interval of indicator changes; comparing the critical interval of indicator changes with the preset standard indicator interval to obtain a list of potential hazard points; performing risk assessment on the list of potential hazard points to obtain a candidate set of key control points; performing time correspondence processing on the candidate set of key control points and the experimental operation records to obtain an operation-related control point mapping; based on the operation-related control point mapping and the detection step optimization suggestions, the control points are prioritized to obtain the key control point list.

[0013] In combination with the first aspect, in the eighth implementation method of the first aspect of the present application, the experimental operation record may be any one of the following operation records: daily medical testing operation records, clinical bedside testing operation records, POCT testing operation records, food safety testing operation records, and surgical operation records.

[0014] In a second aspect, the present application provides an experimental process data recording system based on biometrics, and the experimental process data recording system based on biometrics includes:

[0015] An analysis module is used to analyze the operation records of the real-time voice instructions of the detection operator to obtain the experimental operation records;

[0016] A detection module is used to collect and detect abnormalities of the detection instrument data in real time based on the experimental operation records to obtain safety indicator data and potential risk alarms;

[0017] An optimization module, used to perform step optimization analysis on the experimental operation records, the safety indicator data and the potential risk alarm to obtain optimization suggestions for the detection steps;

[0018] The integration module is used to perform time-series integration processing on the experimental operation records, the safety indicator data, the potential risk alarms and the detection step optimization suggestions to obtain a traceable experimental report and a detection safety level assessment result.

[0019] In the technical solution provided by the present application, by introducing biometric recognition technology, accurate verification of the identity of the detection operator is achieved, which greatly improves the safety and reliability of experimental operations and effectively prevents the risk of unauthorized personnel performing operations. By using voiceprint recognition and natural language processing technology, the operator's voice instructions can be captured and understood in real time and converted into structured experimental operation records, which not only reduces errors in manual recording, but also improves the efficiency and accuracy of data recording; by real-time collection and analysis of data from detection instruments, potential abnormal situations can be quickly identified and risk alarms can be issued in a timely manner, which greatly improves the safety and controllability of the experimental process; based on the collected data, Real-time step optimization analysis provides optimization suggestions for operators, which can improve experimental efficiency and reduce resource waste; through the time-serialized integration of the data of the entire experimental process, highly traceable experimental reports and detection safety level assessment results are generated, which not only meets regulatory requirements, but also provides strong support for subsequent quality improvement and risk management; while ensuring data integrity and reliability, the present invention also greatly reduces manual intervention and reduces the possibility of human errors. It can also promptly detect and respond to abnormal situations in the experimental process, greatly improving the efficiency and reliability of food safety testing, and significantly improving the reliability and traceability of experimental data and the safety of the overall experimental process. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0021] Figure 1 A schematic diagram of an embodiment of a method for recording experimental process data based on biometrics in an embodiment of the present application;

[0022] Figure 2 This is a schematic diagram of an embodiment of performing operation record analysis on a real-time voice command of a detection operator in an embodiment of the present application;

[0023] Figure 3 This is a schematic diagram of an embodiment of an experimental process data recording system based on biometrics in an embodiment of the present application. DETAILED DESCRIPTION

[0024] The present application embodiment provides a kind of experimental process data recording method and system based on biometrics. The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or precedence. It should be understood that the data used in this way can be interchangeable in appropriate circumstances, so that the embodiments described here can be implemented in a sequence other than the content illustrated or described here. In addition, the term "including" or "having" and any variation thereof are intended to cover non-exclusive inclusions, for example, the process, method, system, product or equipment comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or equipment.

[0025] For ease of understanding, the specific process of the embodiment of the present application is described below. Figure 1 In the embodiment of the present application, an embodiment of the method for recording experimental process data based on biometrics includes:

[0026] Step S101, analyzing the operation records of the real-time voice instructions of the detection operator to obtain the experimental operation records;

[0027] It is understandable that the execution subject of the present application can be a biometric-based experimental process data recording system, or a terminal or a server, which is not specifically limited here. The present application embodiment is described by taking the server as the execution subject as an example.

[0028] Specifically, by establishing a preset biometric database, the biometrics of the detection operator are collected and the identity verification is processed, so as to obtain the operator's identity confirmation result and the experimental authorization mark. The specific implementation process is as follows: First, the system collects the experimental instruction voice of the detection operator to obtain the original voiceprint data related to the experiment, and then eliminates the laboratory background noise of these data to obtain the target detection voiceprint data. The experimental operation-related features of the target detection voiceprint data are extracted to obtain the detection voiceprint feature vector, and the laboratory environment adaptability is normalized to generate the standardized detection voiceprint feature. Then, the standardized detection voiceprint feature is calculated with the food inspector voiceprint template pre-stored in the biometric database for similarity to obtain the detection voiceprint matching score. At the same time, the system will also collect the facial image of the detection operator after wearing protective equipment to obtain the original face image of the detection protection state. The features of these images are extracted to obtain the protection state face feature vector. The protection state face feature vector is calculated with the protection state face template of the food inspector pre-stored in the biometric database for similarity to obtain the detection protection state face matching score. After that, the system performs weighted fusion processing on the detection voiceprint matching score and the detection protection status face matching score based on the detection security level to obtain the detection comprehensive identity matching score. Finally, the detection comprehensive identity matching score is graded and compared to obtain the operator identity confirmation result and the experiment authorization mark. This multimodal biometric recognition method greatly improves the accuracy and reliability of identity authentication. For example, in a food safety testing laboratory, an operator is preparing to conduct a pesticide residue detection experiment. The system first collects the operator's voice command "Start pesticide residue detection experiment", and obtains the standardized voiceprint feature after noise elimination and feature extraction. This feature is matched with the pre-stored voiceprint template to obtain a voiceprint matching score of 0.92. At the same time, the system collects the facial image of the operator wearing a mask and goggles, extracts the features, and matches it with the pre-stored protection status face template to obtain a face matching score of 0.88. According to the safety level of pesticide residue detection, the system performs weighted fusion on the voiceprint and face matching scores to obtain a comprehensive matching score of 0.90. This score is higher than the preset threshold of 0.85, so the system confirms that the operator's identity is valid and authorizes him to conduct the pesticide residue detection experiment. This method not only ensures that only authorized personnel can conduct sensitive food safety testing experiments, but also adapts to the special situation in the laboratory environment where operators need to wear protective equipment, greatly improving the accuracy of identity authentication and the safety of experimental operations.

[0029] After obtaining the operator's identity confirmation result and the experimental authorization mark, the real-time voice instructions of the detection operator are further subjected to voiceprint recognition and natural language processing to obtain the experimental operation record that matches the preset experimental step library. First, the real-time voice instructions of the detection operator are collected and processed to obtain the original voice data, and then the data are subjected to laboratory environmental noise filtering to obtain purified voice data. Next, the purified voice data is framed and feature extracted to obtain voice feature vectors, and these vectors are matched with the voiceprint model associated with the operator's identity confirmation result to obtain the voiceprint consistency score. Based on the voiceprint consistency score and the experimental authorization mark, the purified voice data is decoded to obtain the voice instruction text. Then, the voice instruction text is processed for security term recognition to extract key experimental step phrases. These key experimental step phrases are processed for context association to generate structured experimental step description data. The structured experimental step description data is matched with the preset experimental step library to obtain the step matching probability distribution data. Based on these probability distribution data, the structured experimental step description data is subjected to the best matching selection process to determine the current experimental operation step. Finally, the current experimental operation steps, the operator identity confirmation results and the current timestamp are associated to obtain the experimental operation record. This method can not only accurately identify and record the experimental operation steps, but also ensure the compliance and traceability of the operation. For example, in an experiment to detect the heavy metal content in food, the operator issued a voice command "start sample digestion". First, the noise of this voice is filtered, and then the voice feature vector is extracted and matched with the operator's voiceprint model to obtain a voiceprint consistency score of 0.95. After confirming that the voiceprint matches and the operator has experimental authorization, the voice is decoded to obtain the text "start sample digestion". Next, the security term "sample digestion" is identified and combined with the context information to generate a structured step description "perform heavy metal detection sample digestion". This description is matched with the preset experimental step library, and the matching probability is 0.98. Based on this high matching probability, the current operation step is determined to be "heavy metal detection sample digestion", and this step is associated with the operator ID and the current timestamp (such as 2024-07-3010:15:30) to generate a complete experimental operation record. This method not only accurately records the experimental steps, but also ensures the compliance and traceability of the operations, greatly improving the reliability and efficiency of food safety testing.

[0030] Step S102: Based on the experimental operation records, real-time acquisition and abnormality detection processing of the detection instrument data are performed to obtain safety indicator data and potential risk alarms;

[0031] Specifically, based on the experimental operation records, the data of the detection instrument is collected in real time and anomaly detected, and the process of obtaining safety indicator data and potential risk alarms is as follows: First, the experimental operation records are parsed for food safety POCT (Point-of-Care Testing, i.e., on-site rapid testing) detection steps to obtain the rapid detection step identification. According to this identification, the corresponding POCT equipment is triggered to collect data and obtain the original safety indicator data stream. These original data streams are subjected to interference elimination processing to obtain pure safety indicator data. Then, the pure safety indicator data are converted to standard units and the POCT equipment sensitivity is calibrated to obtain standardized indicator values. The standardized indicator values ​​are subjected to indicator category correlation analysis processing to obtain indicator distribution characteristics. Based on the preset indicator threshold data set, the indicator distribution characteristics are subjected to over-standard risk determination processing to obtain indicator abnormality marks. Subsequently, the indicator abnormality marks are subjected to indicator abnormality factor tracing analysis processing to obtain abnormal trend characteristics. These abnormal trend characteristics are evaluated and processed to obtain potential risk levels. Finally, the standardized indicator values ​​and potential risk levels are processed with the detection safety level association to obtain safety indicator data, and based on the potential risk level and the preset indicator warning rules, the potential risks are processed with the detection safety level classification to obtain potential risk alarms. For example, in a pesticide residue detection experiment, the operation record shows that the current step is "organophosphorus pesticide detection". The system triggers the corresponding POCT equipment to collect data and obtains the original data stream. After matrix interference elimination, pure data is obtained. The system converts the data unit to mg / kg and calibrates it according to the equipment sensitivity to obtain a standardized indicator value of 0.15mg / kg. By comparing with the food safety standards for vegetables, the system finds that the value exceeds the limit standard of 0.1mg / kg for organophosphorus pesticides. The system immediately marks the indicator as abnormal, and through supply chain data analysis, it is found that similar abnormalities have occurred in the same batch of vegetables many times recently. Based on this trend, the system evaluates and obtains a high risk level. Finally, the system generates a safety indicator data report containing excess information and issues a high risk alarm to remind relevant personnel to take immediate measures. This method can not only monitor food safety indicators in real time, but also detect potential risks in a timely manner, greatly improving the efficiency and accuracy of food safety detection.

[0032] Step S103: Perform step optimization analysis on the experimental operation records, safety indicator data and potential risk alarms to obtain optimization suggestions for the detection steps;

[0033] Specifically, the test process is decomposed for the experimental operation records to obtain a fine-grained sequence of operation steps. The entire experimental process is broken down into a series of specific and quantifiable operation steps. Then, the safety indicator data and potential risk alerts are correlated and matched to obtain a step-level risk indicator mapping. This step associates each specific operation step with the corresponding safety indicator and risk alert to form a detailed risk distribution map. Next, based on the fine-grained operation step sequence and the step-level risk indicator mapping, the experimental process is evaluated for time efficiency and the step time distribution is obtained. This step identifies the links that may have efficiency problems by analyzing the time consumption of each operation step. Subsequently, the step time distribution and the step-level risk indicator mapping are comprehensively analyzed to obtain the initial optimization plan. This step comprehensively considers time efficiency and risk factors and proposes preliminary optimization suggestions. Next, the initial optimization plan is checked for compliance with the preset standard operating procedure (SOP) to obtain compliance optimization suggestions. This step ensures that all optimization suggestions comply with relevant safety standards and specifications. Finally, based on the compliance optimization suggestions and the current laboratory resource status collected in real time, the feasibility evaluation of the initial optimization plan is carried out to obtain the final optimization suggestions for the detection steps. This step takes into account the actual laboratory conditions and resource limitations to ensure that the optimization suggestions are practical. For example, in an experiment to detect the heavy metal content in food, the system first decomposes the experimental process into fine-grained steps such as sample pretreatment, digestion, and instrument analysis. By analyzing historical data, the system found that the sample digestion step took an average of 120 minutes, and 80% of potential risk alarms appeared in this step. The system calculated that the risk-time index of the sample digestion step was 96 (120 minutes × 80%), which is much higher than other steps. Based on this analysis, the system proposed an initial optimization plan: shorten the digestion time to 90 minutes and add intermediate checkpoints. The system then compared this plan with the heavy metal detection SOP and confirmed that the 90-minute digestion time still met the minimum requirements. Finally, the system checked the equipment status of the laboratory and found that new rapid digestion equipment was available. Based on this information, the system gives the final optimization suggestion: use a new type of rapid digestion equipment, shorten the digestion time to 60 minutes, and add an intermediate checkpoint at 30 minutes. This optimization plan not only improves experimental efficiency, but also reduces potential risks by increasing the inspection frequency, while ensuring compliance with standard operating procedures and practical feasibility.

[0034] Step S104: Perform time-series integration processing on the experimental operation records, safety indicator data, potential risk alerts and detection step optimization suggestions to obtain a traceable experimental report and detection safety level assessment results.

[0035] Specifically, the process of integrating experimental operation records, safety indicator data, potential risk alarms and detection step optimization suggestions through time series processing to obtain traceable experimental reports and detection safety level assessment results is as follows: First, extract and sort the experimental operation records through timestamps to obtain the experimental step time series. This step arranges all experimental operations in chronological order to form a complete experimental timeline. Next, the safety indicator data is time-correlated to obtain the indicator change trend chart. This step associates the numerical changes of various food safety indicators with time to form an intuitive trend chart. Then, the potential risk alarms are time-node mapped to obtain the risk event timeline. This step marks all risk alarms on the experimental timeline according to the time point of occurrence, clearly showing the time distribution of risk occurrence. Subsequently, the detection step optimization suggestions are time-correlated analyzed to obtain the implementation time point of the optimization suggestions. This step associates the optimization suggestions with the specific time points in the experimental process, indicating when it is most appropriate to implement these optimization measures. Based on the experimental step time series, indicator change trend chart, risk event timeline and optimization suggestion implementation time points, multi-dimensional data integration is performed through the time series data fusion algorithm to obtain the time series data of the entire detection process. This step integrates all relevant information into a unified time series framework to form a complete experimental process data set. The key nodes are extracted and the causal relationship analysis is processed for the time series data of the entire detection process to obtain a list of critical control points. This step identifies the most critical control points in the experimental process by analyzing the time relationship between data changes and risk occurrence. Finally, based on the list of critical control points and the preset rating standards, the entire experimental process is comprehensively evaluated to obtain a traceable experimental report and the test safety level assessment results. This step integrates all analysis results, generates a comprehensive and detailed experimental report, and gives the final test safety level assessment. For example, in a pesticide residue detection experiment, the system first extracts and sorts the timestamps of all operation records to form a complete time series from sample pretreatment to final data analysis. Then, the system associates the pesticide residue concentration data with time and generates a trend chart showing the change of pesticide concentration over time. Then, the system marks the two exceeded alarms at the time points after sample digestion and instrument analysis stage respectively. The optimization suggestion "Increase the number of cleaning times in the sample pretreatment step" is associated with the 30-minute time point after the start of the experiment. Through the time series data fusion algorithm, the system integrates this information into a unified time series data set. By analyzing this data set, the system identified the sample digestion step as the critical control point because the first over-limit alarm appeared after this step. Finally, the system generated a detailed experimental report, including the complete experimental process, data change trends, risk alarms and optimization suggestions.Based on the performance of key control points and the preset rating standards, the system gave a "B" test safety rating result, indicating that there is a certain risk but it is within the controllable range. This method not only provides a highly traceable experimental report, but also deeply reveals potential problems and improvement opportunities in the experimental process through time series analysis.

[0036] In the embodiment of the present application, by introducing biometric recognition technology, accurate verification of the identity of the detection operator is achieved, the safety and reliability of the experimental operation is greatly improved, and the risk of unauthorized personnel operating is effectively prevented. By using voiceprint recognition and natural language processing technology, the operator's voice instructions can be captured and understood in real time and converted into structured experimental operation records, which not only reduces the errors of manual recording, but also improves the efficiency and accuracy of data recording; by real-time collection and analysis of data from detection instruments, potential abnormal situations can be quickly identified and risk alarms can be issued in a timely manner, which greatly improves the safety and controllability of the experimental process; real-time analysis based on the collected data The process of optimizing the steps can be analyzed and optimization suggestions can be provided to the operator, which can improve the experimental efficiency and reduce the waste of resources. By integrating the time series of the data of the whole experimental process, a highly traceable experimental report and the detection safety level assessment result are generated, which not only meets the regulatory requirements, but also provides strong support for the subsequent quality improvement and risk management. While ensuring the integrity and reliability of the data, the present invention also greatly reduces the manual intervention and reduces the possibility of human error. It can also timely discover and respond to abnormal situations in the experimental process, which greatly improves the efficiency and reliability of food safety detection, and significantly improves the reliability and traceability of the experimental data and the safety of the overall experimental process.

[0037] In a specific embodiment, if Figure 2 As shown, the process of executing step S101 may specifically include the following steps:

[0038] S201, based on the preset biometric database, perform biometric collection and identity authentication processing on the detection operator to obtain the operator identity confirmation result and the experiment authorization mark;

[0039] S202: Based on the operator identity confirmation result and the experiment authorization mark, the real-time voice command of the detection operator is subjected to voiceprint recognition and natural language processing to obtain an experiment operation record matching a preset experiment step library.

[0040] In a specific embodiment, the process of executing step S201 may specifically include the following steps:

[0041] (1) Collecting and processing the experimental command voice of the detection operator to obtain the original voiceprint data related to the experiment, and performing laboratory background noise elimination processing on the original voiceprint data related to the experiment to obtain the target detection voiceprint data;

[0042] (2) Extracting experimental operation-related features from the target detection voiceprint data to obtain a detection voiceprint feature vector, and performing laboratory environment adaptability normalization on the detection voiceprint feature vector to obtain a standardized detection voiceprint feature;

[0043] (3) Calculate the similarity between the standardized detection voiceprint features and the pre-stored detector voiceprint template in the biometric database to obtain the detection voiceprint matching score;

[0044] (4) collecting and processing the facial image of the detection operator after wearing the protective equipment to obtain the original facial image of the detection protection state;

[0045] (5) performing feature extraction processing on the original face image in the detected protection state to obtain a face feature vector in the protection state;

[0046] (6) Calculate the similarity between the face feature vector of the protection state and the face template of the inspector of the protection state stored in the biometric database to obtain the face matching score of the protection state;

[0047] (7) Performing weighted fusion processing on the detection voiceprint matching score and the detection protection status face matching score based on the detection security level to obtain the detection comprehensive identity matching score;

[0048] (8) Based on the preset detection authority threshold, the comprehensive identity matching score of the detection is graded and compared to obtain the operator identity confirmation result and the experiment authorization mark.

[0049] Specifically, the experimental command voice of the detection operator is collected and processed to obtain the original voiceprint data related to the experiment. In this step, a high-sensitivity microphone array is used to collect the operator's voice commands and convert them into digital signals. Subsequently, the laboratory background noise is eliminated from the original voiceprint data related to the experiment to obtain the target detection voiceprint data. An adaptive filtering algorithm, such as the minimum mean square error (LMS) algorithm, is used to eliminate the background noise in the laboratory environment.

[0050] The experimental operation-related feature extraction process is performed on the target detection voiceprint data to obtain the detection voiceprint feature vector. This step uses the Mel Frequency Cepstral Coefficient (MFCC) algorithm to extract voiceprint features. It should be noted that the calculation process of MFCC is as follows:

[0051] 1. Pre-emphasize and frame the signal. 2. Perform Fast Fourier Transform (FFT) on each frame of the signal. 3. Map the spectrum to the Mel scale. 4. Perform logarithmic operation on the Mel spectrum. 5. Perform discrete cosine transform (DCT). The final MFCC feature vector can be expressed as:

[0052] MFCC=[c1,c2,...,c n ]; where c i Represents the i-th Mel-frequency cepstral coefficient.

[0053] Then, the detection voiceprint feature vector is normalized for laboratory environment adaptability to obtain the standardized detection voiceprint feature. This step uses the mean variance normalization (MVN) method, and the calculation formula is as follows:

[0054]

[0055] Among them, x is the original eigenvalue, μ is the characteristic mean, and σ is the characteristic standard deviation.

[0056] Then, the standardized detection voiceprint features and the pre-stored detector voiceprint template in the biometric database are similarly calculated to obtain the detection voiceprint matching score. The cosine similarity calculation method is used here, and the formula is as follows:

[0057]

[0058] Among them, x i and i are the i-th components of the feature to be matched and the template feature respectively.

[0059] At the same time, the facial images of the operator wearing protective equipment are collected and processed to obtain the original facial images of the protective state. These images are then subjected to feature extraction to obtain the facial feature vectors of the protective state. This step uses a convolutional neural network (CNN) to extract facial features and obtain a feature vector of a fixed dimension.

[0060] The similarity calculation is performed on the face feature vector of the protection state and the pre-stored face template of the inspector in the protection state in the biometric database to obtain the face matching score of the detection protection state. The cosine similarity calculation method is also used here.

[0061] The detection voiceprint matching score and the detection protection status face matching score are weighted and fused based on the detection security level to obtain the detection comprehensive identity matching score. The fusion formula is as follows:

[0062] score final =w1*score voice +w2*score face ;

[0063] Among them, w1 and w2 are weights determined according to the detection safety level, score voice and score face They are the voiceprint and face matching scores respectively.

[0064] Based on the preset detection authority threshold, the comprehensive identity matching score of the detection is graded and compared to obtain the operator identity confirmation result and the experiment authorization mark.

[0065] For example, in a food heavy metal content detection experiment, operator Zhang San put on a mask and protective glasses and said to the microphone "Start the lead content detection experiment". First, this voice is collected and the LMS algorithm is used to eliminate the background noise in the laboratory. Then, the voiceprint features are extracted using the MFCC algorithm to obtain a 13-dimensional feature vector. After MVN normalization, it is matched with the pre-stored Zhang San voiceprint template to obtain a similarity score of 0.92. At the same time, the system collects Zhang San's facial image after wearing protective equipment, uses CNN to extract features, and obtains a 128-dimensional feature vector. It is matched with the pre-stored Zhang San protective state face template to obtain a similarity score of 0.88. Considering the high risk of heavy metal detection, the system assigns weights of 0.6 and 0.4 to the voiceprint and face matching results respectively, and calculates a comprehensive matching score of 0.904. This score is higher than the preset threshold of 0.85. The system confirms that Zhang San's identity is valid and authorizes him to conduct the lead content detection experiment.

[0066] In a specific embodiment, the process of executing step S202 may specifically include the following steps:

[0067] (1) collecting and processing the real-time voice commands of the detection operator to obtain raw voice data, and filtering the raw voice data for laboratory environmental noise to obtain purified voice data;

[0068] (2) Frame and feature extract the purified voice data to obtain a voice feature vector, and match the voice feature vector with a voiceprint model associated with the operator identity confirmation result to obtain a voiceprint consistency score;

[0069] (3) Based on the voiceprint consistency score and the experimental authorization mark, the purified voice data is decoded to obtain the voice command text;

[0070] (4) Perform safety terminology recognition on the voice command text to obtain key experimental step phrases;

[0071] (5) context-related processing is performed on key experimental step phrases to obtain structured experimental step description data;

[0072] (6) Calculating the matching degree between the structured experimental step description data and the preset experimental step library to obtain step matching probability distribution data;

[0073] (7) Based on the step matching probability distribution data, the structured experimental step description data is optimally matched and selected to obtain the current experimental operation steps;

[0074] (8) The current experimental operation steps, the operator identity confirmation results and the current timestamp are associated to obtain the experimental operation record.

[0075] Specifically, the real-time voice instructions of the detected operator are collected and processed to obtain the original voice data. In this step, a high-sensitivity microphone array is used to collect the operator's voice instructions and convert them into digital signals. Subsequently, the original voice data is subjected to laboratory environmental noise filtering to obtain purified voice data. In this step, an adaptive filtering algorithm, such as a Wiener filter, is used to eliminate background noise in the laboratory environment, such as the sound of instrument operation and ventilation system. The purified voice data is framed and feature extracted to obtain a voice feature vector. The framing process divides the voice signal into several overlapping short-time frames, each frame is usually 20-30 milliseconds. Then, the Mel Frequency Cepstral Coefficient (MFCC) algorithm is used to extract the acoustic features of each frame to form a voice feature vector. Subsequently, these voice feature vectors are matched with the voiceprint model associated with the operator's identity confirmation result to obtain a voiceprint consistency score. In this step, a dynamic time warping (DTW) algorithm or a Gaussian mixture model (GMM) is used to calculate the similarity between the feature vector and the pre-stored voiceprint model.

[0076] Based on the voiceprint consistency score and the experimental authorization mark, the purified voice data is decoded to obtain the voice command text. This step uses a deep learning model, such as a speech recognition model based on a long short-term memory network (LSTM), to convert the voice signal into text. Next, the voice command text is processed for security terminology recognition to obtain key experimental step phrases. This step uses named entity recognition (NER) technology, combined with a pre-defined security term dictionary, to extract key experimental step phrases from the text. The key experimental step phrases are context-associated to obtain structured experimental step description data. This step uses dependency syntax analysis technology in natural language processing to clarify the grammatical relationship between phrases and form a structured experimental step description. Next, the structured experimental step description data is matched with the preset experimental step library to obtain the step matching probability distribution data. This step uses a text similarity algorithm, such as TF-IDF (term frequency-inverse document frequency) combined with cosine similarity, to calculate the matching degree between the current step description and each step in the preset step library.

[0077] Based on the step matching probability distribution data, the structured experimental step description data is processed for the best match selection to obtain the current experimental operation step. In this step, the step with the highest matching probability is selected as the current operation step. Finally, the current experimental operation step, the operator identity confirmation result and the current timestamp are associated to obtain the experimental operation record.

[0078] For example, in a food heavy metal content detection experiment, operator Li Si said "Start sample digestion procedure" into the microphone. First, this voice is collected and the background noise of the laboratory, such as the hum of the fume hood, is eliminated using a Wiener filter. Then, the system divides the purified voice signal into 25 millisecond frames and uses the MFCC algorithm to extract the 13-dimensional feature vector of each frame. These feature vectors are matched with Li Si's voiceprint model, and the voiceprint consistency score of 0.95 is calculated using the GMM algorithm.

[0079] Based on this high consistency score and the previous experimental authorization mark, the LSTM model is used to decode the speech into the text "Start sample digestion procedure". Then, the NER model identifies the keyword group "sample digestion" from it. Through dependency syntax analysis, a structured description is obtained: {action: "start", object: "sample digestion procedure"}. This structured description is matched with the preset experimental step library, and the highest matching probability of 0.92 is obtained using TF-IDF and cosine similarity calculations, and the corresponding step is "perform sample digestion". Finally, an experimental operation record is generated, including the operation step "perform sample digestion", the operator "Li Si" and the timestamp "2024-07-3010:15:30". Not only are the experimental steps accurately identified and recorded, but the compliance and traceability of the operation are also ensured.

[0080] In a specific embodiment, the process of executing step S102 may specifically include the following steps:

[0081] (1) Perform POCT test step analysis on the experimental operation record to obtain a rapid test step identifier, and based on the rapid test step identifier, perform data collection trigger processing on the corresponding POCT device to obtain the original safety indicator data stream;

[0082] (2) performing interference elimination processing on the original safety indicator data stream to obtain pure safety indicator data;

[0083] (3) The pure safety index data are converted to standard units and the POCT equipment sensitivity is calibrated to obtain standardized index values;

[0084] (4) Performing correlation analysis on the standardized indicator values ​​to obtain the indicator distribution characteristics;

[0085] (5) Based on the preset indicator threshold data set, the indicator distribution characteristics are processed to determine the risk of exceeding the standard and obtain the indicator abnormality mark;

[0086] (6) Conducting a retrospective analysis of abnormal factors on abnormal indicators to obtain abnormal trend characteristics;

[0087] (7) Evaluate and process abnormal trend characteristics to obtain potential risk levels;

[0088] (8) The standardized indicator values ​​and potential risk levels are correlated with the detection safety level to obtain safety indicator data, and based on the potential risk level and the preset indicator warning rules, the potential risks are graded according to the detection safety level to obtain potential risk alerts.

[0089] Specifically, the method first parses the experimental operation records for food safety POCT (Point-of-Care Testing, i.e., on-site rapid testing) detection steps to obtain rapid detection step identification. Natural language processing techniques, such as named entity recognition (NER) and dependency syntax analysis, are used to extract key information from the operation records and identify the current detection steps. Based on the obtained rapid detection step identification, the corresponding POCT device is triggered for data collection to obtain the original safety indicator data stream. The Internet of Things technology can be used to collect the output data of the POCT device in real time through a preset communication protocol.

[0090] The original safety index data stream is processed for interference elimination to obtain pure safety index data. This step uses signal processing technology, such as wavelet transform or Fourier transform, combined with a pre-established food matrix interference model to eliminate the influence of the sample matrix on the test results. The specific mathematical model can be expressed as:

[0091] S clean (t) = S raw (t)-M interference (t);

[0092] Among them, S clean (t) is a pure signal, S raw (t) is the original signal, M interference (t) is the matrix interference model.

[0093] Then, the pure safety index data is converted to standard units and the POCT equipment sensitivity is calibrated to obtain the standardized index value. The unit conversion uses the preset conversion factor, and the sensitivity calibration uses the linear regression model:

[0094] V standardized =a·V clean +b;

[0095] Among them, V standardized is the standardized value, V clean is a pure value, and a and b are calibration coefficients.

[0096] The standardized indicator values ​​are analyzed for correlation with indicator categories to obtain the indicator distribution characteristics. This step uses principal component analysis (PCA) or factor analysis to extract the characteristic distribution of different food categories. Based on the preset indicator threshold data set, the indicator distribution characteristics are processed for risk determination of exceeding the standard to obtain the indicator abnormality mark. The determination process can be expressed by the following formula:

[0097]

[0098] Among them, R risk is the risk ratio, V indicator is the index value, V limit is the limit standard value. risk When >1, it is marked as abnormal.

[0099] The abnormal indicator marks are traced and analyzed for abnormal indicator factors to obtain abnormal trend characteristics. This step uses graph database technology to build an indicator abnormal factor network and trace the source of the abnormality through the depth-first search (DFS) algorithm. The abnormal trend characteristics are evaluated and processed to obtain the potential risk level. Fuzzy logic or Bayesian network models are used here to comprehensively consider factors such as the degree of abnormality, scope of impact and duration to calculate the risk level. The standardized indicator values ​​and potential risk levels are associated with the detection safety level to obtain safety indicator data, and based on the potential risk level and the preset indicator warning rules, the potential risks are graded for detection safety levels to obtain potential risk alerts. This step uses a decision tree model to map the indicator values ​​and risk levels to specific detection safety levels and alarm levels.

[0100] For example, in a pesticide residue detection experiment, the operation record shows that the current step is "organophosphorus pesticide detection". This step is identified and the corresponding POCT device is triggered to start data collection. The raw data shows that the content of organophosphorus pesticides in the sample is 0.18 mg / kg. After wavelet transform to remove the interference of vegetable matrix, the pure data value is reduced to 0.15 mg / kg. Through the pre-calibrated linear model (a = 1.05, b = -0.01), the standardized value of 0.1475 mg / kg is obtained. This value is compared with the limit standard of 0.1 mg / kg for organophosphorus pesticides in vegetable foods, and the risk ratio is calculated to be 1.475, so it is marked as abnormal. Through supply chain tracing, it is found that the same batch of vegetables has had similar abnormalities many times in the past week. After evaluation using the fuzzy logic model, the potential risk level is obtained as "high". Finally, the system classifies the test result as "unqualified" and issues a "high risk" alarm to remind relevant personnel to take immediate measures. This method can not only accurately detect food safety indicators, but also detect potential risks in a timely manner, providing strong support for food safety management.

[0101] In a specific embodiment, the process of executing step S103 may specifically include the following steps:

[0102] (1) Decompose the experimental operation records into test procedures to obtain a fine-grained operation step sequence;

[0103] (2) Correlation and matching of safety indicator data and potential risk alerts is performed to obtain a step-level risk indicator mapping;

[0104] (3) Based on the mapping of fine-grained operation step sequences and step-level risk indicators, the time efficiency of the experimental process is evaluated and the time distribution of the steps is obtained;

[0105] (4) Comprehensively analyze and process the step time distribution and step-level risk indicator mapping to obtain an initial optimization plan;

[0106] (5) Conduct compliance verification on the initial optimization plan and the preset standard operating procedures to obtain compliance optimization suggestions;

[0107] (6) Based on the compliance optimization suggestions and the current laboratory resource status collected in real time, the feasibility of the initial optimization plan is evaluated and the optimization suggestions for the testing steps are obtained.

[0108] Specifically, the test process of the experimental operation records is decomposed to obtain a fine-grained sequence of operation steps. Natural language processing techniques, such as named entity recognition (NER) and dependency syntax analysis, are used to decompose the entire test process into a series of specific operation steps. Subsequently, the safety indicator data and potential risk alerts are associated and matched to obtain a step-level risk indicator mapping. This process uses association rule mining algorithms, such as the Apriori algorithm, to establish a mapping relationship between operation steps and risk indicators. Based on the fine-grained sequence of operation steps and the step-level risk indicator mapping, the experimental process is evaluated for time efficiency and the step time distribution is obtained. Time series analysis methods, such as exponential smoothing, are used here to calculate the average time and coefficient of variation of each step. The specific calculation formula is as follows:

[0109] T i =αX i +(1-α)T i-1 ;

[0110] Among them, T i is the smoothing time of the ith step, X i is the actual observation value, and α is the smoothing coefficient (0<α<1).

[0111] Then, the step time distribution and the step-level risk indicator mapping are comprehensively analyzed and processed to obtain the initial optimization solution. This step uses a multi-objective optimization algorithm, such as the Pareto optimization method, to consider both time efficiency and risk control. The optimization objective function can be expressed as:

[0112] F(x)=[f1(x),f2(x)];

[0113] Among them, f1(x) represents the time efficiency target, f2(x) represents the risk control target, and x is the adjustment parameter of the operation step.

[0114] Compliance verification is performed on the initial optimization plan and the preset standard operating procedures (SOP) to obtain compliance optimization suggestions. This step uses rule engine technology to match the optimization plan with SOP rules to ensure that the optimization suggestions meet standard requirements. The compliance score can be calculated using the following formula:

[0115]

[0116] Among them, C s is the compliance score, w i is the weight of the i-th rule, r i is the compliance degree of the i-th rule (0 or 1).

[0117] Finally, based on the compliance optimization suggestions and the current laboratory resource status collected in real time, the feasibility of the initial optimization plan is evaluated and the optimization suggestions for the detection steps are obtained. This step uses the constraint satisfaction problem (CSP) solving technology, taking laboratory resources as constraints to solve the optimal implementation plan.

[0118] For example, in a food heavy metal content detection experiment, the system first decomposes the experimental process into fine-grained steps such as "sample pretreatment", "digestion", and "instrument analysis". Through association rule mining, it is found that the "digestion" step is strongly correlated with high-risk alerts (support 0.8, confidence 0.9). Using exponential smoothing (α = 0.3) to analyze historical data, the average time of the "digestion" step is 120 minutes, and the coefficient of variation is 0.15. The multi-objective optimization algorithm considers time efficiency and risk control, and proposes an initial optimization plan to shorten the "digestion" time to 90 minutes and increase the intermediate checkpoint. The rule engine matches this plan with the SOP and calculates a compliance score of 0.95 (full score is 1). Finally, considering that the laboratory has newly purchased rapid digestion equipment, the CSP solution results suggest that the "digestion" time be further shortened to 60 minutes and a checkpoint be added at 30 minutes. This optimization suggestion not only improves experimental efficiency, but also reduces potential risks by increasing the frequency of inspections, while ensuring compliance with standard operating procedures and practical feasibility.

[0119] In a specific embodiment, the process of executing step S104 may specifically include the following steps:

[0120] (1) Extract and sort the timestamps of the experimental operation records to obtain the time sequence of the experimental steps;

[0121] (2) Perform time correlation processing on the safety indicator data to obtain an indicator change trend chart;

[0122] Map the potential risk alerts to the time nodes to obtain the risk event timeline;

[0123] (3) Perform time correlation analysis on the optimization suggestions for the detection steps to obtain the implementation time points of the optimization suggestions;

[0124] (4) Based on the experimental step time series, indicator change trend chart, risk event timeline and optimization suggestion implementation time point, multi-dimensional data integration processing is performed through time series data fusion algorithm to obtain the time series data of the entire detection process;

[0125] (5) Extract key nodes and analyze causal relationships of the time series data of the entire testing process to obtain a list of critical control points;

[0126] (6) Based on the critical control point list and preset rating standards, the entire experimental process is comprehensively evaluated and processed to obtain a traceable experimental report and test safety level assessment results.

[0127] Specifically, the experimental operation records are timestamped and sorted to obtain the experimental step time series. This process uses regular expressions to extract the time information in each operation record, and then uses the quick sorting algorithm to arrange them in chronological order to form an ordered experimental step time series. Next, the safety indicator data is time-correlated to obtain the indicator change trend chart. Here, an interpolation algorithm, such as cubic spline interpolation, is used to make discrete indicator data points continuous, and then the moving average method is used to smooth the data to finally generate a continuous trend curve. Time node mapping is performed on potential risk alerts to obtain the risk event timeline. This step maps the occurrence time of each risk alert to a unified timeline, and uses color or size to indicate the risk level. For the detection step optimization suggestions, time correlation analysis is performed to obtain the optimization suggestion implementation time point. Here, a time series correlation analysis method, such as the cross-correlation function (CCF), is used to determine the best implementation time for each optimization suggestion. Based on the above-generated experimental step time series, indicator change trend chart, risk event timeline and optimization suggestion implementation time point, multi-dimensional data integration is performed through the time series data fusion algorithm to obtain the detection process time series data. This process uses the dynamic time warping (DTW) algorithm to align data of different time scales, and then uses the principal component analysis (PCA) method to reduce the dimensionality, ultimately forming a unified time series dataset.

[0128] The key nodes are extracted and the causal relationship analysis is performed on the time series data of the entire detection process to obtain a list of critical control points. The key node extraction uses a change point detection algorithm, such as the CUSUM (cumulative sum control chart) method, to identify the time points when the data changes significantly. The causal relationship analysis uses the Granger causality test to determine the causal relationship between different events. Based on the critical control point list and the preset rating standards, the entire experimental process is comprehensively evaluated to obtain a traceable experimental report and the detection safety level assessment results. This step uses a multi-criteria decision analysis (MCDA) method, such as the analytic hierarchy process (AHP), to comprehensively consider various indicators and risk factors to obtain the final safety level assessment.

[0129] For example, in a pesticide residue detection experiment, the system first extracts timestamps from the operation records, such as "2024-07-30 09:00:00 sample pretreatment started", "2024-07-30 10:15:30 sample digestion completed", etc., and arranges them in chronological order. For pesticide residue concentration data, the system uses cubic spline interpolation to generate a continuous trend curve, showing that the concentration gradually decreases from the initial 0.18 mg / kg to 0.15 mg / kg. Risk alerts are marked on the timeline, such as the "high risk" alert that appears at 10:20. The optimization suggestion "increase the number of sample cleaning times" is determined by CCF analysis to be best implemented at 09:30. The DTW algorithm is used to time-align these data, and then PCA is used to reduce the dimension to obtain a comprehensive time series data set. The CUSUM algorithm detects three key nodes in this data set: 09:30 (sample processing completed), 10:20 (high risk alert appears), and 11:00 (test results are stable). The Granger causality test found that there was a significant causal relationship between sample processing methods and subsequent risk alerts (p value < 0.05).

[0130] Finally, the AHP method was used to evaluate the entire process. Four evaluation criteria were set: operational standardization (weight 0.3), data reliability (weight 0.3), risk control (weight 0.2) and efficiency (weight 0.2). According to the performance of the key control points, the scores were given as follows: 8, 7, 6, 9 (out of 10 points). The calculated comprehensive score was 7.5, corresponding to the "B" detection safety level, indicating that there is a certain risk but it is within the controllable range. This method not only provides a highly traceable experimental report, but also deeply reveals the potential problems and improvement opportunities in the experimental process through multidimensional data analysis.

[0131] In a specific embodiment, the process of extracting key nodes and analyzing causal relationships of the time series data of the entire detection process to obtain a critical control point list may specifically include the following steps:

[0132] (1) Perform voiceprint recognition event marking processing on the time series data of the entire detection process to obtain the operator's voice command time point set;

[0133] (2) Based on the preset experimental process verification nodes, the operator's voice command time point set is classified and processed to obtain a sequential node set and a parallel node set;

[0134] (3) monitoring and recording each node in the sequential node set in a predetermined order, and generating an alarm message if a sequence error is detected;

[0135] (4) Performing sequence-variable recording processing on the nodes in the parallel node set, recording the actual execution order of the nodes, but not triggering an alarm for sequence changes;

[0136] (5) Based on the sequential node set and the parallel node set, the safety index data is segmented to obtain the key interval of index change;

[0137] (6) Compare the key intervals of indicator changes with the preset standard indicator intervals to obtain a list of potential hazard points;

[0138] (7) Perform risk assessment on the list of potential hazard points to obtain a candidate set of critical control points;

[0139] (8) Perform time correspondence processing on the candidate set of critical control points and the experimental operation records to obtain the operation-related control point mapping;

[0140] (9) Based on the operation-related control point mapping and detection step optimization suggestions, the control points are prioritized to obtain a list of critical control points.

[0141] Specifically, the time series data of the whole detection process is processed by voiceprint recognition event marking to obtain the operator's voice command time point set. In this step, the Mel Frequency Cepstral Coefficient (MFCC) is used to extract voice features, and then the Gaussian Mixture Model (GMM) is used for voiceprint matching to identify the operator's voice command and mark its time point. Next, based on the preset experimental process verification node, the operator's voice command time point set is classified and processed to obtain a sequential node set and a parallel node set. This process is matched with the preset experimental process template to divide the voice command time points into sequential nodes that must be executed in sequence and parallel nodes whose order can be flexibly adjusted. Each node in the sequential node set is monitored and recorded in a predetermined order, and an alarm message is generated if a sequence error is detected. Here, a finite state machine (FSM) model is used to track the execution order of the experimental steps. Once an operation that does not conform to the preset order is detected, the alarm mechanism is immediately triggered. At the same time, the nodes in the parallel node set are processed for variable order recording, and the actual execution order of the nodes is recorded, but no alarm is triggered for sequence changes. This step uses a dynamic programming algorithm to record the actual execution order of parallel nodes for subsequent analysis.

[0142] Based on the sequential node set and the parallel node set, the safety indicator data is segmented to obtain the key interval of indicator change. Here, the segmented linear regression algorithm is used to segmentally fit the indicator data with the node time point as the boundary to identify the interval with significant indicator change. Next, the key interval of indicator change is compared with the preset standard indicator interval to obtain a list of potential hazard points. This step uses the Z-score standardization method to calculate the degree of deviation of each data point. When the Z-score exceeds the preset threshold, the point is marked as a potential hazard point. The list of potential hazard points is risk assessed to obtain a candidate set of key control points. Here, the fuzzy comprehensive evaluation method is used to comprehensively consider factors such as the severity, probability of occurrence and controllability of the hazard, score each potential hazard point, and select the point with the highest score as a candidate for the key control point. Next, the candidate set of key control points and the experimental operation record are processed in time correspondence to obtain the operation-related control point mapping. This step uses the nearest neighbor algorithm to associate each key control point with the operation record closest in time.

[0143] Based on the operation-related control point mapping and detection step optimization suggestions, the control points are prioritized to obtain a list of key control points. A weighted scoring method is used here to comprehensively consider factors such as risk level, operation criticality and optimization potential, calculate a comprehensive score for each control point, and sort them by score.

[0144] For example: In a pesticide residue detection experiment, 10 key voice command time points were identified through MFCC and GMM analysis. After matching these time points with the preset experimental process, 7 sequential nodes (such as "sample pretreatment", "sample digestion", "instrument analysis", etc.) and 3 parallel nodes (such as "data entry", "report generation", and "equipment cleaning") were determined. The FSM model is used to monitor the execution of sequential nodes, and an alarm is immediately generated when it is detected that "instrument analysis" is executed before "sample digestion". For parallel nodes, the actual execution order is recorded as "equipment cleaning", "data entry", and "report generation", but no alarm is triggered.

[0145] The segmented linear regression analysis found that in the "sample digestion" stage, the slope of the pesticide residue concentration data changed from -0.02mg / kg / min to -0.05mg / kg / min, which was marked as a critical change interval. Z-score analysis showed that there was a data point in this interval with a Z-score of 2.8, which exceeded the preset threshold of 2.5 and was listed as a potential hazard point. The fuzzy comprehensive evaluation gave this point 85 points (out of 100) and included it in the candidate set of critical control points. The nearest neighbor algorithm associated this point with the operation record "adjusting the digestion temperature". Finally, the weighted scoring method considered the risk level of this point (weight 0.4, score 90), operational criticality (weight 0.3, score 85) and optimization potential (weight 0.3, score 80), and calculated a comprehensive score of 85.5, ranking first among all control points, so it was listed as the most important food safety critical control point.

[0146] In a specific embodiment, the experimental operation record may be any one of the following operation records: daily medical test operation record, clinical bedside test operation record, POCT test operation record, food safety test operation record and surgical operation record.

[0147] The above describes the experimental process data recording method based on biometrics in the embodiment of the present application. The following describes the experimental process data recording system based on biometrics in the embodiment of the present application. Figure 3 In the embodiment of the present application, an embodiment of the experimental process data recording system based on biometrics includes:

[0148] The analysis module 301 is used to analyze the operation records of the real-time voice instructions of the detection operator to obtain the experimental operation records;

[0149] The detection module 302 is used to collect the detection instrument data in real time and perform abnormal detection processing based on the experimental operation record to obtain safety indicator data and potential risk alarms;

[0150] The optimization module 303 is used to perform step optimization analysis on the experimental operation record, the safety indicator data and the potential risk alarm to obtain detection step optimization suggestions;

[0151] The integration module 304 is used to perform time-series integration processing on the experimental operation records, the safety indicator data, the potential risk alarms and the detection step optimization suggestions to obtain a traceable experimental report and a detection safety level assessment result.

[0152] Through the coordinated cooperation of the above-mentioned components and the introduction of biometric recognition technology, the accurate verification of the identity of the detection operator is achieved, which greatly improves the safety and reliability of experimental operations and effectively prevents the risk of unauthorized personnel performing operations. By using voiceprint recognition and natural language processing technology, the operator's voice instructions can be captured and understood in real time and converted into structured experimental operation records, which not only reduces the errors of manual recording, but also improves the efficiency and accuracy of data recording; by collecting and analyzing the data of the detection instrument in real time, potential abnormal situations can be quickly identified and risk alarms can be issued in time, which greatly improves the safety and controllability of the experimental process; based on the collected data Real-time step optimization analysis can be carried out to provide optimization suggestions for operators, which can improve experimental efficiency and reduce resource waste. Through the time-serialized integration of the data of the entire experimental process, highly traceable experimental reports and detection safety level assessment results are generated, which not only meets regulatory requirements, but also provides strong support for subsequent quality improvement and risk management. While ensuring data integrity and reliability, the present invention also greatly reduces manual intervention and reduces the possibility of human errors. It can also promptly detect and respond to abnormal situations in the experimental process, greatly improving the efficiency and reliability of food safety testing, and significantly improving the reliability and traceability of experimental data and the safety of the overall experimental process.

[0153] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0154] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for recording experimental process data based on biometrics, characterized in that: The experimental process data recording method based on biometrics includes: The real-time voice commands of the detection operator are analyzed and recorded to obtain the experimental operation records; Based on the experimental operation records, real-time collection and abnormal detection processing of detection instrument data are performed to obtain safety indicator data and potential risk alarms; Performing step optimization analysis on the experimental operation records, the safety indicator data and the potential risk alarm to obtain detection step optimization suggestions; The experimental operation records, the safety indicator data, the potential risk alarms and the detection step optimization suggestions are integrated and processed in a time series manner to obtain a traceable experimental report and detection safety level assessment results, including: Extracting and sorting the timestamps of the experimental operation records to obtain a time sequence of experimental steps; Performing time correlation processing on the safety indicator data to obtain an indicator change trend graph; Performing time node mapping processing on the potential risk alert to obtain a risk event timeline; Performing time correlation analysis on the optimization suggestions for the detection steps to obtain a time point for implementing the optimization suggestions; Based on the experimental step time series, the indicator change trend chart, the risk event timeline and the optimization suggestion implementation time point, multi-dimensional data integration processing is performed through a time series data fusion algorithm to obtain the time series data of the entire detection process; Extract key nodes and perform causal relationship analysis on the time series data of the entire detection process to obtain a list of key control points; Based on the critical control point list and the preset rating standards, a comprehensive evaluation is performed on the entire experimental process to obtain the traceable experimental report and the detection safety level assessment result; The key nodes are extracted and the causal relationship analysis is performed on the time series data of the whole detection process to obtain a list of key control points, including: Performing voiceprint recognition event marking processing on the time series data of the entire detection process to obtain a set of operator voice command time points; Based on the preset experimental process verification nodes, the operator voice instruction time point set is classified and processed to obtain a sequential node set and a parallel node set; Monitor and record each node in the sequential node set in a predetermined order, and generate an alarm message if a sequence error is detected; Performing sequence-variable recording processing on the nodes in the parallel node set, recording the actual execution sequence of the nodes, but not triggering an alarm for sequence changes; Based on the sequential node set and the parallel node set, the security indicator data is processed in sections to obtain a critical interval of indicator change; Compare the critical interval of the indicator change with the preset standard indicator interval to obtain a list of potential hazard points; Performing risk assessment on the potential hazard point list to obtain a candidate set of critical control points; Performing time correspondence processing on the candidate set of key control points and the experimental operation records to obtain a control point mapping related to the operation; Based on the control point mapping related to the operation and the detection step optimization suggestions, the control points are prioritized to obtain the key control point list.

2. The method for recording experimental process data based on biometrics according to claim 1, characterized in that: The operation record analysis of the real-time voice command of the detection operator to obtain the experimental operation record includes: Based on the preset biometric database, the biometrics of the test operator are collected and the identity verification is processed to obtain the operator identity confirmation result and the experiment authorization mark; Based on the operator identity confirmation result and the experiment authorization mark, voiceprint recognition and natural language processing are performed on the real-time voice instructions of the detection operator to obtain an experimental operation record that matches the preset experimental step library.

3. The method for recording experimental process data based on biometrics according to claim 2, characterized in that: Based on the preset biometric database, the biometrics of the detection operator are collected and the identity verification is processed to obtain the operator identity confirmation result and the experiment authorization mark, including: The experimental command voice of the detection operator is collected and processed to obtain original voiceprint data related to the experiment, and the laboratory background noise is eliminated from the original voiceprint data related to the experiment to obtain target detection voiceprint data; Performing experimental operation-related feature extraction processing on the target detection voiceprint data to obtain a detection voiceprint feature vector, and performing laboratory environment adaptability normalization processing on the detection voiceprint feature vector to obtain a standardized detection voiceprint feature; Performing similarity calculation on the standardized detection voiceprint feature and the pre-stored detector voiceprint template in the biometric database to obtain a detection voiceprint matching score; The facial image of the detection operator wearing the protective equipment is collected and processed to obtain an original facial image in the detection protection state; Performing feature extraction processing on the original face image in the detected protection state to obtain a face feature vector in the protection state; Performing similarity calculation on the protection state face feature vector and the protection state face template of the inspector stored in the biometric database to obtain a detection protection state face matching score; Performing weighted fusion processing on the detected voiceprint matching score and the detected protection status face matching score based on the detection security level to obtain a detection comprehensive identity matching score; Based on a preset detection authority threshold, the detection comprehensive identity matching score is subjected to a hierarchical comparison process to obtain the operator identity confirmation result and the experiment authorization mark.

4. The method for recording experimental process data based on biometrics according to claim 2, characterized in that: Based on the operator identity confirmation result and the experiment authorization mark, voiceprint recognition and natural language processing are performed on the real-time voice command of the detection operator to obtain an experimental operation record matching a preset experimental step library, including: The real-time voice instructions of the detection operator are collected and processed to obtain original voice data, and the original voice data is subjected to laboratory environmental noise filtering to obtain purified voice data; Performing frame division and feature extraction processing on the purified voice data to obtain a voice feature vector, and performing matching processing on the voice feature vector and a voiceprint model associated with the operator identity confirmation result to obtain a voiceprint consistency score; Based on the voiceprint consistency score and the experiment authorization mark, decoding the purified voice data to obtain a voice command text; Performing safety terminology recognition processing on the voice instruction text to obtain key experimental step phrases; Performing context association processing on the key experimental step phrases to obtain structured experimental step description data; Performing matching degree calculation processing on the structured experimental step description data and the preset experimental step library to obtain step matching probability distribution data; Based on the step matching probability distribution data, the structured experimental step description data is subjected to the best matching selection process to obtain the current experimental operation steps; The current experimental operation steps, the operator identity confirmation result and the current timestamp are associated with each other to obtain the experimental operation record.

5. The method for recording experimental process data based on biometrics according to claim 1, characterized in that: Based on the experimental operation records, real-time collection and abnormal detection processing of detection instrument data are performed to obtain safety indicator data and potential risk alarms, including: Performing on-site rapid detection POCT detection step parsing processing on the experimental operation record to obtain a rapid detection step identifier, and based on the rapid detection step identifier, performing data collection trigger processing on the corresponding on-site rapid detection POCT device to obtain an original safety indicator data stream; Performing interference elimination processing on the original safety indicator data stream to obtain pure safety indicator data; The pure safety index data is converted into standard units and the sensitivity of the on-site rapid detection POCT equipment is calibrated to obtain a standardized index value; Performing indicator category correlation analysis on the standardized indicator values ​​to obtain indicator distribution characteristics; Based on a preset indicator threshold data set, the indicator distribution characteristics are subjected to over-limit risk determination processing to obtain an indicator abnormality mark; Performing a retrospective analysis of the abnormal factors of the indicators on the abnormal indicator marks to obtain abnormal trend characteristics; Evaluate and process the abnormal trend characteristics to obtain a potential risk level; The standardized indicator value and the potential risk level are subjected to detection safety level association processing to obtain the safety indicator data, and based on the potential risk level and preset indicator warning rules, the potential risk is subjected to detection safety level classification processing to obtain the potential risk alarm.

6. The method for recording experimental process data based on biometrics according to claim 5, characterized in that: The step optimization analysis of the experimental operation record, the safety indicator data and the potential risk alarm to obtain the detection step optimization suggestions includes: Decomposing the experimental operation record into a detection process to obtain a fine-grained operation step sequence; Performing correlation and matching processing on the safety indicator data and the potential risk alert to obtain a step-level risk indicator mapping; Based on the fine-grained operation step sequence and the risk indicator mapping of the step level, the time efficiency of the experimental process is evaluated to obtain the time distribution of the steps; Comprehensively analyzing and processing the time distribution of the steps and the risk indicator mapping of the step levels to obtain an initial optimization plan; Perform compliance verification on the initial optimization plan and the preset standard operating procedures to obtain compliance optimization suggestions; Based on the compliance optimization suggestions and the current laboratory resource status collected in real time, the initial optimization plan is evaluated for feasibility to obtain the detection step optimization suggestions.

7. The method for recording experimental process data based on biometrics according to claim 1, characterized in that: The experimental operation record is any one of the following operation records: daily medical test operation record, on-site rapid test POCT test operation record, food safety test operation record and surgical operation record.

8. An experimental process data recording system based on biometrics, characterized in that: Used to perform the method for recording experimental process data based on biometric identification as described in any one of claims 1 to 7, the experimental process data recording system based on biometric identification comprises: An analysis module is used to analyze the operation records of the real-time voice instructions of the detection operator to obtain the experimental operation records; A detection module is used to collect and detect abnormalities of the detection instrument data in real time based on the experimental operation records to obtain safety indicator data and potential risk alarms; An optimization module, used to perform step optimization analysis on the experimental operation records, the safety indicator data and the potential risk alarm to obtain optimization suggestions for the detection steps; The integration module is used to perform time-series integration processing on the experimental operation records, the safety indicator data, the potential risk alarms and the detection step optimization suggestions to obtain a traceable experimental report and a detection safety level assessment result.

Images