Gas recognition method, device, equipment, storage medium and program product

Abstract

The present application relates to a gas recognition method, including: obtaining multiple groups of gas data collected for a target object through multiple gas sensors, determining the characteristics of each group of gas data, and obtaining multiple gas characteristics; classifying the multiple gas characteristics into important characteristics and secondary characteristics; for each important characteristic, performing a pooling operation on the important characteristic to obtain an important pooled characteristic, and encoding the important pooled characteristic through an important encoder to obtain an important encoded characteristic corresponding to the important characteristic; for each secondary characteristic, performing a pooling operation on the secondary characteristic to obtain a secondary pooled characteristic, and encoding the secondary pooled characteristic through a secondary encoder to obtain a secondary encoded characteristic corresponding to the secondary characteristic; identifying the gas category to which the exhaled gas of the target object belongs according to the important encoded characteristics corresponding to the respective important characteristics and the secondary encoded characteristics corresponding to the respective secondary characteristics. The use of this method can improve the diagnostic accuracy of lung diseases.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular, to a gas recognition method, device, equipment, storage medium, and program product. Background Art

[0002] Lung diseases, such as lung cancer, pneumonia, and COPD (Chronic Obstructive Pulmonary Disease), are the main causes of death and disability worldwide. Early detection and accurate diagnosis are crucial for improving the treatment effect and patient survival rate. Traditional diagnostic methods rely on manual film reading, which has problems such as strong subjectivity, low efficiency, and high misdiagnosis rate. Summary of the Invention

[0003] Based on this, it is necessary to provide a gas recognition method, device, equipment, storage medium, and program product that can improve the diagnostic accuracy of lung diseases for the above technical problems.

[0004] In a first aspect, the present application provides a gas recognition method, the method comprising:

[0005] Obtaining multiple sets of gas data collected for a target object through multiple gas sensors, determining the characteristics of each set of gas data, and obtaining multiple gas characteristics;

[0006] Classifying the multiple gas characteristics into important characteristics and secondary characteristics;

[0007] For each important characteristic, performing a pooling operation on the important characteristic to obtain an important pooling feature of the important characteristic, and encoding the important pooling feature through an important encoder to obtain an important encoding feature corresponding to the important characteristic;

[0008] For each secondary characteristic, performing a pooling operation on the secondary characteristic to obtain a secondary pooling feature of the secondary characteristic, and encoding the secondary pooling feature through a secondary encoder to obtain a secondary encoding feature corresponding to the secondary characteristic;

[0009] Identifying the gas category to which the exhaled gas of the target object belongs according to the important encoding features corresponding to the respective important characteristics and the secondary encoding features corresponding to the respective secondary characteristics.

[0010] In a second aspect, the present application provides a gas recognition device, the device comprising:

[0011] An acquisition module, configured to obtain multiple sets of gas data collected for a target object through multiple gas sensors, determine the characteristics of each set of gas data, and obtain multiple gas characteristics;

[0012] A classification module for classifying the multiple gas features into important features and secondary features;

[0013] A feature processing module for, for each important feature, performing a pooling operation on the targeted important feature to obtain an important pooled feature of the targeted important feature, and encoding the important pooled feature through an important encoder to obtain an important encoded feature corresponding to the targeted important feature; for each secondary feature, performing a pooling operation on the targeted secondary feature to obtain a secondary pooled feature of the targeted secondary feature, and encoding the secondary pooled feature through a secondary encoder to obtain a secondary encoded feature corresponding to the targeted secondary feature;

[0014] An identification module for identifying the gas category to which the exhaled gas of the target object belongs according to the important encoded features respectively corresponding to the respective important features and the secondary encoded features respectively corresponding to the respective secondary features.

[0015] In a third aspect, the present application provides a computer device, including a memory and a processor, where a computer program is stored in the memory, and when the processor executes the computer program, the steps in the method embodiments of the present application are implemented.

[0016] In a fourth aspect, the present application provides a computer-readable storage medium storing a computer program, and when the computer program is executed by a processor, the steps in the method embodiments of the present application are implemented.

[0017] In a fifth aspect, the present application provides a computer program product including a computer program, and when the computer program is executed by a processor, the steps in the method embodiments of the present application are implemented.

[0018] The above gas recognition method, device, equipment, storage medium and program product obtain multiple sets of gas data collected for a target object through multiple gas sensors, determine the characteristics of each set of gas data, and obtain multiple gas characteristics; classify the multiple gas characteristics into important characteristics and secondary characteristics; for each important characteristic, perform a pooling operation on the important characteristic to obtain an important pooling characteristic of the important characteristic, and encode the important pooling characteristic through an important encoder to obtain an important encoded characteristic corresponding to the important characteristic; for each secondary characteristic, perform a pooling operation on the secondary characteristic to obtain a secondary pooling characteristic of the secondary characteristic, and encode the secondary pooling characteristic through a secondary encoder to obtain a secondary encoded characteristic corresponding to the secondary characteristic; identify the gas category to which the exhaled gas of the target object belongs according to the important encoded characteristics corresponding to the respective important characteristics and the secondary encoded characteristics corresponding to the respective secondary characteristics. In this way, by classifying gas characteristics into important characteristics and secondary characteristics, and combining the pooling operation and the dual-encoder architecture, the present application differentiates the processing of important characteristics and secondary characteristics, which helps to pay more attention to and process important characteristics more carefully during gas recognition, thereby improving the accuracy of gas recognition, and can assist in the diagnosis process of lung diseases and improve the accuracy of lung disease diagnosis. Description of the Drawings

[0019] Figure 1 It is a schematic flowchart of the gas recognition method in one embodiment;

[0020] Figure 2 It is a schematic structural diagram of the gas recognition model in one embodiment;

[0021] Figure 3 It is a schematic structural diagram of the pyramid pooling network in one embodiment;

[0022] Figure 4 It is a block diagram of the structure of the gas recognition device in one embodiment;

[0023] Figure 5 It is an internal structural diagram of a computer device in one embodiment;

[0024] Figure 6 It is an internal structural diagram of a computer device in another embodiment. Detailed Embodiments

[0025] In order to make the objectives, technical solutions and advantages of the present application clearer and more understandable, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0026] In one embodiment, as Figure 1As shown, a gas recognition method is provided. This method can be applied to a computer device, which can be a terminal or a server. That is, this method can be executed independently by the terminal or the server, or can be implemented through the interaction between the terminal and the server. This embodiment takes the application of this method to a computer device as an example for illustration, including the following steps:

[0027] Step 102, obtain multiple groups of gas data collected for a target object by multiple gas sensors, determine the characteristics of each group of gas data, and obtain multiple gas characteristics.

[0028] Among them, a gas sensor is a sensor used to collect gas data of the gas exhaled by an object. For example, the gas data of the gas exhaled by a patient can be collected through a gas sensor.

[0029] Specifically, the gas sensor can collect the gas data of the gas exhaled by the target object and send the collected gas data to the computer device. The computer device can obtain multiple groups of gas data collected for the target object by multiple gas sensors and determine the characteristics of each group of gas data to obtain multiple gas characteristics.

[0030] In one embodiment, determining the characteristics of each group of gas data and obtaining multiple gas characteristics includes: performing discrete wavelet transform on each group of gas data respectively to determine the characteristics of each group of gas data and obtain multiple gas characteristics. Specifically, the gas data is a discrete signal of a time series, and the process of discrete wavelet transform can be represented by the following formula:

[0031] X j,k =∑ n x[n]·ψ j,k [n];

[0032] A j,k =∑ n x[n]·φ j,k [n];

[0033] Among them, X j,k represents the detail coefficient, which is used to capture the high-frequency characteristics of the signal, and A j,k represents the approximation coefficient, which is used to capture the low-frequency characteristics of the signal. x[n] is a discrete signal of a time series, and ψ j,k [n] is the discrete wavelet function at scale j and translation k, which is used to extract the detail information of the signal. φ j,k[n] is the discrete scaling function at scale j and translation k, which is used to extract the approximate information of the signal. j represents the number of decomposition levels or scales, determining the frequency range of analysis, and k represents the position of the wavelet function on the time axis. In this way, through the discrete wavelet transform, the characteristics of the time series can be analyzed and understood from multiple scales, so as to achieve feature selection, which can significantly reduce the number of features while retaining the key features of the original signal, thereby reducing the resource consumption of data processing.

[0034] For example, the computer device can perform the fourth-level decomposition based on wavelet on the signals of the gas data collected by 8 gas sensors. The gas data collected by each original gas sensor contains 4000 data points. Through the discrete wavelet transform, the gas data of each gas sensor can be effectively compressed into 4 new feature points, which mainly represent the low-frequency components, while the high-frequency components are discarded.

[0035] Step 104, classify multiple gas features into important features and secondary features.

[0036] In one embodiment, classifying multiple gas features into important features and secondary features includes: determining the importance scores of the multiple gas features respectively; for each gas feature, the importance score of the targeted gas feature is positively correlated with the contribution degree of the targeted gas feature to the gas recognition result; classifying the multiple gas features into important features and secondary features according to the importance scores of the multiple gas features respectively; the importance score of the important feature is greater than the importance score of the secondary feature. It can be understood that the contribution degree of the important feature to the gas recognition result is greater than the contribution degree of the secondary feature to the gas recognition result.

[0037] Step 106, for each important feature, perform a pooling operation on the targeted important feature to obtain the important pooling feature of the targeted important feature, and encode the important pooling feature through the important encoder to obtain the important encoded feature corresponding to the targeted important feature.

[0038] In one embodiment, for each important feature, the computer device can input the targeted important feature into the pooling network to perform a pooling operation on the targeted important feature through the pooling network to obtain the important pooling feature of the targeted important feature. Furthermore, the computer device can input the important pooling feature of the targeted important feature into the important encoder to encode the important pooling feature of the targeted important feature through the important encoder to obtain the important encoded feature corresponding to the targeted important feature.

[0039] Step 108, for each secondary feature, perform a pooling operation on the targeted secondary feature to obtain the secondary pooling feature of the targeted secondary feature, and encode the secondary pooling feature through the secondary encoder to obtain the secondary encoded feature corresponding to the targeted secondary feature.

[0040] It can be understood that the important encoder is an encoder for processing important pooling features, and the secondary encoder is an encoder for processing secondary pooling features. The important encoder and the secondary encoder have different configurations. The important encoder has a higher dimension than the secondary encoder, and the ability of the important encoder with a higher dimension to refine features is stronger than that of the secondary encoder with a lower dimension. It can be understood that in the process of gas recognition, giving more attention and more detailed processing to important features can improve the gas recognition accuracy on the premise of ensuring the recognition efficiency.

[0041] In one embodiment, for each secondary feature, the computer device may input the secondary feature targeted into a pooling network to perform a pooling operation on the secondary feature targeted through the pooling network to obtain the secondary pooling feature of the secondary feature targeted. Further, the computer device may input the secondary pooling feature of the secondary feature targeted into the secondary encoder to encode the secondary pooling feature of the secondary feature targeted through the secondary encoder to obtain the secondary coding feature corresponding to the secondary feature targeted.

[0042] Step 110, identify the gas category to which the exhaled gas of the target object belongs according to the important coding features respectively corresponding to the respective important features and the secondary coding features respectively corresponding to the respective secondary features.

[0043] Among them, the gas category is the category to which the gas to be recognized belongs. For example, the gas category may include but is not limited to smokers, COPD (Chronic Obstructive Pulmonary Disease), healthy people, and air.

[0044] In one embodiment, the computer device may fuse the important coding features respectively corresponding to the respective important features and the secondary coding features respectively corresponding to the respective secondary features to obtain a fused feature, and identify the gas category to which the exhaled gas of the target object belongs according to the fused feature.

[0045] In one embodiment, the computer device may determine the fusion weights of the important coding features respectively corresponding to the respective important features and determine the fusion weights of the secondary coding features respectively corresponding to the respective secondary features. Among them, the fusion weight of the important coding feature is greater than the fusion weight of the secondary coding feature. Further, the computer device may fuse the respective important codings and the secondary coding features based on the fusion weights of the respective important coding features and the fusion weights of the respective secondary coding features to obtain a fused feature, and identify the gas category to which the exhaled gas of the target object belongs according to the fused feature. In the process of gas recognition, by assigning a higher fusion weight to the important coding feature compared to the secondary coding feature to further give more attention to the important feature, the gas recognition accuracy can be improved.

[0046] In the above gas recognition method, by acquiring multiple sets of gas data collected for a target object through multiple gas sensors, determining the characteristics of each set of gas data, and obtaining multiple gas characteristics; classifying the multiple gas characteristics into important characteristics and secondary characteristics; for each important characteristic, performing a pooling operation on the important characteristic to obtain an important pooling characteristic of the important characteristic, and encoding the important pooling characteristic through an important encoder to obtain an important encoding characteristic corresponding to the important characteristic; for each secondary characteristic, performing a pooling operation on the secondary characteristic to obtain a secondary pooling characteristic of the secondary characteristic, and encoding the secondary pooling characteristic through a secondary encoder to obtain a secondary encoding characteristic corresponding to the secondary characteristic; identifying the gas category to which the exhaled gas of the target object belongs according to the important encoding characteristics corresponding to the respective important characteristics and the secondary encoding characteristics corresponding to the respective secondary characteristics. In this way, by classifying gas characteristics into important characteristics and secondary characteristics, and combining the pooling operation and the dual-encoder architecture, the important characteristics and secondary characteristics are processed in a differentiated manner, which helps to pay more attention to and process important characteristics more carefully during the gas recognition process, thereby improving the accuracy of gas recognition, and can thus assist in the diagnosis process of lung diseases and improve the accuracy of lung disease diagnosis.

[0047] In one embodiment, classifying the multiple gas characteristics into important characteristics and secondary characteristics includes: determining the absolute value of the SHAP value of each of the multiple gas characteristics; for each gas characteristic, the absolute value of the SHAP value of the gas characteristic is positively correlated with the degree of contribution of the gas characteristic to the gas recognition result; classifying the multiple gas characteristics into important characteristics and secondary characteristics according to the absolute value of the SHAP value of each of the multiple gas characteristics; the absolute value of the SHAP value of the important characteristic is greater than the absolute value of the SHAP value of the secondary characteristic.

[0048] Among them, SHAP (Shapley Additive Explanations) is a model interpretation technique, and the SHAP analysis technique can be used to evaluate the degree of contribution of gas characteristics to the gas recognition result.

[0049] Specifically, the computer device can calculate the absolute value of the SHAP value of each of multiple gas characteristics. Among them, for each gas characteristic, the absolute value of the SHAP value of the targeted gas characteristic is positively correlated with the degree of contribution of the targeted gas characteristic to the gas recognition result. It can be understood that the larger the absolute value of the SHAP value of a gas characteristic, the greater the degree of contribution of the gas characteristic to the gas recognition result. Conversely, the smaller the absolute value of the SHAP value of a gas characteristic, the smaller the degree of contribution of the gas characteristic to the gas recognition result. The computer device can classify multiple gas characteristics into important characteristics and secondary characteristics according to the absolute values of the SHAP values of each of the multiple gas characteristics. Among them, the absolute value of the SHAP value of the important characteristic is greater than the absolute value of the SHAP value of the secondary characteristic.

[0050] In one embodiment, the number of gas sensors is 8, so the number of gas characteristics is also 8. The computer device can calculate the absolute value of the SHAP value of each of these 8 gas characteristics, sort these 8 gas characteristics in descending order according to the absolute value of the SHAP value, and determine the first 4 gas characteristics with larger absolute values of the SHAP value as important characteristics, and determine the last 4 gas characteristics with smaller absolute values of the SHAP value as secondary characteristics.

[0051] In the above embodiment, the SHAP analysis technology is used to evaluate the degree of contribution of each gas characteristic to the gas recognition result, so as to classify each gas characteristic into important characteristics and secondary characteristics, so that more attention and more refined processing can be given to the important characteristics in the subsequent gas recognition process, and the accuracy of gas recognition can be improved.

[0052] In one embodiment, determining the absolute value of the SHAP value of each of multiple gas characteristics includes:

[0053]

[0054] Among them, S represents the set composed of all feature combinations of multiple gas characteristics, f represents the recognition function, f(s) represents the recognition result when S exists, and f(s∪{i}) is the recognition result after the gas characteristic i is added to S; represents the marginal contribution value of the gas characteristic i in S; N represents the set composed of multiple gas characteristics, |S| represents the size of S, and |N| represents the total number of features; φ i represents the SHAP value of the gas characteristic i.

[0055] In the above embodiment, by providing the specific calculation method of the SHAP value, each gas characteristic can be more accurately classified into important characteristics and secondary characteristics, improving the classification accuracy of gas characteristics, and further improving the accuracy of gas recognition.

[0056] In one embodiment, for each important feature, a pooling operation is performed on the targeted important feature to obtain an important pooled feature of the targeted important feature, including: for each important feature, performing pooling operations at multiple scales on the targeted important feature to obtain important scale features corresponding to each of the multiple scales; and performing feature fusion on the important scale features corresponding to each of the multiple scales to obtain an important pooled feature of the targeted important feature.

[0057] In one embodiment, for each important feature, the computer device can input the targeted important feature into a pyramid pooling network including multiple pooling layers, where each of the multiple pooling layers corresponds to a different scale, so as to perform pooling operations at multiple scales on the targeted important feature through the pyramid pooling network to obtain important scale features corresponding to each of the multiple scales, and perform feature fusion on the important scale features corresponding to each of the multiple scales to obtain an important pooled feature of the targeted important feature.

[0058] In the above embodiment, by performing pooling operations at multiple scales on the important feature respectively, important scale features corresponding to each of the multiple scales are obtained, and the important feature is understood from different scales, capturing information of the important feature at different degrees from local to global. Furthermore, by performing feature fusion on the important scale features corresponding to each of the multiple scales, an important pooled feature integrating multi-scale information can be obtained, improving the expression ability of the important feature.

[0059] In one embodiment, for each secondary feature, a pooling operation is performed on the targeted secondary feature to obtain a secondary pooled feature of the targeted secondary feature, including: for each secondary feature, performing pooling operations at multiple scales on the targeted secondary feature to obtain secondary scale features corresponding to each of the multiple scales; and performing feature fusion on the secondary scale features corresponding to each of the multiple scales to obtain a secondary pooled feature of the targeted secondary feature.

[0060] In one embodiment, for each secondary feature, the computer device can input the targeted secondary feature into a pyramid pooling network including multiple pooling layers, where each of the multiple pooling layers corresponds to a different scale, so as to perform pooling operations at multiple scales on the targeted secondary feature through the pyramid pooling network to obtain secondary scale features corresponding to each of the multiple scales, and perform feature fusion on the secondary scale features corresponding to each of the multiple scales to obtain a secondary pooled feature of the targeted secondary feature.

[0061] In the above embodiments, by performing pooling operations on secondary features at multiple scales respectively, secondary scale features corresponding to each scale are obtained, and the secondary features are understood from different scales, capturing information of the secondary features at different degrees from local to global. Furthermore, by fusing the secondary scale features corresponding to each scale, secondary pooling features integrating multi-scale information can be obtained, enhancing the expression ability of the secondary features.

[0062] In one embodiment, the gas category to which the exhaled gas of the target object belongs is identified by a trained gas recognition model; the gas recognition model includes a feature classification network, a pyramid pooling network, and a dual encoder network; the dual encoder network includes a primary encoder and a secondary encoder; the primary feature and the secondary feature are classified by the feature classification network; the primary pooling feature is obtained by processing the primary encoder in the pyramid pooling network; the secondary pooling feature is obtained by processing the secondary encoder in the pyramid pooling network.

[0063] In one embodiment, the gas recognition model includes a feature extraction network, a feature classification network, a pyramid pooling network, a dual encoder network, and a recognition network. Among them, the dual encoder network includes a primary encoder and a secondary encoder. Specifically, the computer device can obtain multiple sets of gas data collected for the target object by multiple gas sensors, and input the multiple sets of gas data into the feature extraction network to determine the features of each set of gas data through the feature extraction network, obtaining multiple gas features. The computer device can input the multiple gas features into the feature classification network to classify the multiple gas features into primary features and secondary features through the feature classification network. For each primary feature, the computer device can input the corresponding primary feature into the pyramid pooling network to perform a pooling operation on the corresponding primary feature through the pyramid pooling network, obtaining the primary pooling feature of the corresponding primary feature, and input the primary pooling feature of the corresponding primary feature into the primary encoder to encode the input primary pooling feature through the primary encoder, obtaining the primary encoded feature corresponding to the corresponding primary feature. For each secondary feature, the computer device can input the corresponding secondary feature into the pyramid pooling network to perform a pooling operation on the corresponding secondary feature through the pyramid pooling network, obtaining the secondary pooling feature of the corresponding secondary feature, and input the secondary pooling feature of the corresponding secondary feature into the secondary encoder to encode the input secondary pooling feature through the secondary encoder, obtaining the secondary encoded feature corresponding to the corresponding secondary feature. The computer device can identify the gas category to which the exhaled gas of the target object belongs through the recognition network in the gas recognition model according to the primary encoded features corresponding to each primary feature and the secondary encoded features corresponding to each secondary feature.

[0064] In the above embodiments, by using the trained gas recognition model to recognize the gas category to which the exhaled gas of the target object belongs, the accuracy of gas recognition can be further improved.

[0065] In one embodiment, as Figure 2 shown, the gas recognition model to be trained includes a feature extraction network, a feature classification network, a pyramid pooling network, a dual encoder network, and a recognition network. Among them, the dual encoder network includes a primary encoder and a secondary encoder. Specifically, the computer device can obtain multiple sets of original gas data collected for a sample object through multiple gas sensors, and through a series of data augmentation techniques, expand the number of samples to several times that of the original gas data to obtain a large amount of sample gas data. The computer device can add white noise to the sample gas data through the sliding window technique to ensure the diversity of the sample data and avoid overfitting caused by some small samples during the training process. Then, the computer device can input the sample gas data into the feature extraction network to extract the respective features of the sample gas data through discrete wavelet transform by the feature extraction network, obtaining multiple sample gas features. The computer device can input the multiple sample gas features into the feature classification network to classify the multiple sample gas features into primary features and secondary features based on the SHAP analysis technique by the feature classification network. For each primary feature, the computer device can input the targeted primary feature into the pyramid pooling network including 4 pooling layers to perform multi-scale pooling operations on the targeted primary feature through the pyramid pooling network, obtaining the important pooling feature of the targeted primary feature, and input the important pooling feature of the targeted primary feature into the primary encoder to encode the input important pooling feature through the primary encoder, obtaining the important encoded feature corresponding to the targeted primary feature. For each secondary feature, the computer device can input the targeted secondary feature into the pyramid pooling network including 4 pooling layers to perform multi-scale pooling operations on the targeted secondary feature through the pyramid pooling network, obtaining the secondary pooling feature of the targeted secondary feature, and input the secondary pooling feature of the targeted secondary feature into the secondary encoder to encode the input secondary pooling feature through the secondary encoder, obtaining the secondary encoded feature corresponding to the targeted secondary feature. The computer device can use the recognition network in the gas recognition model to recognize the gas category to which the exhaled gas of the sample object belongs according to the important encoded features corresponding to each primary feature and the secondary encoded features corresponding to each secondary feature, so as to iteratively train the gas recognition model to be trained and obtain the trained gas recognition model.

[0066] In one embodiment, as Figure 3As shown in the figure, the pyramid pooling network includes four pooling layers, a fusion layer, a convolutional layer, a normalization layer, and an activation function. Each pooling layer corresponds to a different scale. It can be understood that the input of each pooling layer is a feature map, and the sizes of the feature maps required for input by each pooling layer are different. Specifically, for each important feature, the computer device can input the important feature to the pyramid pooling network including four pooling layers. The four pooling layers each correspond to a different scale, so as to perform pooling operations on the important feature at four scales respectively through the pyramid pooling network, obtain important scale features corresponding to the four scales respectively, and input the important scale features corresponding to the four scales to the fusion layer for feature splicing in the channel dimension, and convert the spliced features into the original number of channels through the convolutional layer, and then through the normalization layer and the activation function to enhance the non-linear expression ability and generalization ability of the network, and finally output the important pooling features of the important feature. For each secondary feature, the computer device can input the secondary feature to the pyramid pooling network including four pooling layers. The four pooling layers each correspond to a different scale, so as to perform pooling operations on the secondary feature at four scales respectively through the pyramid pooling network, obtain secondary scale features corresponding to the four scales respectively, and input the secondary scale features corresponding to the four scales to the fusion layer for feature splicing in the channel dimension, and convert the spliced features into the original number of channels through the convolutional layer, and then through the normalization layer and the activation function to enhance the non-linear expression ability and generalization ability of the network, and finally output the secondary pooling features of the secondary feature.

[0067] It should be understood that although the steps in the flowcharts of the above embodiments are shown in sequence, these steps are not necessarily executed in sequence. Unless there is a clear description in this article, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. Moreover, at least a part of the steps in the above embodiments may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed alternately or alternately with at least a part of other steps or sub-steps or stages of other steps.

[0068] In one embodiment, as Figure 4 shown, a gas recognition device 400 is provided. The device specifically includes:

[0069] An acquisition module 402, configured to acquire multiple sets of gas data collected by a plurality of gas sensors for a target object, determine the features of each set of gas data, and obtain a plurality of gas features;

[0070] A classification module 404 for classifying multiple gas features into important features and secondary features;

[0071] A feature processing module 406 for, for each important feature, performing a pooling operation on the targeted important feature to obtain an important pooled feature of the targeted important feature, and encoding the important pooled feature through an important encoder to obtain an important encoded feature corresponding to the targeted important feature; for each secondary feature, performing a pooling operation on the targeted secondary feature to obtain a secondary pooled feature of the targeted secondary feature, and encoding the secondary pooled feature through a secondary encoder to obtain a secondary encoded feature corresponding to the targeted secondary feature;

[0072] An identification module 408 for identifying the gas category to which the exhaled gas of the target object belongs according to the important encoded features respectively corresponding to the respective important features and the secondary encoded features respectively corresponding to the respective secondary features.

[0073] In one embodiment, the classification module 404 is further configured to determine the absolute value of the SHAP value of each of the multiple gas features; for each gas feature, the absolute value of the SHAP value of the targeted gas feature is positively correlated with the contribution degree of the targeted gas feature to the gas identification result; classify the multiple gas features into important features and secondary features according to the absolute values of the SHAP values of the multiple gas features; the absolute value of the SHAP value of the important feature is greater than the absolute value of the SHAP value of the secondary feature.

[0074] In one embodiment, the classification module 404 is further configured to calculate:

[0075]

[0076] where S represents a set composed of all feature combinations of multiple gas features, f represents an identification function, f(S) represents the identification result when S exists, and f(S∪{i}) is the identification result after the gas feature i is added to S; represents the marginal contribution value of the gas feature i in S; N represents a set composed of multiple gas features, |S| represents the size of S, and |N| represents the total number of features; φ i represents the SHAP value of the gas feature i.

[0077] In one embodiment, the feature processing module 406 is further configured to, for each important feature, perform pooling operations at multiple scales on the targeted important feature to obtain important scale features corresponding to the respective scales; perform feature fusion on the important scale features corresponding to the respective scales to obtain an important pooled feature of the targeted important feature.

[0078] In one embodiment, the feature processing module 406 is further configured to, for each secondary feature, perform pooling operations at multiple scales on the targeted secondary feature respectively to obtain secondary scale features corresponding to each scale; and fuse the secondary scale features corresponding to each scale to obtain the secondary pooling feature of the targeted secondary feature.

[0079] In one embodiment, the gas category to which the exhaled gas of the target object belongs is identified by a trained gas recognition model; the gas recognition model includes a feature classification network, a pyramid pooling network, and a dual encoder network; the dual encoder network includes a primary encoder and a secondary encoder; the primary feature and the secondary feature are classified by the feature classification network; the primary pooling feature is processed by the primary encoder in the pyramid pooling network; and the secondary pooling feature is processed by the secondary encoder in the pyramid pooling network.

[0080] The above gas recognition device acquires multiple sets of gas data collected for the target object by multiple gas sensors, determines the features of each set of gas data, and obtains multiple gas features; classifies the multiple gas features into primary features and secondary features; for each primary feature, performs a pooling operation on the targeted primary feature to obtain the primary pooling feature of the targeted primary feature, and encodes the primary pooling feature through the primary encoder to obtain the primary encoded feature corresponding to the targeted primary feature; for each secondary feature, performs a pooling operation on the targeted secondary feature to obtain the secondary pooling feature of the targeted secondary feature, and encodes the secondary pooling feature through the secondary encoder to obtain the secondary encoded feature corresponding to the targeted secondary feature; and identifies the gas category to which the exhaled gas of the target object belongs according to the primary encoded features respectively corresponding to the respective primary features and the secondary encoded features respectively corresponding to the respective secondary features. In this way, in this application, by classifying gas features into primary features and secondary features, and combining pooling operations and a dual encoder architecture to perform differentiated processing on primary features and secondary features, it helps to pay more attention to and process primary features more carefully during gas recognition, thereby improving the accuracy of gas recognition, which can assist in the diagnosis process of lung diseases and improve the accuracy of lung disease diagnosis.

[0081] Each module in the above gas recognition device can be implemented in whole or in part by software, hardware, and their combination. The above modules can be embedded in the processor in the computer device in hardware form or be independent of it, or can be stored in the memory in the computer device in software form so that the processor can call and execute the operations corresponding to the above respective modules.

[0082] In one embodiment, a computer device is provided. The computer device can be a server, and its internal structure diagram can be as Figure 5As shown in the figure. The computer device includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O), and a communication interface. Among them, the processor, the memory, and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and external devices. The communication interface of the computer device is used to communicate with external terminals through a network connection. When the computer program is executed by the processor, it implements a gas recognition method.

[0083] In one embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as Figure 6 shown in the figure. The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit, and an input device. Among them, the processor, the memory, and the input / output interface are connected through a system bus, and the communication interface, the display unit, and the input device are connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and external devices. The communication interface of the computer device is used to communicate with external terminals in a wired or wireless manner. The wireless manner can be achieved through WIFI, a mobile cellular network, NFC (Near Field Communication), or other technologies. When the computer program is executed by the processor, it implements a gas recognition method. The display unit of the computer device is used to form a visually visible picture, which can be a display screen, a projection device, or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer covering the display screen, or a button, a trackball, or a touchpad provided on the computer device housing, or an external keyboard, touchpad, or mouse, etc.

[0084] Those skilled in the art can understand that Figure 5 and Figure 6The structure shown is only a block diagram of some structures related to the solution of this application, and does not constitute a limitation on the computer device to which the solution of this application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine some components, or have different component arrangements.

[0085] In one embodiment, a computer device is further provided, including a memory and a processor. A computer program is stored in the memory, and when the processor executes the computer program, the steps in the above method embodiments are implemented.

[0086] In one embodiment, a computer-readable storage medium is provided, storing a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0087] In one embodiment, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0088] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use, and processing of relevant data need to comply with the relevant laws, regulations, and standards of relevant countries and regions.

[0089] Those of ordinary skill in the art can understand that all or part of the processes of implementing the methods in the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it may include the processes of the above method embodiments. Among them, any reference to a memory, storage, database, or other medium used in the various embodiments provided in this application may include at least one of non-volatile and volatile memories. Non-volatile memory may include read-only memory (ROM), magnetic tape, floppy disk, flash memory, or optical memory, etc. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0090] The technical features of the above embodiments can be combined arbitrarily. For the sake of concise description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope recorded in this specification.

[0091] The above-described embodiments merely represent several implementation manners of the present application. The description is relatively specific and detailed, but it should not be construed as a limitation on the scope of the invention patent. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can still be made, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the appended claims.

Claims

1. A gas identification method, characterized in that: The method comprises: Acquire multiple groups of gas data collected from a target object by multiple gas sensors, determine the characteristics of each of the multiple groups of gas data, and obtain multiple gas characteristics; Determine the absolute value of the SHAP value of each of the multiple gas features; for each gas feature, the absolute value of the SHAP value of the targeted gas feature is positively correlated with the contribution of the targeted gas feature to the gas identification result; Classifying the plurality of gas features into important features and minor features according to the absolute values ​​of the SHAP values ​​of the plurality of gas features; the absolute value of the SHAP value of the important feature is greater than the absolute value of the SHAP value of the minor feature; For each important feature, a pooling operation is performed on the important feature to obtain an important pooling feature of the important feature, and the important pooling feature is encoded by an important encoder to obtain an important encoding feature corresponding to the important feature; For each secondary feature, a pooling operation is performed on the secondary feature to obtain a secondary pooling feature of the secondary feature, and the secondary pooling feature is encoded by a secondary encoder to obtain a secondary encoding feature corresponding to the secondary feature; the important encoder and the secondary encoder have different configurations, the important encoder has a higher dimension than the secondary encoder, and the ability of the important encoder with a higher dimension to refine features is stronger than that of the secondary encoder with a lower dimension; The gas category to which the exhaled gas of the target object belongs is identified according to the important coding features corresponding to each important feature and the secondary coding features corresponding to each secondary feature.

2. The method according to claim 1, characterized in that: Determining the absolute value of the SHAP value of each of the plurality of gas characteristics comprises: Wherein, S represents a set consisting of all feature combinations of the plurality of gas features, f represents a recognition function, f(S) represents the recognition result when S exists, and f(S∪{i}) is the recognition result after gas feature i is added to S; represents the marginal contribution value of gas feature i in S; N represents the set consisting of the multiple gas features, |S| represents the size of S, and |N| represents the total number of features; φ i represents the SHAP value of gas feature i.

3. The method according to claim 1, characterized in that For each important feature, a pooling operation is performed on the important feature to obtain an important pooling feature of the important feature, including: For each important feature, a pooling operation of multiple scales is performed on the important feature to obtain important scale features corresponding to each of the multiple scales; The important scale features corresponding to the multiple scales are respectively fused to obtain important pooled features for the important features.

4. The method according to claim 1, characterized in that: For each secondary feature, performing a pooling operation on the secondary feature to obtain a secondary pooling feature of the secondary feature includes: For each secondary feature, perform pooling operations at multiple scales on the secondary feature to obtain secondary scale features corresponding to each of the multiple scales; The secondary scale features corresponding to the multiple scales are respectively fused to obtain secondary pooling features for the secondary features.

5. The method according to any one of claims 1 to 4, characterized in that The gas category to which the exhaled gas of the target object belongs is identified by a trained gas recognition model; the gas recognition model includes a feature classification network, a pyramid pooling network and a dual encoder network; the dual encoder network includes the important encoder and the secondary encoder; the important features and the secondary features are classified by the feature classification network; the important pooling features are obtained by processing the important encoder in the pyramid pooling network; the secondary pooling features are obtained by processing the secondary encoder in the pyramid pooling network.

6. A gas identification device, characterized in that: The device comprises: An acquisition module, used to acquire multiple groups of gas data collected by multiple gas sensors for a target object, determine the characteristics of each of the multiple groups of gas data, and obtain multiple gas characteristics; A classification module, used to determine the absolute value of the SHAP value of each of the multiple gas features; for each gas feature, the absolute value of the SHAP value of the gas feature is positively correlated with the contribution of the gas feature to the gas identification result; according to the absolute value of the SHAP value of each of the multiple gas features, the multiple gas features are classified into important features and secondary features; the absolute value of the SHAP value of the important feature is greater than the absolute value of the SHAP value of the secondary feature; A feature processing module, for each important feature, performing a pooling operation on the important feature to obtain an important pooling feature of the important feature, encoding the important pooling feature through an important encoder to obtain an important coded feature corresponding to the important feature; for each secondary feature, performing a pooling operation on the secondary feature to obtain a secondary pooling feature of the secondary feature, encoding the secondary pooling feature through a secondary encoder to obtain a secondary coded feature corresponding to the secondary feature; the important encoder and the secondary encoder have different configurations, the important encoder has a higher dimension than the secondary encoder, and the ability of the important encoder with a higher dimension to refine features is stronger than that of the secondary encoder with a lower dimension; The identification module is used to identify the gas category to which the exhaled gas of the target object belongs according to the important coding features corresponding to each important feature and the secondary coding features corresponding to each secondary feature.

7. The device according to claim 6, characterized in that The feature processing module is also used to perform pooling operations of multiple scales on each important feature to obtain important scale features corresponding to each of the multiple scales; The important scale features corresponding to the multiple scales are respectively fused to obtain important pooled features for the important features.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 5 are implemented.

9. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

Images