Feature dynamic embedding-based traditional Chinese medicine constitution iterative identification method, system and terminal

Abstract

The invention discloses a characteristic dynamic embedding-based traditional Chinese medicine constitution iterative identification method, system and terminal, and relates to the field of artificial intelligence, and the key points of the technical scheme are as follows: under the condition that the number of initially provided constitution characteristics is limited, identifiable to-be-identified characteristics are dynamically determined according to the real-time traditional Chinese medicine constitution identification condition, so that the identification accuracy of the to-be-identified characteristics is improved; according to the method, the to-be-identified characteristics with the identifiability are obtained, the to-be-identified characteristics with the identifiability are fused with the existing physical characteristics, so that rapid identification of the traditional Chinese medicine physique is realized, and the data volume of source data acquisition is effectively reduced under the condition that the identification accuracy of the traditional Chinese medicine physique is guaranteed.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence, and more specifically, to a method, system and terminal for iterative identification of TCM constitution based on dynamic feature embedding. Background Art

[0002] TCM constitution refers to the unique physiological and pathological characteristics of an individual formed during the growth and development process under the influence of congenital genetics and acquired environmental factors. In general, TCM divides constitution into the following nine basic types: peaceful constitution, qi deficiency constitution, yang deficiency constitution, yin deficiency constitution, phlegm-damp constitution, damp-heat constitution, blood stasis constitution, qi stagnation constitution, and special constitution. TCM constitution identification can help people better understand their physical condition and health trends, so as to take more personalized and effective health management measures.

[0003] With the continuous development of artificial intelligence technology, the existing TCM constitution identification method mainly mines and analyzes a large amount of constitution data through machine learning and deep learning algorithms, which mainly consider facial features, tongue features, pulse features and fingerprint features, and some also consider features such as emotions, living habits, eating habits, voice, body shape and posture. However, due to the different feature dimensions of different TCM constitutions, some TCM constitutions can be accurately identified by fewer significant features, while some TCM constitutions require more features to be accurately identified; in addition, due to the existence of the same or similar features of the features shown by different TCM constitutions, and the influence of the systematic errors brought by the feature extraction process, TCM constitution misidentification is more likely to occur, especially when there is less individual data. On this basis, if the accuracy of TCM constitution identification is to be improved, the number of features required for the TCM constitution identification process needs to be increased, which not only increases the difficulty of applying TCM constitution identification technology, but also consumes a lot of network resources, which is not conducive to large-scale promotion and application.

[0004] Therefore, how to study and design a method, system and terminal for iterative identification of TCM constitution based on dynamic feature embedding that can overcome the above-mentioned defects is a problem that we urgently need to solve. Summary of the invention

[0005] In order to address the deficiencies in the prior art, the purpose of the present invention is to provide a method, system and terminal for iterative identification of TCM constitution based on dynamic feature embedding. When the number of constitution features initially provided is limited, the identifiable features to be identified are dynamically determined according to the real-time TCM constitution identification situation, and the identifiable features to be identified are integrated with the previously existing constitution features, thereby realizing rapid identification of TCM constitution, and effectively reducing the amount of source data collected while ensuring the accuracy of TCM constitution identification.

[0006] The above technical objectives of the present invention are achieved through the following technical solutions:

[0007] In the first aspect, a method for iterative identification of TCM constitution based on dynamic feature embedding is provided, comprising the following steps:

[0008] Collecting a facial image, a tongue image and / or a fingerprint image of a target object to obtain a target image;

[0009] Extracting initial physical features from the target image to obtain a physical feature set;

[0010] Select the characteristics of a single TCM constitution from the constitution characteristic set to obtain a sub-characteristic set corresponding to each TCM constitution;

[0011] Determine at least one feature to be identified based on the first N sub-feature sets with the largest number of features and the full feature set of the TCM constitution corresponding to the first N sub-feature sets, and generate an information collection strategy based on the detection attributes of the feature to be identified;

[0012] Execute the information collection strategy to collect feature information of the feature to be identified, and extract the newly added physical features from the feature information;

[0013] The newly added physical characteristics are embedded and fused with the sub-feature set to obtain an updated sub-feature set;

[0014] The updated sub-feature sets are iterated to embed and fuse the newly added physical features until the difference in the number of features between the first two sub-feature sets with the largest number of features among all updated sub-feature sets exceeds the feature threshold;

[0015] The TCM constitution corresponding to the updated sub-feature set with the largest number of features is taken as the TCM constitution identification result and output.

[0016] Furthermore, the process of selecting the characteristics of a single TCM constitution from the set of constitution characteristics is specifically as follows:

[0017] Statistically analyze all the physical characteristics of a single TCM constitution from historical data and establish a complete feature set;

[0018] Solve the intersection of the full feature set and the constitution feature set to obtain the sub-feature set of the corresponding TCM constitution;

[0019] If there are two mutually exclusive physical characteristics in the sub-feature set, the two mutually exclusive physical characteristics are deleted.

[0020] Furthermore, the determination expression of the feature to be identified is specifically:

[0021] Filter out the top N sub-feature sets with the largest number of features from all sub-feature sets;

[0022] All feature items with the same identification object are screened out from the full feature set of TCM constitution corresponding to the first N sub-feature sets to obtain an initial feature item set;

[0023] Delete the feature items in the initial feature item set contained in any of the first N sub-feature sets to obtain the final feature item set;

[0024] Each feature item in the final feature item set is used as a feature to be identified.

[0025] Furthermore, the method also includes:

[0026] If the number of feature items in the final feature item set is zero, the first N-1 sub-feature sets with the largest number of features are used to replace the first N sub-feature sets with the largest number of features to iterate and determine the final feature item set.

[0027] If the number of feature items in the final feature item set determined by the first two sub-feature sets with the largest number of features is zero, then one feature item that is not included in the corresponding sub-feature set is selected from the full feature set of the TCM constitution corresponding to the first two sub-feature sets, and the two selected feature items are combined to construct a feature to be identified.

[0028] Furthermore, the process of generating an information collection strategy based on the detection attributes of the feature to be identified is specifically as follows:

[0029] If the detection attribute of the feature to be identified is image detection, the generated information acquisition strategy includes: controlling the image acquisition device to start and stop, preprocessing the acquired image, and extracting the physical characteristics corresponding to the feature to be identified from the acquired image;

[0030] If the detection attribute of the feature to be identified is text detection, the generated information collection strategy includes: simulating the generation of question information corresponding to the feature to be identified, receiving answer information corresponding to the question information, and extracting the physical characteristics corresponding to the feature to be identified from the answer information.

[0031] Furthermore, the process of embedding and fusing the newly added physical features with the sub-feature set is specifically as follows:

[0032] If the newly added physical characteristics are included in the full feature set corresponding to the corresponding sub-feature set, the newly added physical characteristics are written into the corresponding sub-feature set to obtain an updated sub-feature set;

[0033] If the newly added physical characteristic is not included in the full characteristic set corresponding to the corresponding sub-characteristic set, the corresponding sub-characteristic set will not be updated.

[0034] Furthermore, the process of determining the characteristic threshold is specifically as follows:

[0035] Determine the judgment threshold, the standard threshold and the number of real-time features of the Nth sub-feature set with the largest number of features;

[0036] If the number of real-time features is less than or equal to the judgment threshold, the feature threshold is taken as the standard threshold;

[0037] If the number of real-time features is greater than the judgment threshold, the real-time ratio of the number of real-time features to the judgment threshold is calculated, and the feature threshold is calculated by the product of the real-time ratio and the standard threshold;

[0038] Alternatively, the feature threshold adopts a standard threshold.

[0039] Secondly, a TCM constitution iterative identification system based on dynamic feature embedding is provided, including:

[0040] An image acquisition module is used to acquire a facial image, a tongue image and / or a fingerprint image of a target object to obtain a target image;

[0041] A feature extraction module is used to extract initial physical features from the target image to obtain a physical feature set;

[0042] A feature selection module is used to select the features of a single TCM constitution from the constitution feature set to obtain a sub-feature set corresponding to each TCM constitution;

[0043] A strategy generation module, used to determine at least one feature to be identified based on the first N sub-feature sets with the largest number of features and the full feature set of the TCM constitution corresponding to the first N sub-feature sets, and generate an information collection strategy based on the detection attributes of the feature to be identified;

[0044] An information collection module, used to execute the information collection strategy to collect feature information of the feature to be identified, and extract the newly added physical features from the feature information;

[0045] A feature fusion module is used to embed and fuse the newly added physical features with the sub-feature set to obtain an updated sub-feature set;

[0046] An iterative identification module is used to iterate the updated sub-feature sets to embed and fuse the newly added physical features until the difference in the number of features between the first two sub-feature sets with the largest number of features among all the updated sub-feature sets exceeds a feature threshold;

[0047] The result output module is used to take the TCM constitution corresponding to the updated sub-feature set with the largest number of features as the TCM constitution identification result and output it.

[0048] In a third aspect, a computer terminal is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the iterative identification method for TCM constitution based on dynamic feature embedding as described in any one of the first aspects is implemented.

[0049] In a fourth aspect, a computer-readable medium is provided, on which a computer program is stored, and the computer program is executed by a processor to implement the method for iterative identification of TCM constitution based on dynamic feature embedding as described in any one of the first aspects.

[0050] Compared with the prior art, the present invention has the following beneficial effects:

[0051] 1. The TCM constitution iterative identification method based on dynamic feature embedding provided by the present invention dynamically determines the identifiable features to be identified based on the real-time TCM constitution identification situation when the number of initially provided constitution features is limited, and merges the identifiable features to be identified with the previously existing constitution features, thereby realizing rapid TCM constitution identification and effectively reducing the amount of source data collection while ensuring the accuracy of TCM constitution identification;

[0052] 2. In the process of TCM constitution identification, the present invention takes into account the influence of source data error and system error in feature extraction process. There are two mutually exclusive constitution features in the sub-feature set, and the two mutually exclusive constitution features are deleted, thereby ensuring the accuracy and reliability of TCM constitution identification;

[0053] 3. When determining the features to be identified, the present invention takes the first N sub-feature sets with the largest number of features as the basis, and selects the feature items with the same identification object and not included in any of the first N sub-feature sets as the features to be identified, which can effectively ensure the identifiability of the features to be identified;

[0054] 4. The present invention dynamically determines the size of the feature threshold according to the number of real-time features of the sub-feature set, and can maintain the difference requirement of the overall recognition when the number of features increases. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] The drawings described herein are used to provide a further understanding of the embodiments of the present invention, constitute a part of this application, and do not constitute a limitation of the embodiments of the present invention. In the drawings:

[0056] Figure 1 is a flow chart of Embodiment 1 of the present invention;

[0057] Figure 2 It is a system block diagram in Example 2 of the present invention. DETAILED DESCRIPTION

[0058] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with embodiments and drawings. The exemplary embodiments of the present invention and their description are only used to explain the present invention and are not intended to limit the present invention.

[0059] Embodiment 1: TCM constitution iterative identification method based on dynamic feature embedding, such as Figure 1 As shown, the following steps are included:

[0060] S1: Collecting a facial image, a tongue image and / or a fingerprint image of a target object to obtain a target image;

[0061] S2: extracting initial physical features from the target image to obtain a physical feature set;

[0062] S3: Select the characteristics of a single TCM constitution from the constitution characteristic set to obtain a sub-feature set corresponding to each TCM constitution;

[0063] S4: determining at least one feature to be identified based on the first N sub-feature sets with the largest number of features and the full feature set of the TCM constitution corresponding to the first N sub-feature sets, and generating an information collection strategy based on the detection attributes of the feature to be identified;

[0064] S5: executing the information collection strategy to collect feature information of the feature to be identified, and extracting the newly added physical features from the feature information;

[0065] S6: embed and fuse the newly added physical characteristics with the sub-feature set to obtain an updated sub-feature set;

[0066] S7: iterating the updated sub-feature sets to embed and fuse the newly added physical features, until the difference in the number of features between the first two sub-feature sets with the largest number of features among all updated sub-feature sets exceeds the feature threshold;

[0067] S8: The TCM constitution corresponding to the updated sub-feature set with the largest number of features is taken as the TCM constitution identification result and output.

[0068] In step S1, considering the convenience of data collection and ensuring the rationality of the initially provided physical characteristics, the present invention extracts physical characteristics by collecting some image information of the target object as source data. The collected image can be at least one of a facial image, a tongue image, and a fingerprint image, which is not limited here.

[0069] In step S2, if there are multiple types of collected images, all feature items related to TCM constitution identification are extracted from different images respectively, and each feature item may show different characteristics for different TCM constitutions.

[0070] 1. Taking facial images as an example, the feature items related to TCM constitution contained in facial images include but are not limited to complexion, skin texture, facial luster, eyelid puffiness, facial morphology, eyebrow thickness, and lip thickness.

[0071] Specifically, (1) changes in complexion can reflect the state of the body's qi and blood. For example, a pale complexion may be related to qi and blood deficiency, a sallow complexion may be related to spleen and stomach weakness, a dark complexion may be related to kidney deficiency or blood stasis, a bluish complexion may be related to qi and blood stasis, and a red complexion may be related to heat syndrome or yin deficiency. (2) Changes in skin texture, such as the number and distribution of wrinkles, can also reflect the state of the body. For example, people with yin deficiency often have dry skin and are prone to wrinkles. (3) The glossiness of the face can reflect the state of the body's spirit. Good glossiness usually indicates sufficient qi and blood, while a dull complexion may indicate insufficient qi and blood or other health problems. (4) Eyelid swelling may be related to poor kidney function or internal water retention. (5) Facial shape, such as the fatness or thinness of the face, can also reflect the body's constitution. For example, facial obesity may be related to phlegm-dampness constitution. (6) The thickness and shape of the eyebrows may be related to the release of liver qi. (7) The thickness and color of the lips can reflect the health of the spleen and stomach.

[0072] In addition, the feature items in the facial image can also include facial muscle state and facial temperature, etc. The tension or relaxation of facial muscles can also provide physical information. The facial temperature distribution may be related to the circulation of qi and blood and the functional state of internal organs.

[0073] 2. Taking the tongue image as an example, the feature items related to TCM constitution contained in the tongue image include but are not limited to tongue color, tongue coating, tongue shape, tongue quality, tooth marks, sublingual veins and tongue body dynamics.

[0074] Specifically, (1) changes in tongue color can reflect the state of Qi and blood. For example, a pale tongue may be associated with Qi and blood deficiency, a red tongue may be associated with heat syndrome or Yin deficiency, and a purple tongue may be associated with Qi and blood stasis. (2) The color and thickness of the tongue coating can reflect the state of dampness, heat, cold and dampness in the body. For example, a white coating may be associated with cold syndrome or external syndrome, a yellow coating may be associated with heat syndrome, a thick coating may be associated with phlegm dampness or food accumulation, and a thin coating may be associated with Yin deficiency or external syndrome. (3) The shape of the tongue, such as fat, thin, or cracked, can also reflect the state of the body constitution. For example, a fat tongue may be associated with phlegm dampness constitution, and a thin tongue may be associated with Yin deficiency or Qi and blood deficiency. (4) The wetness and elasticity of the tongue can reflect the sufficiency of body fluids and Qi and blood. A dry tongue may be associated with Yin deficiency or heat syndrome, and a moist tongue may be associated with internal water retention. (5) Tooth marks on the edge of the tongue can reflect spleen deficiency or internal water retention. (6) The fullness and color of the sublingual veins can reflect the blood circulation and blood stasis. (7) The movement and extension of the tongue can reflect the functional state of the nervous system.

[0075] 3. Taking fingerprint images as an example, the feature items related to TCM constitution contained in the fingerprint images include but are not limited to feature items such as fingerprint texture, morphology and distribution.

[0076] The physical characteristics extracted from each source data are combined to form a physical characteristic set.

[0077] In step S3, the process of selecting the characteristics exhibited by a single TCM constitution from the constitution characteristic set is specifically as follows: statistically analyzing all the constitution characteristics exhibited by a single TCM constitution from historical data to establish a full feature set; performing intersection solving on the full feature set and the constitution characteristic set to obtain a sub-feature set of the corresponding TCM constitution; wherein, if there are two mutually exclusive constitution characteristics in the sub-feature set, the two mutually exclusive constitution characteristics are deleted.

[0078] For example, if the constitution feature set contains 11 constitution features, {a, b, c, d, e, f, g, h, I, j, k}, and assuming that the full feature set of Yin deficiency constitution is {a, b, f, j, k, m, q, s, …}, then the sub-feature set corresponding to Yin deficiency constitution selected from the constitution feature set is {a, b, f, j, k}.

[0079] In the process of TCM constitution identification, the present invention takes into account the influence of source data errors and system errors in the feature extraction process. If there are two mutually exclusive constitution features in the sub-feature set, the two mutually exclusive constitution features are deleted, thereby ensuring the accuracy and reliability of TCM constitution identification.

[0080] For example, if the constitution feature a and the constitution feature b do not exist simultaneously in the Yin deficiency constitution, the constitution feature a and the constitution feature b can be deleted from the sub-feature set at the same time, and the final sub-feature set is {f, j, k}.

[0081] In step S4, the determination expression of the feature to be identified is specifically as follows: the first N sub-feature sets with the largest number of features are selected from all sub-feature sets; all feature items with the same identification object are selected from the full feature set of the TCM constitution corresponding to the first N sub-feature sets to obtain an initial feature item set; the feature items in the initial feature item set that are included in any of the first N sub-feature sets are deleted to obtain a final feature item set; each feature item in the final feature item set is used as a feature to be identified.

[0082] For example, the top three sub-feature sets with the largest number of features are: phlegm-damp constitution {a, f, g, h, I, j}, damp-heat constitution {a, b, e, f, k}, and yin deficiency constitution {f, j, k}.

[0083] The full feature set of phlegm-dampness constitution is {a,f,g,h,I,j,r,y,z,s,…}, the full feature set of damp-heat constitution is {a,b,e,f,k,g,z,x,s,…}, and the full feature set of yin deficiency constitution is {a,b,f,j,k,m,q,s,…}. All feature items with the same identification object are screened out as the feature items corresponding to a and s. Since a is in the above sub-feature set, the final feature item set is {S}, where S is the feature item corresponding to s.

[0084] In addition, if the number of feature items in the final feature item set is zero, the first N-1 sub-feature sets with the largest number of features are used to replace the first N sub-feature sets with the largest number of features to iteratively determine the final feature item set. If the number of feature items in the final feature item set determined by the first two sub-feature sets with the largest number of features is zero, one feature item that is not included in the corresponding sub-feature set is selected from each of the full feature sets of the TCM constitutions corresponding to the first two sub-feature sets, and the two selected feature items are combined to construct a feature to be identified.

[0085] The process of generating an information collection strategy based on the detection attribute of the feature to be identified is specifically as follows: if the detection attribute of the feature to be identified is image detection, the generated information collection strategy includes: controlling the image acquisition device to start and stop, preprocessing the acquired image, and extracting the physical characteristics corresponding to the feature to be identified from the acquired image; if the detection attribute of the feature to be identified is text detection, the generated information collection strategy includes: simulating the generation of question information corresponding to the feature to be identified, receiving answer information corresponding to the question information, and extracting the physical characteristics corresponding to the feature to be identified from the answer information.

[0086] In step S5, assuming that the feature to be identified is eating habits, the detection attribute is text detection, and the information collection strategy can be voice questioning or text questioning. To this end, an artificial voice template can be used to generate a series of question information, such as: "Do you like spicy food?", "Do you like greasy food?", etc.

[0087] The process of executing the information collection strategy also includes controlling the start-up of devices such as voice players and sound collectors to achieve automatic information collection.

[0088] In step S6, the process of embedding and fusing the newly added physical characteristics with the sub-feature set is specifically as follows: if the newly added physical characteristics are included in the full feature set corresponding to the corresponding sub-feature set, the newly added physical characteristics are written into the corresponding sub-feature set to obtain an updated sub-feature set; if the newly added physical characteristics are not included in the full feature set corresponding to the corresponding sub-feature set, the corresponding sub-feature set is not updated.

[0089] If the newly added physical feature is s, the updated sub-feature set is {f, j, k, s}.

[0090] In step S7, after each update, it is determined whether the difference between the feature numbers of the first two sub-feature sets with the largest feature numbers in the sub-feature set exceeds the feature threshold. If not, the iteration continues; if so, the identification process terminates.

[0091] As an optional implementation, the feature threshold may be set as a fixed standard threshold based on historical data.

[0092] As another optional implementation, the process of determining the feature threshold is specifically as follows: determining the judgment threshold, the standard threshold and the number of real-time features of the Nth sub-feature set with the largest number of features; if the number of real-time features is less than or equal to the judgment threshold, the feature threshold is taken as the standard threshold; if the number of real-time features is greater than the judgment threshold, the real-time ratio of the real-time feature number to the judgment threshold is calculated, and the feature threshold is calculated as the product of the real-time ratio and the standard threshold.

[0093] The present invention dynamically determines the size of the feature threshold according to the number of real-time features of the sub-feature set, and can maintain the difference requirement of the overall recognition when the number of features increases.

[0094] In step S8, the TCM constitution identification result can be sent to a printing device for printing, or it can be connected to other electronic devices, terminals, software apps, etc. to remotely transmit the TCM constitution identification result, which is not limited here.

[0095] Embodiment 2: A system for iterative identification of TCM constitution based on dynamic feature embedding, the system is used to implement the method for iterative identification of TCM constitution based on dynamic feature embedding as described in Embodiment 1, Figure 2 As shown, it includes an image acquisition module, a feature extraction module, a feature selection module, a strategy generation module, an information acquisition module, a feature fusion module, an iterative identification module and a result output module.

[0096] Among them, the image acquisition module is used to collect the facial image, tongue image and / or fingerprint image of the target object to obtain the target image; the feature extraction module is used to extract the initial physical characteristics from the target image to obtain the physical characteristic set; the feature selection module is used to select the characteristics of a single TCM constitution from the physical characteristic set to obtain a sub-feature set corresponding to the TCM constitution; the strategy generation module is used to determine at least one feature to be identified based on the first N sub-feature sets with the largest number of features and the full feature set of the TCM constitution corresponding to the first N sub-feature sets, and generate an information collection strategy based on the detection attributes of the feature to be identified; the information The acquisition module is used to execute the information acquisition strategy to collect the feature information of the feature to be identified, and extract the newly added physical characteristics from the feature information; the feature fusion module is used to embed and fuse the newly added physical characteristics with the sub-feature set to obtain an updated sub-feature set; the iterative identification module is used to iterate the updated sub-feature set to embed and fuse the newly added physical characteristics until the difference in the number of features of the first two sub-feature sets with the largest number of features in all updated sub-feature sets exceeds the feature threshold; the result output module is used to use the TCM constitution corresponding to the updated sub-feature set with the largest number of features as the TCM constitution identification result, and output it.

[0097] The present invention also records a computer terminal, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the iterative identification method for TCM constitution based on dynamic feature embedding as described in Example 1 is implemented.

[0098] The present invention also records a computer-readable medium on which a computer program is stored. The computer program is executed by a processor to implement the method for iterative identification of TCM constitution based on dynamic feature embedding as described in Example 1.

[0099] Working principle: When the number of physical characteristics initially provided is limited, the present invention dynamically determines the identifiable features to be identified based on the real-time TCM constitution identification situation, and merges the identifiable features to be identified with the previously existing physical characteristics, thereby realizing the rapid identification of TCM constitution, and effectively reducing the amount of source data collection while ensuring the accuracy of TCM constitution identification.

[0100] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0101] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0102] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0103] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0104] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. The iterative identification method of TCM constitution based on dynamic feature embedding is characterized by: The following steps are involved: Collecting a facial image, a tongue image and / or a fingerprint image of a target object to obtain a target image; Extracting initial physical features from the target image to obtain a physical feature set; Select the characteristics of a single TCM constitution from the constitution characteristic set to obtain a sub-characteristic set corresponding to each TCM constitution; Determine at least one feature to be identified based on the first N sub-feature sets with the largest number of features and the full feature set of the TCM constitution corresponding to the first N sub-feature sets, and generate an information collection strategy based on the detection attributes of the feature to be identified; Execute the information collection strategy to collect feature information of the feature to be identified, and extract the newly added physical features from the feature information; The newly added physical characteristics are embedded and fused with the sub-feature set to obtain an updated sub-feature set; The updated sub-feature sets are iterated to embed and fuse the newly added physical features until the difference in the number of features between the first two sub-feature sets with the largest number of features among all updated sub-feature sets exceeds the feature threshold; The TCM constitution corresponding to the updated sub-feature set with the largest number of features is taken as the TCM constitution identification result and output.

2. The TCM constitution iterative identification method based on dynamic feature embedding according to claim 1 is characterized in that: The process of selecting the characteristics of a single TCM constitution from the constitution characteristics set is specifically as follows: Statistically analyze all the physical characteristics of a single TCM constitution from historical data and establish a complete feature set; Solve the intersection of the full feature set and the constitution feature set to obtain the sub-feature set of the corresponding TCM constitution; If there are two mutually exclusive physical characteristics in the sub-feature set, the two mutually exclusive physical characteristics are deleted.

3. The TCM constitution iterative identification method based on dynamic feature embedding according to claim 1 is characterized in that: The specific expression for determining the feature to be identified is: Filter out the top N sub-feature sets with the largest number of features from all sub-feature sets; All feature items with the same identification object are screened out from the full feature set of TCM constitution corresponding to the first N sub-feature sets to obtain an initial feature item set; Delete the feature items in the initial feature item set contained in any of the first N sub-feature sets to obtain the final feature item set; Each feature item in the final feature item set is used as a feature to be identified.

4. The TCM constitution iterative identification method based on dynamic feature embedding according to claim 3 is characterized in that: The method further includes: If the number of feature items in the final feature item set is zero, the first N-1 sub-feature sets with the largest number of features are used to replace the first N sub-feature sets with the largest number of features to iterate and determine the final feature item set. If the number of feature items in the final feature item set determined by the first two sub-feature sets with the largest number of features is zero, then one feature item that is not included in the corresponding sub-feature set is selected from the full feature set of the TCM constitution corresponding to the first two sub-feature sets, and the two selected feature items are combined to construct a feature to be identified.

5. The TCM constitution iterative identification method based on dynamic feature embedding according to claim 1 is characterized in that: The process of generating an information collection strategy based on the detection attributes of the feature to be identified is specifically as follows: If the detection attribute of the feature to be identified is image detection, the generated information acquisition strategy includes: controlling the image acquisition device to start and stop, preprocessing the acquired image, and extracting the physical characteristics corresponding to the feature to be identified from the acquired image; If the detection attribute of the feature to be identified is text detection, the generated information collection strategy includes: simulating the generation of question information corresponding to the feature to be identified, receiving answer information corresponding to the question information, and extracting the physical characteristics corresponding to the feature to be identified from the answer information.

6. The TCM constitution iterative identification method based on dynamic feature embedding according to claim 1 is characterized in that: The process of embedding and fusing the newly added physical features with the sub-feature set is specifically as follows: If the newly added physical characteristics are included in the full feature set corresponding to the corresponding sub-feature set, the newly added physical characteristics are written into the corresponding sub-feature set to obtain an updated sub-feature set; If the newly added physical characteristic is not included in the full characteristic set corresponding to the corresponding sub-characteristic set, the corresponding sub-characteristic set will not be updated.

7. The TCM constitution iterative identification method based on dynamic feature embedding according to claim 1 is characterized in that: The process of determining the characteristic threshold is specifically as follows: Determine the judgment threshold, the standard threshold and the number of real-time features of the Nth sub-feature set with the largest number of features; If the number of real-time features is less than or equal to the judgment threshold, the feature threshold is taken as the standard threshold; If the number of real-time features is greater than the judgment threshold, the real-time ratio of the number of real-time features to the judgment threshold is calculated, and the feature threshold is calculated by the product of the real-time ratio and the standard threshold; Alternatively, the feature threshold adopts a standard threshold.

8. The TCM constitution iterative identification system based on dynamic feature embedding is characterized by: include: An image acquisition module is used to acquire a facial image, a tongue image and / or a fingerprint image of a target object to obtain a target image; A feature extraction module is used to extract initial physical features from the target image to obtain a physical feature set; A feature selection module is used to select the features of a single TCM constitution from the constitution feature set to obtain a sub-feature set corresponding to each TCM constitution; A strategy generation module, used to determine at least one feature to be identified based on the first N sub-feature sets with the largest number of features and the full feature set of the TCM constitution corresponding to the first N sub-feature sets, and generate an information collection strategy based on the detection attributes of the feature to be identified; An information collection module, used to execute the information collection strategy to collect feature information of the feature to be identified, and extract the newly added physical features from the feature information; A feature fusion module is used to embed and fuse the newly added physical features with the sub-feature set to obtain an updated sub-feature set; An iterative identification module is used to iterate the updated sub-feature sets to embed and fuse the newly added physical features until the difference in the number of features between the first two sub-feature sets with the largest number of features among all the updated sub-feature sets exceeds a feature threshold; The result output module is used to take the TCM constitution corresponding to the updated sub-feature set with the largest number of features as the TCM constitution identification result and output it.

9. A computer terminal comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the iterative identification method of TCM constitution based on dynamic feature embedding as described in any one of claims 1 to 7 is implemented.

10. A computer readable medium having a computer program stored thereon, characterized in that: The computer program is executed by a processor to implement the TCM constitution iterative identification method based on dynamic feature embedding as described in any one of claims 1 to 7.

Images