A cognitive state detection method and apparatus

Abstract

The application provides a cognitive state detection method, comprising: obtaining target cognitive judgment item data based on detection or testing, including response time, memory ability and attention concentration; extracting feature data related to cognitive state in the cognitive judgment item data according to a preset feature extraction rule, classifying and normalizing the feature data to generate sample data; constructing a sample model, inputting the sample data into the sample model, and calculating and outputting more training data and verification data than the sample data through the sample model; constructing a cognitive state detection neural network, training the cognitive state detection neural network based on the training data and verification data, and obtaining a cognitive state detection model; obtaining target judgment item data to be detected and inputting the target judgment item data into the cognitive state detection model to calculate and obtain a cognitive state. The application can help to improve the accuracy of model testing after training by constructing a large amount of training data.

Description

Technical Field

[0001] The present application relates to the field of cognitive detection, and in particular to a cognitive state detection method and a cognitive state detection device. Background Art

[0002] Cognitive abilities (cognitive ability) refer to the human brain's ability to process, store, and retrieve information. Specifically, it refers to a person's ability to understand the structure of things, their relationships with other things, the driving forces and direction of development, and the underlying laws governing them. It is the most important psychological condition for people to successfully complete activities. Perception, memory, attention, thinking, and imagination are all considered cognitive abilities.

[0003] Cognitive ability testing is a test that measures a person's ability to learn and complete a task. It can be divided into language ability, calculation ability, perception speed, spatial ability and reasoning ability.

[0004] Currently, cognitive ability tests are performed by obtaining relevant data from a target through testing or examination. After evaluating and identifying the data, the target's cognitive ability level is determined. With the advent of computer technology and deep neural networks, cognitive ability tests can now be output through intelligent models. However, this requires a large amount of training data for more accurate predictions. Summary of the Invention

[0005] The purpose of this application is to overcome the problem of insufficient training data for intelligent models in cognitive ability analysis in the prior art and to provide a cognitive state detection method. This application also relates to a cognitive state detection device.

[0006] The present application provides a method for detecting cognitive status, comprising:

[0007] Data acquisition: obtaining target cognitive judgment data based on detection or testing, including response time, memory ability and attention concentration;

[0008] Extracting feature data related to cognitive status from the cognitive judgment item data according to preset feature extraction rules, classifying and normalizing the feature data to generate sample data;

[0009] Constructing a sample model, inputting the sample data into the sample model, and outputting training data and verification data that are larger than the sample data through calculation by the sample model;

[0010] Constructing a cognitive state detection neural network, and training the cognitive state detection neural network based on the training data and the verification data to obtain a cognitive state detection model;

[0011] The judgment item data of the target to be detected is obtained, input into the cognitive state detection model, and the cognitive state is calculated and obtained.

[0012] Optionally, build a sample model, including:

[0013] Establishing a category relevance threshold based on the sample data;

[0014] Based on the relevance threshold of the category, generating a random number within the threshold;

[0015] Based on the random number generation within the threshold, training data and verification data are generated.

[0016] Optionally, generating the correlation threshold includes:

[0017] Associate any two categories of data and calculate the range expression of the two categories;

[0018] Select a category data and generate a threshold based on the range expression.

[0019] Optionally, the cognitive state includes: normal state and abnormal state.

[0020] Optionally, the training data is data generated by a sample model; the test data is target cognitive judgment item data obtained based on detection or testing.

[0021] The present application also provides a cognitive state detection device, comprising:

[0022] The acquisition module is used for data acquisition, which acquires target cognitive judgment data based on detection or testing, including response time, memory ability and attention concentration;

[0023] a processing module, configured to extract feature data related to cognitive state from the cognitive judgment item data according to preset feature extraction rules, classify and normalize the feature data, and generate sample data;

[0024] A generation module, configured to construct a sample model, input the sample data into the sample model, and output training data and verification data that are larger than the sample data through calculation by the sample model;

[0025] A training module is used to construct a cognitive state detection neural network, and train the cognitive state detection neural network based on the training data and the verification data to obtain a cognitive state detection model;

[0026] The detection module is used to obtain the judgment item data of the target to be detected, input it into the cognitive state detection model, and calculate and obtain the cognitive state.

[0027] Optionally, the generation module constructs a sample model, including:

[0028] Establishing a category relevance threshold based on the sample data;

[0029] Based on the relevance threshold of the category, generating a random number within the threshold;

[0030] Based on the random number generation within the threshold, training data and verification data are generated.

[0031] Optionally, the generating of the correlation threshold in the processing module includes:

[0032] Associate any two categories of data and calculate the range expression of the two categories;

[0033] Select a category data and generate a threshold based on the range expression.

[0034] Optionally, the cognitive state includes: normal state and abnormal state.

[0035] Optionally, the training data is data generated by a sample model; the test data is target cognitive judgment item data obtained based on detection or testing.

[0036] Advantages and beneficial effects of this application:

[0037] The present application provides a method for detecting a cognitive state, including: data acquisition, obtaining target cognitive judgment item data based on detection or testing, including response time, memory ability, and attention concentration; extracting feature data related to the cognitive state from the cognitive judgment item data according to preset feature extraction rules, classifying and normalizing the feature data, and generating sample data; constructing a sample model, inputting the sample data into the sample model, and outputting training data and verification data that are greater than the sample data through calculation by the sample model; constructing a cognitive state detection neural network, training the cognitive state detection neural network based on the training data and verification data, and obtaining a cognitive state detection model; obtaining target judgment item data to be detected, inputting it into the cognitive state detection model, and calculating and obtaining the cognitive state. By constructing a large amount of training data, the present application helps to improve the accuracy of the test after model training. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 This is a schematic diagram of the cognitive state detection process in this application.

[0039] Figure 2 This is a schematic diagram of the sample model execution flow in this application.

[0040] Figure 3 This is a schematic diagram of the structure of the cognitive state detection device in this application. DETAILED DESCRIPTION

[0041] The present application is further described below with reference to the accompanying drawings and specific embodiments so that those skilled in the art can better understand the present application and implement it.

[0042] The following contents are all examples of specific implementation processes provided for detailed description of the technical solutions to be protected by this application. However, this application can also be implemented in other ways different from the descriptions here. Those skilled in the art can adopt different technical means to implement this application under the guidance of the concept of this application. Therefore, this application is not limited to the specific embodiments below.

[0043] The present application provides a method for detecting a cognitive state, including: data acquisition, obtaining target cognitive judgment item data based on detection or testing, including response time, memory ability, and attention concentration; extracting feature data related to the cognitive state from the cognitive judgment item data according to preset feature extraction rules, classifying and normalizing the feature data, and generating sample data; constructing a sample model, inputting the sample data into the sample model, and outputting training data and verification data that are greater than the sample data through calculation by the sample model; constructing a cognitive state detection neural network, training the cognitive state detection neural network based on the training data and verification data, and obtaining a cognitive state detection model; obtaining target judgment item data to be detected, inputting it into the cognitive state detection model, and calculating and obtaining the cognitive state. By constructing a large amount of training data, the present application helps to improve the accuracy of the test after model training.

[0044] Figure 1 This is a schematic diagram of the cognitive state detection process in this application.

[0045] Please refer to Figure 1 As shown, S101 data acquisition, obtaining target cognitive judgment item data based on detection or testing, including response time, memory ability and attention concentration;

[0046] Data acquisition is based on detection or testing. Detection refers to testing using instruments, such as detecting changes in the target's brain waves or skin waves. Car testing involves asking and answering questions about the target, as well as other similar tests, such as test questions.

[0047] Before acquiring data, the cognitive ability assessment items need to be determined. This needs to be determined based on the specific situation. For example, cognitive ability testing can be performed on children, the elderly, the general public, or patients with brain diseases. In addition to the initial test, cognitive ability testing can also be performed on intoxicated individuals. Different target cognitive assessment items can be set based on different actual situations.

[0048] In the embodiments provided in the present application, the response time, memory capacity and attention concentration are preferably used as the cognitive judgment items for cognitive ability detection. In general, the cognitive judgment items described in the present application can be selected according to the specific actual situation, but in the case where the result requirement is not a specific judgment of a certain data, the cognitive judgment items provided in the present application can be used for cognitive ability judgment in various types of actual situations.

[0049] According to the above, a plurality of groups of data can be collected, and the data can be saved or stored.

[0050] Please refer to Figure 1 As shown in the figure, S102 extracts the feature data related to the cognitive state in the cognitive judgment item data according to the preset feature extraction rule, classifies and normalizes the feature data, and generates sample data.

[0051] The feature extraction rule described in the present application is to extract the data that needs to be retained from the data, delete the data that does not need to be retained, and then form a data set that can be used.

[0052] In the feature extraction rule, the data that needs to be retained refers to the data name and data value, and the rest of the data should be deleted, for example, in the response time and memory capacity, the response time includes: name: XXX, response time: XX seconds; the memory capacity includes: name: XXX, memory time: XX seconds, or name: XXX, memory content length: XX, etc. The data name is the feature data, the data value is called the feature value, and the feature name and the feature data are collectively called the data feature.

[0053] After extracting the data feature, the cognitive judgment item data is classified based on the feature data. The data feature has been extracted as described above, so this step can also be called data feature classification.

[0054] Specifically, the data features with the same feature data are classified into a category, and then normalized, that is, the data format is unified to form sample data. It should be noted that the sample data mentioned in the present application is not the training data and test data for final model training.

[0055] Please refer to Figure 1 As shown in the figure, S103 constructs a sample model, inputs the sample data into the sample model, and calculates and outputs more training data and validation data than the sample data through the sample model.

[0056] The sample model described in the present application can generate more data according to the sample data. As shown in the figure, Figure 2 The execution process of the sample model is as follows:

[0057] S201: establishing a category correlation threshold based on the sample data;

[0058] In the previous step S101, the data features have been categorized, with each data category having different values. Furthermore, since the conditions for data collection vary, coordinates (conditions, values) can be established, and a curve equation can be constructed based on these coordinates for description. Based on the descriptions of the two categories, a correlation between the categories is described.

[0059] Preferably, time can be used as the connection point of the correlation, or other acquisition conditions can be used as the correlation connection point, the characteristic of which is that the acquisition condition can correlate two categories.

[0060] The test data is based on the detection or test to obtain the target cognitive judgment item data. In this application, with the time as the condition, the description f(x) and g(x) of any two types of data features can be obtained, and then the following relationship is established:

[0061] f(x)=K g(x)+c

[0062] K∈[a,b]

[0063] g(x)=T

[0064] At this time, the formula can be changed to:

[0065] f(x)∈[ag(x)+c, bg(x)+c], g(x)=T

[0066] Where T represents a specific eigenvalue in a category, f(x) is a description of the type of the eigenvalue, g(x) represents a description of another type, K represents the relationship between the two types, and its value range is (a, b), and c is a constant.

[0067] Then, based on the above formula, a correlation range [ag(x)+c, bg(x)+c] can be determined, and the values ​​of the two end points of the range are the correlation thresholds described in this application.

[0068] S202 generates a random number within the threshold based on the correlation threshold of the category;

[0069] In the above process, the range of (ag(x)+c, bg(x)+c) is determined, and then a closed data area is formed based on this range and the mentioned time bar x, that is, random numbers are generated within the range of (x1, x1), [ag(x)+c, bg(x)+c] and added to the sample data.

[0070] S203 generates training data and validation data based on the threshold value within the random number generation.

[0071] Finally, based on the original sample data and later generated sample data, the training data and validation data are divided. One way of division is to shuffle the original sample data and the later generated sample data and re-divide them. Another way of division is to use the original sample data as test data and the later generated sample data as training data.

[0072] Please refer to Figure 1 S104 builds a cognitive state detection neural network, trains the cognitive state detection neural network based on the training data and validation data, and obtains a cognitive state detection model.

[0073] The cognitive state detection neural network is built based on a deep neural network. When the training data is input to the cognitive state detection neural network, the cognitive state detection model is formed after the testing data is tested and qualified.

[0074] Please refer to Figure 1 S105 obtains the judgment item data of the target to be detected, inputs it to the cognitive state detection model, calculates and obtains the cognitive state.

[0075] Finally, when detecting, the judgment item data of the target is obtained, input to the cognitive state detection model, and the final result is obtained. The result can include normal state and abnormal state.

[0076] The application also provides a cognitive state detection device, which comprises an acquisition module 301, a processing module 302, a generation module 303, a training module 304, and a detection module 305.

[0077] Figure 3 It is a structure diagram of the cognitive state detection device in the application.

[0078] Please refer to Figure 3 The acquisition module 301 is used for data acquisition, and obtains the cognitive judgment item data of the target based on detection or test, including response time, memory ability and attention concentration.

[0079] The data acquisition is realized based on detection or test. The detection refers to detection by instrument equipment, such as detection of brain wave change of the target, change of skin wave, etc. The automobile test refers to test acquisition of the target by asking and answering or other same type of test acquisition, such as test questions and answers, etc.

[0080] Before acquiring data, the cognitive ability assessment items need to be determined. This needs to be determined based on the specific situation. For example, cognitive ability testing can be performed on children, the elderly, the general public, or patients with brain diseases. In addition to the initial test, cognitive ability testing can also be performed on intoxicated individuals. Different target cognitive assessment items can be set based on different actual situations.

[0081] In the examples provided herein, response time, memory capacity, and attention concentration are preferably used as cognitive assessment items for cognitive ability testing. In general, the cognitive assessment items described herein can be selected based on specific circumstances. However, if the result requirement is not specific to a particular data set, the cognitive assessment items provided herein can be used to assess cognitive ability in various situations.

[0082] According to the above, multiple sets of data can be collected and saved or stored.

[0083] Please refer to Figure 3 As shown, the processing module 302 is used to extract feature data related to cognitive state from the cognitive judgment item data according to a preset feature extraction rule, classify and normalize the feature data, and generate sample data.

[0084] The feature extraction rules described in this application specifically extract the data that needs to be retained from the data, delete the data that does not need to be retained, and then form a data set that can be used.

[0085] In the feature extraction rules, the data that needs to be retained refers to the data name and data value, and the rest of the data should be deleted. For example, in the response time and memory capacity, the response time includes: name: XXX, response time: XX seconds; the memory capacity includes: name: XXX, memory time: XX seconds, or name: XXX, memory content length: XX, etc. The data name is the feature data, the data value is called the feature value, and the feature name and feature data are collectively called data features.

[0086] After extracting the data features, the cognitive judgment item data is classified based on the feature data. Since the data features have been extracted as mentioned above, this step can also be called data feature classification.

[0087] Specifically, the data features with the same feature data are classified into one category, and finally normalized, that is, the data format is unified to form sample data. It should be noted that the sample data mentioned in this application is not the training data and test data for the final model training.

[0088] Please refer to Figure 3 As shown, the generating module 303 is used to construct a sample model, input the sample data into the sample model, and output training data and verification data that are larger than the sample data through calculation by the sample model;

[0089] The sample model described in this application can generate more data based on the sample data. Figure 2 As shown, the sample model execution process is as follows:

[0090] S201: establishing a category correlation threshold based on the sample data;

[0091] In the previous step S101, the data features have been categorized, with each data category having different values. Furthermore, since the conditions for data collection vary, coordinates (conditions, values) can be established, and a curve equation can be constructed based on these coordinates for description. Based on the descriptions of the two categories, a correlation between the categories is described.

[0092] Preferably, time can be used as the connection point of the correlation, or other acquisition conditions can be used as the correlation connection point, the characteristic of which is that the acquisition condition can correlate two categories.

[0093] The test data is based on the detection or test to obtain the target cognitive judgment item data. In this application, with the time as the condition, the description f(x) and g(x) of any two types of data features can be obtained, and then the following relationship is established:

[0094] f(x)=K g(x)+c

[0095] K∈[a,b]

[0096] g(x)=T

[0097] At this time, the formula can be changed to:

[0098] f(x)∈[ag(x)+c, bg(x)+c], g(x)=T

[0099] Where T represents a specific eigenvalue in a category, f(x) is a description of the type of the eigenvalue, g(x) represents a description of another type, K represents the relationship between the two types, and its value range is (a, b), and c is a constant.

[0100] Then, based on the above formula, a correlation range [ag(x)+c, bg(x)+c] can be determined, and the values ​​of the two end points of the range are the correlation thresholds described in this application.

[0101] S202 generates a random number within the threshold based on the correlation threshold of the category;

[0102] In the above process, the range of (ag(x)+c, bg(x)+c) is determined, and then a closed data area is formed based on this range and the mentioned time bar x, that is, random numbers are generated within the range of (x1, x1), [ag(x)+c, bg(x)+c] and added to the sample data.

[0103] S203 generates training data and verification data based on the random number generation within the threshold.

[0104] Finally, the previously acquired sample data and the subsequently generated sample data are divided into training data and validation data. One way to divide the data is to shuffle the previously acquired sample data and the subsequently generated sample data and then divide them again; another way to divide the data is to use the previously acquired sample data as test data and the subsequently generated sample data as training data.

[0105] Please refer to Figure 3 As shown, the training module 304 is used to construct a cognitive state detection neural network, and train the cognitive state detection neural network based on the training data and the verification data to obtain a cognitive state detection model;

[0106] The cognitive state detection neural network constructed based on the deep neural network forms a cognitive state detection model when the training data is input to train the cognitive state detection neural network and the test data is tested and qualified.

[0107] Please refer to Figure 3 As shown, the detection module 305 is used to obtain the judgment item data of the target to be detected, input it into the cognitive state detection model, and calculate and obtain the cognitive state.

[0108] When performing the detection, the judgment item data of the target is obtained, input into the cognitive state detection model, and the final result is obtained, and the result may include: normal state, abnormal state.

[0109] Although the preferred embodiments of the present application have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present application.

[0110] Obviously, those skilled in the art may make various changes and modifications to this application without departing from the spirit and scope of this application. Thus, if these modifications and variations of this application fall within the scope of the claims of this application and their equivalents, this application is intended to include these modifications and variations.

Claims

1. A method for detecting cognitive status, characterized in that: include: Data acquisition: obtaining target cognitive judgment data based on the test, including response time, memory ability and attention concentration; Extracting feature data related to cognitive status from the cognitive judgment item data according to preset feature extraction rules, classifying and normalizing the feature data to generate sample data; Constructing a sample model, inputting the sample data into the sample model, and outputting training data and verification data that are larger than the sample data through calculation by the sample model; The constructing of the sample model includes: establishing a correlation threshold between multiple categories based on the sample data, wherein the correlation threshold is determined by associating any two categories of data and calculating a range expression for the any two categories; generating a random number within a data interval defined by the correlation threshold, i.e., the range expression; and generating training data and validation data that are larger than the sample data based on the random number; Constructing a cognitive state detection neural network, and training the cognitive state detection neural network based on the training data and the verification data to obtain a cognitive state detection model; The judgment item data of the target to be detected is obtained, input into the cognitive state detection model, and the cognitive state is calculated and obtained.

2. The cognitive state detection method according to claim 1, characterized in that: Build a sample model, including: Establishing a category relevance threshold based on the sample data; Based on the relevance threshold of the category, generating a random number within the threshold; Based on the random number generation within the threshold, training data and verification data are generated.

3. The cognitive state detection method according to claim 2, characterized in that: The generating of the correlation threshold comprises: Associate any two categories of data and calculate the range expression of the two categories; Select a category data and generate a threshold based on the range expression.

4. The cognitive state detection method according to claim 1, characterized in that: The cognitive state includes: a normal state and an abnormal state.

5. The cognitive state detection method according to claim 1, characterized in that: The training data is data generated by a sample model.

6. A cognitive state detection device, characterized in that: include: The acquisition module is used for data acquisition, which acquires target cognitive judgment data based on the test, including response time, memory ability and attention concentration; a processing module, configured to extract feature data related to cognitive state from the cognitive judgment item data according to preset feature extraction rules, classify and normalize the feature data, and generate sample data; A generation module, configured to construct a sample model, input the sample data into the sample model, and output training data and verification data that are larger than the sample data through calculation by the sample model; The constructing of the sample model includes: establishing a correlation threshold between multiple categories based on the sample data, wherein the correlation threshold is determined by associating any two categories of data and calculating a range expression for the any two categories; generating a random number within a data interval defined by the correlation threshold, i.e., the range expression; and generating training data and validation data that are larger than the sample data based on the random number; A training module is used to construct a cognitive state detection neural network, and train the cognitive state detection neural network based on the training data and the verification data to obtain a cognitive state detection model; The detection module is used to obtain the judgment item data of the target to be detected, input it into the cognitive state detection model, and calculate and obtain the cognitive state.

7. The cognitive state detection device according to claim 6, characterized in that: The generation module constructs a sample model, including: Establishing a category relevance threshold based on the sample data; Based on the relevance threshold of the category, generating a random number within the threshold; Based on the random number generation within the threshold, training data and verification data are generated.

8. The cognitive state detection device according to claim 7, characterized in that: The generation of the correlation threshold in the processing module includes: Associate any two categories of data and calculate the range expression of the two categories; Select a category data and generate a threshold based on the range expression.

9. The cognitive state detection device according to claim 6, characterized in that: The cognitive state includes: a normal state and an abnormal state.

10. The cognitive state detection device according to claim 6, characterized in that: The training data is data generated by a sample model.

Images