Emotion recognition system and method based on large model

Abstract

The invention discloses an emotion recognition system and method based on a large model, and relates to the field of data analysis, and the method comprises the steps: collecting facial physiological dynamic data, including specific muscle group motion data and photoelectric volume pulse wave signals, of a person in a calm state, and establishing a reference contraction frequency and a reference heart rate; analyzing data characteristics and calculating an amplification threshold and an enhancement threshold; collecting face data of a target person in real time, and extracting real-time contraction frequency, duration of the real-time contraction frequency exceeding a critical value and real-time heart rate; and judging the emotional state by comparing the amplification and enhancement of the real-time data and the reference value. The system comprises a reference establishment module, a threshold calculation module, a real-time monitoring module, a state identification module and the like. According to the method, non-invasive, objective and accurate emotion recognition is realized through multi-modal physiological signal analysis, and the method has the advantages of high real-time performance and high accuracy.

Description

Technical Field

[0001] This invention relates to the field of data analysis, specifically to a large-scale model-based emotion recognition system and method. Background Technology

[0002] Emotion recognition has significant application value in fields such as human-computer interaction, mental health monitoring, and safe driving. Traditional emotion recognition methods mainly rely on facial expression analysis, speech feature extraction, or contact-based physiological signal acquisition. Among these, computer vision-based facial expression analysis technology is easily affected by ambient lighting, individual expression habits, and subjective concealment behaviors, limiting its recognition accuracy. On the other hand, contact-based physiological monitoring, such as electrocardiograms and skin conductance responses, requires wearing dedicated sensors, which is inconvenient for users and has poor comfort during long-term monitoring.

[0003] In recent years, non-contact physiological signal acquisition technology based on facial videos has made some progress, especially in extracting physiological parameters such as heart rate by analyzing facial photoplethysmography (PPG) signals. However, single-modal physiological parameters have limitations in emotion recognition, failing to comprehensively reflect changes in emotional state. Furthermore, most existing methods rely on preset fixed thresholds, lacking adaptability to individual physiological differences, resulting in significant fluctuations in recognition accuracy across different populations.

[0004] Furthermore, current emotion recognition systems often focus on identifying obvious emotional states, lacking the ability to monitor subtle emotional changes or transitions. Especially in practical applications, distinguishing between a calm state and a target emotional state, and establishing an effective judgment mechanism, remains a technical challenge that needs to be addressed.

[0005] Therefore, there is an urgent need in this field for an emotion recognition scheme that can combine multimodal physiological signals and adaptively set recognition thresholds to improve the accuracy, reliability and practicality of emotion recognition. Summary of the Invention

[0006] The purpose of this invention is to provide an emotion recognition system and method based on a large model to solve the problems raised in the prior art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: an emotion recognition system and method based on a large model, the method comprising: Step S1: Collect facial physiological dynamic data of several individuals in a calm state; the facial physiological dynamic data includes the motion dynamic data of the target facial muscle groups and the photoplethysmography (PPG) signal of the face; wherein the target facial muscle groups include the orbicularis oculi, zygomaticus major, depressor supercilii, and corrugator supercilii; based on the facial physiological dynamic data, calculate the baseline contraction frequency and baseline heart rate. Step S2: By analyzing the change characteristics in the facial physiological dynamic data, calculate the amplification threshold and enhancement threshold used for state recognition; Step S3: Select the person whose emotion is to be identified as the target person, collect real-time facial physiological dynamic data of the target person, including real-time motion dynamic data of the facial target muscle group and real-time photoplethysmography (PPG) wave signal of the face; extract the real-time contraction frequency of the target muscle group and the duration of the real-time contraction frequency exceeding a certain critical contraction frequency; and extract the real-time heart rate from the real-time PPG wave signal. Step S4: Calculate the increase in the target person's real-time contraction frequency compared to the baseline contraction frequency, and simultaneously calculate the increase in the target person's real-time heart rate compared to the baseline heart rate; the target person is determined to be in a non-calm state when any of the following conditions are met: Condition 1: The increase in the real-time contraction frequency reaches or exceeds the increase threshold, and the duration of the real-time contraction frequency being above the critical level reaches or exceeds the duration threshold; Condition 2: The increase in the real-time heart rate reaches or exceeds the increase threshold; if none of the above conditions are met, the target person is determined to be in the calm state.

[0008] Furthermore, step S1 includes: Step S1-1: Collect facial physiological dynamic data of several people in a calm state. The facial physiological dynamics include the motion dynamic data of the target facial muscle groups and the photoplethysmography (PPG) signal of the face. Step S1-2: Sample at fixed time intervals to obtain the set of sample values ​​of the contraction frequency of the target muscle group for the i-th person. ,in This represents the contraction frequency value of the first time interval for the i-th person. This represents the contraction frequency value of the second time interval for the i-th person. This represents the contraction frequency value of the i-th person during the n-th time interval; it also yields the set of heart rate sample values ​​for the i-th person. ,in This represents the heart rate value of the i-th person in the first time interval. This represents the heart rate value of the i-th person in the second time interval. This represents the heart rate value of the i-th person at the n-th time interval; Step S1-3: For the i-th person, calculate the set of contraction frequency sample values ​​for the i-th person. The arithmetic mean of the values ​​is used as the baseline contraction frequency for the i-th person. ; Calculate the set of heart rate sample values ​​for the i-th person. The arithmetic mean of the values ​​is used as the baseline heart rate for the i-th person. ; Steps S1-4: Calculate the overall baseline systolic frequency based on the baseline systolic frequency and baseline heart rate of all N individuals. The overall baseline heart rate was calculated. .

[0009] Furthermore, step S2 includes: Step S2-1: Based on the sample values ​​of systolic frequency and heart rate of all personnel in the resting state, calculate the standard deviation of the sample values ​​of systolic frequency. and the standard deviation of sample values ​​of heart rate Wherein, the overall standard deviation The calculation formula is: Where M is the total number of sample values ​​for the contraction frequency of all personnel. This represents the j-th contraction frequency sample value. The reference shrinkage frequency; the overall standard deviation The calculation formula is: Where M is the total number of heart rate samples from all individuals. Represents the j-th heart rate sample value. The reference heart rate; Step S2-2: Based on the reference contraction frequency and the overall standard deviation The increase threshold is calculated. The calculation formula is: , where k is a preset amplification coefficient; Step S2-3: Based on the aforementioned baseline heart rate and the overall standard deviation The enhancement threshold is calculated. The calculation formula is as follows: , where l is the preset enhancement coefficient; Step S2-4: Based on the facial physiological dynamic data in the calm state, calculate the duration for which the contraction frequency of the target muscle group exceeds the baseline contraction frequency for all individuals. After sorting the durations by numerical value, take the 95th percentile as the duration threshold. .

[0010] Furthermore, step S3 includes: Step S3-1: Collect real-time facial physiological dynamic data of the target person, including real-time motion dynamic data of the facial target muscle groups and real-time photoplethysmography (PPG) signals of the face. Step S3-2: Extract the real-time contraction frequency of the target muscle group from the collected real-time motion dynamic data of the facial target muscle group. And calculate that the frequency value continues to exceed the critical value. The continuous time, as the duration , where the critical value Real-time heart rate was extracted from the real-time photoplethysmography (PPG) signals collected from the face. .

[0011] Furthermore, step S4 includes: Step S4-1: Calculate the real-time percentage increase in muscle contraction frequency. The formula is: ; Calculate the real-time percentage increase in heart rate The formula is: ; Step S4-2: The target person's emotional state is determined to be non-calm when any of the following conditions are met: Condition 1: The real-time percentage increase in the muscle contraction frequency. Reaching or exceeding the aforementioned increase threshold And the duration Reaching or exceeding the duration threshold ; Condition 2: The real-time increase percentage of the heart rate Reaching or exceeding the enhancement threshold ; If neither condition one nor condition two is met, the state is determined to be calm.

[0012] The system includes: a baseline establishment module, a threshold calculation module, a real-time monitoring module, and a status recognition module; The baseline establishment module is used to collect facial physiological dynamic data of several people in a calm state, and calculate the baseline contraction frequency and baseline heart rate based on the data. The threshold calculation module is used to analyze the changing characteristics in the facial physiological dynamic data and calculate the amplification threshold and enhancement threshold for state recognition. The real-time monitoring module is used to collect real-time physiological dynamic data of the target person's face, extract the real-time contraction frequency and the duration of exceeding the critical contraction frequency, and extract the real-time heart rate. The status recognition module is used to calculate the increase in real-time contraction frequency compared to the baseline contraction frequency and the increase in real-time heart rate compared to the baseline heart rate, and to determine the emotional state of the target person based on preset conditions.

[0013] The benchmark establishment module includes: The data acquisition unit is used to collect dynamic facial physiological data of several people in a calm state. The sample processing unit is used to sample and obtain a set of systolic frequency sample values ​​and a set of heart rate sample values ​​for each person at fixed time intervals. Individual baseline calculation unit, used to calculate individual baseline systolic frequency and individual baseline heart rate for each individual; The overall baseline calculation unit is used to calculate the overall baseline systolic frequency and overall baseline heart rate based on the individual baseline values ​​of all personnel.

[0014] The threshold calculation module includes: The standard deviation calculation unit is used to calculate the standard deviation of systolic frequency sample values ​​and the standard deviation of heart rate sample values. An amplification threshold calculation unit is used to calculate the amplification threshold based on the baseline contraction frequency and the standard deviation of the contraction frequency sample values. An enhanced threshold calculation unit is used to calculate the enhanced threshold based on the baseline heart rate and the standard deviation of heart rate sample values; Duration threshold calculation unit, used to calculate duration threshold based on calm state data.

[0015] The real-time monitoring module includes: The real-time data acquisition unit is used to collect real-time physiological dynamic data of the target person's face. The feature extraction unit is used to extract the real-time contraction frequency, calculate the duration exceeding the critical contraction frequency, and extract the real-time heart rate.

[0016] The status recognition module includes: The deviation calculation unit is used to calculate the percentage increase in real-time contraction frequency and the percentage increase in real-time heart rate; The state determination unit is used to determine the emotional state of the target person based on the comparison results of the increase percentage, enhancement percentage and corresponding threshold.

[0017] Compared with the prior art, the beneficial effects of the present invention are: 1. A multimodal physiological feature analysis system was constructed by fusing dynamic data of facial muscle group movement with photoplethysmography (PPG) pulse wave signals. This approach overcomes the limitations of single-modal data and effectively overcomes the shortcomings of traditional facial expression analysis, which is susceptible to interference from subjective concealment behaviors, through a dual verification mechanism of increased muscle contraction frequency and enhanced heart rate, thereby substantially improving the accuracy of emotion recognition.

[0018] 2. The baseline and adaptive threshold calculation model, established based on group calm state data, solves the problem of poor applicability of traditional fixed threshold methods across different individuals. By introducing standard deviation statistical analysis and personalized threshold settings, the system can automatically adapt to the physiological differences of different users, significantly improving the stability and reliability of identification in different populations.

[0019] 3. The non-contact data collection method greatly enhances the user experience, making long-term, continuous emotion monitoring possible. Simultaneously, by setting duration thresholds and critical level judgment mechanisms, the system can effectively distinguish between instantaneous physiological fluctuations and genuine emotional state changes, demonstrating higher practical value and reliability in real-world application scenarios and providing effective technical support for fields such as mental health monitoring and intelligent human-computer interaction. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of the structure of the emotion recognition system based on a large model according to the present invention; Figure 2 This is a schematic diagram of the threshold setting module of the large-model-based emotion recognition system of the present invention; Figure 3 This is a flowchart illustrating the emotion recognition method based on a large model according to the present invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] Example: Figure 1 , Figure 2 and Figure 3 As shown, this invention provides a technical solution: an emotion recognition system and method based on a large model. The methods include: Step S1: Collect facial physiological dynamic data of several individuals in a calm state; the facial physiological dynamic data includes the motion dynamic data of the target facial muscle groups and the photoplethysmography (PPG) signal of the face; wherein the target facial muscle groups include the orbicularis oculi, zygomaticus major, depressor supercilii, and corrugator supercilii; based on the facial physiological dynamic data, calculate the baseline contraction frequency and baseline heart rate. Step S2: By analyzing the change characteristics in the facial physiological dynamic data, calculate the amplification threshold and enhancement threshold used for state recognition; Step S3: Select the person whose emotion is to be identified as the target person, collect the real-time physiological dynamic data of the target person's face, including the real-time motion dynamic data of the target facial muscle group and the real-time photoplethysmography (PPG) wave signal of the face; and extract the real-time contraction frequency of the target muscle group and the duration of the real-time contraction frequency exceeding a certain critical contraction frequency, and extract the real-time heart rate from the real-time PPG wave signal. Step S4: Calculate the increase in the target person's real-time contraction frequency compared to the baseline contraction frequency, and simultaneously calculate the increase in the target person's real-time heart rate compared to the baseline heart rate; the target person is determined to be in a non-calm state when any of the following conditions are met: Condition 1: The increase in the real-time contraction frequency reaches or exceeds the increase threshold, and the duration of the real-time contraction frequency being above the critical level reaches or exceeds the duration threshold; Condition 2: The increase in the real-time heart rate reaches or exceeds the increase threshold; if none of the above conditions are met, the target person is determined to be in the calm state.

[0023] Furthermore, step S1 includes: Step S1-1: Collect facial physiological dynamic data of 50 test subjects in a resting state. A 16-channel surface electromyography (EMG) sensor was used to collect dynamic motion data of the corrugator supercilii and orbicularis oculi muscles; simultaneously, a near-infrared camera was used to collect photoplethysmography (PPG) signals of the face.

[0024] Step S1-2: Resample the system at 1-second intervals to obtain a sample set for each person. Using the data from 6 consecutive time intervals of the 3rd tester: Muscle contraction frequency sample value set Hz; Heart rate sample value set 70 beats / minute; 70 beats / minute means that the heart rate of the third test subject was calculated to be 70 beats / minute by analyzing the photoplethysmography signal on his face during the first 1-second time interval. Step S1-3: Calculate the baseline value for the 3rd tester: Reference shrinkage frequency Hz; Baseline heart rate times / minute; Steps S1-4: Summarize the individual baseline values ​​of all 50 testers and calculate the overall baseline value; Overall contraction frequency: Hz; Overall heart rate: times / minute; Overall reference shrinkage frequency Hz; Overall baseline heart rate times per minute.

[0025] Furthermore, step S2 includes: Step S2-1: Based on the sample values ​​of systolic frequency and heart rate of all personnel in a resting state, calculate the standard deviation of the sample values ​​respectively: Standard deviation of sample values ​​of shrinkage frequency The calculation formula is: Hz; Standard deviation of heart rate sample values The calculation formula is: times / minute; Step S2-2: Set k=2.5 and calculate the amplification threshold. : ; Step S2-3: Set l=2.5 and calculate the enhancement threshold. : ; Step S2-4: Record 320 instances of muscle contraction frequency exceeding the baseline value for all participants in a resting state, and determine the duration of each instance in the sample set. The duration threshold is calculated by taking the 95th percentile after sorting the values ​​by size. Second.

[0026] Furthermore, step S3 includes: Step S3-1: Collect real-time physiological dynamic data of the target person's face using the same sensor configuration and sampling parameters; Step S3-2: Critical contraction frequency: Hz; The real-time contraction frequency is calculated based on the electromyographic signals within the current time window: Hz; Duration : Second; Real-time heart rate times per minute.

[0027] Furthermore, step S4 includes: Step S4-1: Percentage increase in contraction frequency: ; Heart rate increase percentage: ; Step S5-2: State determination: Condition one: ,and Second Seconds, the condition is met; Condition two: The conditions are met; Since conditions one and two are met, the target person is determined to be in a non-calm state.

[0028] The system includes: a baseline establishment module, a threshold calculation module, a real-time monitoring module, and a status recognition module; The baseline establishment module is used to collect facial physiological dynamic data of several people in a calm state, and calculate the baseline contraction frequency and baseline heart rate based on the data. The threshold calculation module is used to analyze the changing characteristics in the facial physiological dynamic data and calculate the amplification threshold and enhancement threshold for state recognition. The real-time monitoring module is used to collect real-time physiological dynamic data of the target person's face, extract the real-time contraction frequency and the duration of exceeding the critical contraction frequency, and extract the real-time heart rate. The status recognition module is used to calculate the increase in real-time contraction frequency compared to the baseline contraction frequency and the increase in real-time heart rate compared to the baseline heart rate, and to determine the emotional state of the target person based on preset conditions.

[0029] The benchmark establishment module includes: The data acquisition unit is used to collect dynamic facial physiological data of several people in a calm state. The sample processing unit is used to sample and obtain a set of systolic frequency sample values ​​and a set of heart rate sample values ​​for each person at fixed time intervals. Individual baseline calculation unit, used to calculate individual baseline systolic frequency and individual baseline heart rate for each individual; The overall baseline calculation unit is used to calculate the overall baseline systolic frequency and overall baseline heart rate based on the individual baseline values ​​of all personnel.

[0030] The threshold calculation module includes: The standard deviation calculation unit is used to calculate the standard deviation of systolic frequency sample values ​​and the standard deviation of heart rate sample values. An amplification threshold calculation unit is used to calculate the amplification threshold based on the baseline contraction frequency and the standard deviation of the contraction frequency sample values. An enhanced threshold calculation unit is used to calculate the enhanced threshold based on the baseline heart rate and the standard deviation of heart rate sample values; Duration threshold calculation unit, used to calculate duration threshold based on calm state data.

[0031] The real-time monitoring module includes: The real-time data acquisition unit is used to collect real-time physiological dynamic data of the target person's face. The feature extraction unit is used to extract the real-time contraction frequency, calculate the duration exceeding the critical contraction frequency, and extract the real-time heart rate.

[0032] The status recognition module includes: The deviation calculation unit is used to calculate the percentage increase in real-time contraction frequency and the percentage increase in real-time heart rate; The state determination unit is used to determine the emotional state of the target person based on the comparison results of the increase percentage, enhancement percentage and corresponding threshold.

[0033] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. A large-model-based emotion recognition method, characterized by: Includes the following steps: Step S1: Collect facial physiological dynamic data of several individuals in a calm state; the facial physiological dynamic data includes the motion dynamic data of the target facial muscle groups and the photoplethysmography (PPG) signal of the face; wherein the target facial muscle groups include the orbicularis oculi, zygomaticus major, depressor supercilii, and corrugator supercilii; based on the facial physiological dynamic data, calculate the baseline contraction frequency and baseline heart rate. Step S2: By analyzing the change characteristics in the facial physiological dynamic data, calculate the amplification threshold and enhancement threshold used for state recognition; Step S3: Select the person whose emotion is to be identified as the target person, collect the real-time physiological dynamic data of the target person's face, including the real-time motion dynamic data of the target facial muscle group and the real-time photoplethysmography (PPG) wave signal of the face; and extract the real-time contraction frequency of the target muscle group and the duration of the real-time contraction frequency exceeding a certain critical contraction frequency, and extract the real-time heart rate from the real-time PPG wave signal. Step S4: Calculate the increase in the target person's real-time contraction frequency compared to the baseline contraction frequency, and simultaneously calculate the increase in the target person's real-time heart rate compared to the baseline heart rate; the target person is determined to be in a non-calm state when any of the following conditions are met: Condition 1: The increase in the real-time contraction frequency reaches or exceeds the increase threshold, and the duration of the real-time contraction frequency being above the critical level reaches or exceeds the duration threshold; Condition 2: The increase in the real-time heart rate reaches or exceeds the increase threshold; if none of the above conditions are met, the target person is determined to be in the calm state.

2. The emotion recognition method based on a large model according to claim 1, characterized in that: Step S1 includes: Step S1-1: Collect facial physiological dynamic data of several people in a calm state. The facial physiological dynamics include the motion dynamic data of the target facial muscle groups and the photoplethysmography (PPG) signal of the face. Step S1-2: Sample at fixed time intervals to obtain the set of sample values ​​of the contraction frequency of the target muscle group for the i-th person. ,in This represents the contraction frequency value of the first time interval for the i-th person. This represents the contraction frequency value of the second time interval for the i-th person. This represents the contraction frequency value of the i-th person during the n-th time interval; it also yields the set of heart rate sample values ​​for the i-th person. ,in This represents the heart rate value of the i-th person in the first time interval. This represents the heart rate value of the i-th person in the second time interval. This represents the heart rate value of the i-th person at the n-th time interval; Step S1-3: For the i-th person, calculate the set of contraction frequency sample values ​​for the i-th person. The arithmetic mean of the values ​​is used as the baseline contraction frequency for the i-th person. ; Calculate the set of heart rate sample values ​​for the i-th person. The arithmetic mean of the values ​​is used as the baseline heart rate for the i-th person. ; Steps S1-4: Calculate the overall baseline systolic frequency based on the baseline systolic frequency and baseline heart rate of all N individuals. The overall baseline heart rate was calculated. .

3. The emotion recognition method based on a large model according to claim 2, characterized in that: Step S2 includes: Step S2-1: Based on the sample values ​​of systolic frequency and heart rate of all personnel in the resting state, calculate the standard deviation of the sample values ​​of systolic frequency. and the standard deviation of sample values ​​of heart rate ; Step S2-2: Based on the reference contraction frequency and the overall standard deviation The increase threshold is calculated. The calculation formula is: , where k is a preset amplification coefficient; Step S2-3: Based on the aforementioned baseline heart rate and the overall standard deviation The enhancement threshold is calculated. The calculation formula is as follows: , where l is the preset enhancement coefficient; Step S2-4: Based on the facial physiological dynamic data in the calm state, calculate the duration for which the contraction frequency of the target muscle group exceeds the baseline contraction frequency for all individuals. After sorting the durations by numerical value, take the 95th percentile as the duration threshold. .

4. The emotion recognition method based on a large model according to claim 3, characterized in that: Step S3 includes: Step S3-1: Collect real-time facial physiological dynamic data of the target person, including real-time motion dynamic data of the facial target muscle groups and real-time photoplethysmography (PPG) signals of the face. Step S3-2: Extract the real-time contraction frequency of the target muscle group from the collected real-time motion dynamic data of the facial target muscle group. And calculate that the frequency value continues to exceed the critical value. The continuous time, as the duration , where the critical value Real-time heart rate was extracted from the real-time photoplethysmography (PPG) signals collected from the face. .

5. The emotion recognition method based on a large model according to claim 4, characterized in that: Step S4 includes: Step S4-1: Calculate the real-time percentage increase in muscle contraction frequency. The formula is: ; Calculate the real-time percentage increase in heart rate The formula is: ; Step S4-2: The target person's emotional state is determined to be non-calm when any of the following conditions are met: Condition 1: The real-time percentage increase in the muscle contraction frequency. Reaching or exceeding the aforementioned increase threshold And the duration Reaching or exceeding the duration threshold ; Condition 2: The real-time increase percentage of the heart rate Reaching or exceeding the enhancement threshold ; If neither condition one nor condition two is met, the state is determined to be calm.

6. A large-model-based emotion recognition system, used to execute the large-model-based emotion recognition method according to any one of claims 1-5, characterized in that: The system includes: The module includes a baseline establishment module, a threshold calculation module, a real-time monitoring module, and a status recognition module. The baseline establishment module is used to collect facial physiological dynamic data of several people in a calm state, and calculate the baseline contraction frequency and baseline heart rate based on the data. The threshold calculation module is used to analyze the changing characteristics in the facial physiological dynamic data and calculate the amplification threshold and enhancement threshold for state recognition. The real-time monitoring module is used to collect real-time physiological dynamic data of the target person's face, extract the real-time contraction frequency and the duration of exceeding the critical contraction frequency, and extract the real-time heart rate. The status recognition module is used to calculate the increase in real-time contraction frequency compared to the baseline contraction frequency and the increase in real-time heart rate compared to the baseline heart rate, and to determine the emotional state of the target person based on preset conditions.

7. The emotion recognition system based on a large model according to claim 6, characterized in that: The benchmark establishment module includes: The data acquisition unit is used to collect dynamic facial physiological data of several people in a calm state. The sample processing unit is used to sample and obtain a set of systolic frequency sample values ​​and a set of heart rate sample values ​​for each person at fixed time intervals. Individual baseline calculation unit, used to calculate individual baseline systolic frequency and individual baseline heart rate for each individual; The overall baseline calculation unit is used to calculate the overall baseline systolic frequency and overall baseline heart rate based on the individual baseline values ​​of all personnel.

8. The emotion recognition system based on a large model according to claim 7, characterized in that: The threshold calculation module includes: The standard deviation calculation unit is used to calculate the standard deviation of systolic frequency sample values ​​and the standard deviation of heart rate sample values. An amplification threshold calculation unit is used to calculate the amplification threshold based on the baseline contraction frequency and the standard deviation of the contraction frequency sample values. An enhanced threshold calculation unit is used to calculate the enhanced threshold based on the baseline heart rate and the standard deviation of heart rate sample values; Duration threshold calculation unit, used to calculate duration threshold based on calm state data.

9. The emotion recognition system based on a large model according to claim 8, characterized in that: The real-time monitoring module includes: The real-time data acquisition unit is used to collect real-time physiological dynamic data of the target person's face. The feature extraction unit is used to extract the real-time contraction frequency, calculate the duration exceeding the critical contraction frequency, and extract the real-time heart rate.

10. The emotion recognition system based on a large model according to claim 9, characterized in that: The status recognition module includes: The deviation calculation unit is used to calculate the percentage increase in real-time contraction frequency and the percentage increase in real-time heart rate; The state determination unit is used to determine the emotional state of the target person based on the comparison results of the increase percentage, enhancement percentage and corresponding threshold.

Images