Robot emotional interaction ethical boundary dynamic evaluation system and method

Abstract

The invention discloses a robot emotional interaction ethical boundary dynamic evaluation system and method, and relates to the technical field of robots, and the system comprises the following modules: a speech emotion analysis module which extracts fundamental frequency, energy, formant and speech speed acoustic parameters, extracts advanced features in combination with a Mel frequency cepstrum coefficient, constructs an end-to-end classification system through a recurrent neural network, and obtains a speech emotion analysis result; performing preliminary judgment on the emotion of the current user; the text semantic analysis module is used for analyzing emotional tendency and implicit emotions in a text through a natural language processing technology, and the user character feature collection module comprises a character character recent feature module, a character character long-term feature module and a character character real-time feature module. The user physique feature acquisition module comprises a character physique recent feature module, a character physique long-term feature module and a character physique real-time feature module, performs long-term, recent and real-time three-dimensional detection on characters and physiques of characters, and provides more accurate data for emotion interaction dynamic evaluation.

Description

Technical Field

[0001] The present invention relates to the technical field of robots, and particularly to a dynamic evaluation system and method for the ethical boundary of robot emotional interaction. Background Art

[0002] A robot is commonly known as an automatically controlled machine (Robot). An automatically controlled machine includes all machines that simulate human behavior or thoughts and other organisms (such as robot dogs, robot cats, etc.). There are many classification methods and controversies in the narrow definition of robots, and some computer programs are even called robots. In contemporary industry, a robot refers to an artificial machine device that can automatically execute tasks to replace or assist humans in work. An ideal high-fidelity robot is the product of the advanced integration of cybernetics, mechatronics, computer and artificial intelligence, materials science, and bionics. Currently, the scientific community is researching and developing in this direction.

[0003] In the existing robot emotional interaction system, according to CN107480291B - Emotional Interaction Method and Robot System Based on Humor Generation, it is recorded that "when the humor level of the robot can be output externally, the robot can directly output the humor level of the recognized user speech segment by recognizing the user speech segment. When the humor level of the robot is not directly output externally, the robot first vividly expresses the recognition result through emotional expression, and then the humor detection device indirectly estimates the humor level recognized by the robot based on the recognition result; the humor detection device judges the accuracy of humor recognition by comparing the recognized humor level with the true humor level. Through the accuracy, the robot's ability to recognize humor can be tested. If the accuracy reaches or exceeds the expected value, it means that the robot has strong humor recognition ability. If the accuracy does not reach the expected value, other existing humor recognition methods of the robot or improved humor recognition methods can be selected to conduct experiments again until the accuracy reaches the expected value, so that the robot has strong humor recognition ability". From this, it can be seen that the robot dynamically evaluates the ethical boundary of emotional interaction by recognizing the user speech segment.

[0004] According to CN107480291B - Emotional Interaction Method and Robot System Based on Humor Generation, it is recorded that "the present invention outputs by generating a humorous speech segment, triggers and changes the user's emotion, and monitors the emotion feedback by the user to the output humorous speech segment, that is, the effect of humor is tested through emotion recognition. The humorous speech segment corresponding to the preset emotion in the emotion recognition result feedback by the user is judged to have the characteristics of humor, and the humorous speech segment with the characteristics of humor and its related information are stored in the humor big data knowledge base. In the future, the humorous speech segment can be selected through the humor big data knowledge base, so as to improve the robot's ability to generate humor". From this, it can be seen that the robot adds a monitoring feedback function to evaluate the emotional interaction boundary.

[0005] However, there is a real problem that when the robot collects the user's data, it lacks the reference of historical data and makes judgments based on real-time data, which affects the accuracy of dynamic evaluation.

[0006] In summary, a dynamic evaluation system and method for the ethical boundary of robot emotional interaction are designed. Summary of the Invention

[0007] In order to overcome the above deficiencies, the present invention provides a dynamic evaluation system and method for the ethical boundary of robot emotional interaction.

[0008] The present invention realizes the above object through the following technical solutions:

[0009] A dynamic evaluation system for the ethical boundary of robot emotional interaction includes the following modules:

[0010] The voice emotion analysis module extracts acoustic parameters such as fundamental frequency, energy, formant, and speech rate, combines with Mel-frequency cepstral coefficients to extract high-level features, and then constructs an end-to-end classification system through a recurrent neural network to preliminarily judge the emotion of the user at that time;

[0011] The text semantic parsing module analyzes the emotional tendency and implicit emotion in the text through natural language processing technology, and simultaneously uses the BERT pre-trained model for context emotion reasoning;

[0012] The visual emotion recognition module captures visual information such as facial expressions and body movements through a camera to recognize emotions, decomposes facial muscle movements into action units, recognizes expressions through combination (such as frowning corresponding to anger), then processes the image with a convolutional neural network, extracts expression features, and then makes an emotion judgment;

[0013] The physiological information monitoring module infers emotions by collecting heart rate, skin conductance, and electroencephalogram physiological signals through wearable devices, extracts parameters of heart rate variability and skin conductance response, then denoises through wavelet transform and extracts time-frequency features, and uses a support vector machine model to establish a mapping relationship between physiological signals and emotions;

[0014] The big data information collection module includes a user personality feature collection module and a user physical feature collection module. First, the user identity information is confirmed, and then the relevant information about the user's personality and physical characteristics is collected through the local server and the cloud server;

[0015] The emotional interaction dynamic evaluation module makes an emotional dynamic evaluation of the user through the emotional ethics data judgment of the voice emotion analysis module, the text semantic parsing module, the visual emotion recognition module, and the physiological information monitoring module, plus the data collection of the big data information collection module.

[0016] Preferably, the user personality feature acquisition module includes a recent personality feature module, a long-term personality feature module, and a real-time personality feature module. The recent personality feature module collects information on the user's recent social media dynamics, consumption record graph, and motion sensing data through a cloud server as recent personality feature data. The long-term personality feature module collects information on the user's living area, birth area, and family characteristics through a cloud server as long-term personality feature data. The real-time personality feature module collects the real-time action state, real-time medical record, and current environmental state through a cloud server as real-time personality feature data.

[0017] Preferably, in the real-time personality feature module, the cloud server collects the real-time action state of the person by calling the networked devices around the person's location. For example, through the hotel where the person is located, data is collected from the intelligent devices in use. When the curtains are closed and the bathroom light and exhaust fan are turned on, the person is in a relaxed state. When the TV is turned on and the air conditioner is set to a suitable temperature, the person is in a comfortable state. The cloud server queries the recent medical records of the person, such as the frequency of medical treatment and the condition, by calling the public server of the local hospital to measure the real-time personality state of the person. The cloud server monitors the environment where the person is located by calling the surrounding networked monitoring devices. If the environment where the person is located is noisy, the person's real-time personality will be more anxious. If the environment where the person is located is quiet, the person's real-time personality will be relatively calm.

[0018] Preferably, the user physical feature acquisition module includes a recent physical feature module, a long-term physical feature module, and a real-time physical feature module. The recent physical feature module collects information on the user's diet, work information, and climate environment information through a cloud server as the recent physical feature module. The long-term physical feature module collects information on the user's age characteristics, genetic information, and living habits information through a cloud server as the long-term physical feature module. The real-time physical feature module collects the person's real-time medical record, real-time environmental conditions (such as the current location, the impact of high altitude areas on the physical condition of people living in plain areas), and real-time stress state (by measuring the current psychological stress state of the person in real time through intelligent wearables, thereby having a real-time impact on the physical condition).

[0019] A method for dynamically evaluating the ethical boundary of robot emotional interaction as described above includes the following steps:

[0020] Step 1: Speech emotion analysis. The speech emotion analysis module extracts fundamental frequency, energy, formants, and speech rate acoustic parameters, combines them with Mel-frequency cepstral coefficients to extract high-level features, and then constructs an end-to-end classification system through a recurrent neural network to make a preliminary judgment on the user's emotion at that time;

[0021] Step 2: Text semantic parsing. The text semantic parsing module analyzes the emotional tendency and implicit emotions in the text through natural language processing techniques, and at the same time uses the BERT pre-trained model for context emotion reasoning;

[0022] Step 3: Visual emotion recognition. The visual emotion recognition module captures facial expressions and limb movement visual information through a camera to recognize emotions, decomposes facial muscle movements into action units, recognizes expressions by combination (such as frowning corresponding to anger), then uses a convolutional neural network to process images, extracts expression features, and then makes an emotion judgment;

[0023] Step 4: Physiological information monitoring. The physiological information monitoring module infers emotions by collecting heart rate, skin conductance, and electroencephalogram physiological signals through wearable devices, extracts parameters of heart rate variability and skin conductance response, then denoises and extracts time-frequency features through wavelet transform, and uses a support vector machine model to establish a mapping relationship between physiological signals and emotions;

[0024] Step 5: Big data information collection. The user personality feature collection module and the user physical feature collection module first confirm the user's identity information, and then collect information related to the user's personality and physical characteristics through the local server and the cloud server;

[0025] Step 6: Emotional interaction dynamic assessment. The emotional interaction dynamic assessment module makes an emotional dynamic assessment of the user through the emotional ethics data judgment of the speech emotion analysis module, the text semantic parsing module, the visual emotion recognition module, and the physiological information monitoring module, plus the data collection of the big data information collection module.

[0026] Preferably, the evaluation algorithm in Step 6 is as follows:

[0027]

[0028] Where:

[0029] S is the speech emotion analysis value, and the value range is [0, 1];

[0030] T is the text semantic parsing value, and the value range is [-1, 1];

[0031] V is the visual emotion recognition value, and the value range is [0, 2π);

[0032] P is the physiological emotion analysis value, and the value range is [0, 100];

[0033] The basic weight vector, with a value range of [0, 4] and an initial value of [0.3, 0.25, 0.2, 0.25], satisfies Σw = 1 through L1 regularization constraint;

[0034] A m (x m ) is the weight modulation function of the influence quantity A (the calculation result of the data collected by the user character feature acquisition module);

[0035] B m (x m ) is the weight modulation function of the influence quantity B (the calculation result of the data collected by the user physical feature acquisition module);

[0036] f m (x m ) is the parameter standardization function.

[0037] Preferably, the interval values of the speech emotion analysis value S correspond to different weights of the influence quantity A and the weight of the influence quantity B.

[0038] Preferably, the quantization calculation formula of the influence quantity B is as follows:

[0039]

[0040] Wherein, Automatically discovered through density clustering, Optimally determined through ridge regression.

[0041] Preferably, the mathematical expression of the influence quantity A adopts a piecewise linear modulation model based on neuroscience.

[0042] The beneficial effects of the present invention are as follows: In the robot emotional interaction ethical boundary dynamic evaluation system and method:

[0043] 1. The user character feature acquisition module includes a character recent character feature module, a character long-term character feature module, and a character real-time character feature module, and the user physical feature acquisition module includes a person physical recent feature module, a person physical long-term feature module, and a person physical real-time feature module, so as to be able to detect the character and physical characteristics of a person in three dimensions of long-term, recent, and real-time, providing more accurate data for the dynamic evaluation of emotional interaction;

[0044] 2. In the character real-time character feature module, the cloud server collects the real-time action state of the person by calling the networked devices around the address where the person is located, realizing the connection between the system and the external intelligent system;

[0045] 3. The emotional interaction dynamic evaluation module conducts an emotional dynamic evaluation of the user by judging the emotional and ethical data of the speech emotion analysis module, text semantic parsing module, visual emotion recognition module, and physiological information monitoring module, and adding the data collection of the big data information collection module. After integrating the four individual emotional evaluation data, and then making it more in line with the actual situation of the person in terms of personality and physique, a more accurate dynamic evaluation of emotional interaction is carried out. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] The present invention will be described by way of examples with reference to the accompanying drawings, wherein:

[0047] Figure 1 is the system schematic diagram of the present invention;

[0048] Figure 2 is the method step diagram of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0049] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are all simplified schematic diagrams, only showing the basic structure of the present invention in a schematic manner, so they only show the components related to the present invention.

[0050] As Figure 1 shown, a robot emotional interaction ethical boundary dynamic evaluation system includes the following modules:

[0051] The speech emotion analysis module extracts acoustic parameters such as fundamental frequency, energy, formant, and speech rate, combines them with Mel frequency cepstral coefficients to extract high-level features, and then constructs an end-to-end classification system through a recurrent neural network to preliminarily judge the emotion of the user at that time;

[0052] The text semantic parsing module analyzes the emotional tendency and implicit emotions in the text through natural language processing technology, and at the same time uses the BERT pre-trained model for context emotional reasoning;

[0053] The visual emotion recognition module captures facial expressions and limb movement visual information through a camera to recognize emotions, decomposes facial muscle movements into action units, recognizes expressions by combination (such as frowning corresponding to anger), then processes the image using a convolutional neural network, extracts expression features, and then makes an emotional judgment;

[0054] The physiological information monitoring module infers emotions by collecting heart rate, skin conductance, and electroencephalogram physiological signals through wearable devices, extracts parameters of heart rate variability and skin conductance response, then denoises and extracts time-frequency features through wavelet transform, and uses a support vector machine model to establish a mapping relationship between physiological signals and emotions;

[0055] The big data information collection module, including the user personality trait collection module and the user physical trait collection module, first confirms the user identity information, and then collects information related to the user's personality and physical characteristics through the local server and the cloud server;

[0056] The emotional interaction dynamic evaluation module conducts an emotional dynamic evaluation of the user through the judgment of emotional ethics data by the voice emotion analysis module, the text semantic parsing module, the visual emotion recognition module, and the physiological information monitoring module, combined with the data collection of the big data information collection module.

[0057] Specifically, the user personality trait collection module includes the recent personality trait module of the person, the long-term personality trait module of the person, and the real-time personality trait module of the person. The recent personality trait module of the person collects information on the user's recent social media dynamics, consumption record graph, and motion sensing data through the cloud server as the recent personality trait data of the person; the long-term personality trait module of the person collects information on the user's living area, birth area, and family characteristics through the cloud server as the long-term personality trait data of the person; the real-time personality trait module of the person collects the real-time action state, real-time medical record, and current environmental state through the cloud server as the real-time personality trait data of the person.

[0058] Specifically, in the real-time personality trait module of the person, the cloud server collects the real-time action state of the person by calling the networked devices around the address where the person is located. For example, through the hotel where the person is located, data is collected from the intelligent devices in use. When the curtains are closed and the bathroom light and exhaust fan are turned on, the person is in a relaxed state. When the TV is turned on and the air conditioner is set to a suitable temperature, the person is in a comfortable state. The cloud server queries the recent medical records of the person, such as the frequency of medical treatment, the condition, etc., by calling the public server of the local hospital to measure the real-time personality state of the person. The cloud server monitors the environment where the person is located by calling the surrounding networked monitoring devices. If it is detected that the environment where the person is located is noisy, the real-time personality of the person will be more anxious. If it is detected that the environment where the person is located is quiet, the real-time personality of the person will be relatively calm.

[0059] Specifically, the user physical characteristic acquisition module includes a recent physical characteristic module of the person, a long-term physical characteristic module of the person, and a real-time physical characteristic module of the person. The recent physical characteristic module of the person collects the user's diet and catering information, the person's work information, and the climate and environmental information through the cloud server as the recent physical characteristic module of the person. The long-term physical characteristic module of the person collects the user's age characteristic information, genetic gene information, and living routine information through the cloud server as the long-term physical characteristic module of the person. The real-time physical characteristic module of the person collects the person's real-time medical records, real-time environmental conditions (such as the current location, the impact of the plateau area on the physical characteristics of people living in the plain area), and real-time stress status (intelligently measure the current psychological stress status of the person through wearable devices, so as to cause real-time impact on the physical characteristics) through the cloud server.

[0060] As Figure 2 shown, a method for dynamically evaluating the ethical boundary of robot emotional interaction includes the following steps:

[0061] Step 1: Speech emotion analysis. The speech emotion analysis module extracts acoustic parameters such as fundamental frequency, energy, formant, and speech rate, combines them with Mel-frequency cepstral coefficients to extract high-level features, and then constructs an end-to-end classification system through a recurrent neural network to make a preliminary judgment on the emotion of the user at that time.

[0062] Step 2: Text semantic parsing. The text semantic parsing module analyzes the emotional tendency and implicit emotions in the text through natural language processing technology, and at the same time uses the BERT pre-trained model for context emotion reasoning.

[0063] Step 3: Visual emotion recognition. The visual emotion recognition module captures facial expressions and limb movement visual information through a camera to recognize emotions, decomposes facial muscle movements into action units, recognizes expressions by combination (such as frowning corresponding to anger), then processes the image using a convolutional neural network, extracts expression features, and then makes an emotion judgment.

[0064] Step 4: Physiological information monitoring. The physiological information monitoring module collects heart rate, skin conductance, and electroencephalogram physiological signals through wearable devices to infer emotions, extracts parameters of heart rate variability and skin conductance response, then denoises through wavelet transform and extracts time-frequency features, and uses a support vector machine model to establish the mapping relationship between physiological signals and emotions.

[0065] Step 5: Big data information collection. The user personality characteristic acquisition module and the user physical characteristic acquisition module first confirm the user's identity information, and then collect the user's personality and physical related information through the local server and the cloud server.

[0066] Step 6: Emotional interaction dynamic assessment. The emotional interaction dynamic assessment module judges through the emotional ethics data of the speech emotion analysis module, text semantic analysis module, visual emotion recognition module, and physiological information monitoring module, and combined with the data collection of the big data information collection module, conducts dynamic emotional assessment on the user.

[0067] Specifically, the evaluation algorithm for Step 6 is as follows:

[0068]

[0069] Where:

[0070] S is the speech emotion analysis value, and its value range is [0, 1];

[0071] T is the text semantic analysis value, and its value range is [-1, 1];

[0072] V is the visual emotion recognition value, and its value range is [0, 2π);

[0073] P is the speech emotion analysis value, and its value range is [0, 100];

[0074] is the basic weight vector, and its value range is [0, 4], initial value: [0.3, 0.25, 0.2, 0.25], and it satisfies Σw = 1 through L1 regularization constraint;

[0075] A m (x m ) is the weight modulation function of the influence quantity A (the calculation result of the data collected by the user personality feature collection module);

[0076] B m (x m ) is the weight modulation function of the influence quantity B (the calculation result of the data collected by the user physical feature collection module);

[0077] f m (x m ) is the parameter normalization function.

[0078] Specifically, the interval values of the speech emotion analysis value S correspond to different weights of the influence quantity A and the influence quantity B.

[0079] Table of interval weights of speech emotion value S

[0080]

[0081] Specifically, the quantization calculation formula for the influence quantity B is as follows:

[0082]

[0083] Where, Automatically discovered through density clustering Determined through optimization by ridge regression

[0084] Specifically, the mathematical expression of the influence quantity A adopts a piecewise linear modulation model based on neuroscience

[0085] Based on the inspiration of the present invention, through the above description, relevant staff can completely make various changes and modifications without departing from the technical idea of this invention. The technical scope of this invention is not limited to the content in the specification, and its technical scope must be determined according to the scope of the claims

Claims

1. A dynamic evaluation system for the ethical boundaries of robot emotional interaction, characterized in that: It includes the following modules: The speech emotion analysis module extracts acoustic parameters such as fundamental frequency, energy, formant, and speech rate, combines them with Mel Frequency Cepstral Coefficients to extract high-level features, and then constructs an end-to-end classification system through a recurrent neural network to make a preliminary judgment on the emotion of the user at that time; The text semantic parsing module analyzes the emotional tendency and implicit emotions in the text through natural language processing technology, and at the same time uses the BERT pre-trained model for context emotion reasoning; The visual emotion recognition module captures visual information such as facial expressions and body movements through a camera to recognize emotions, decomposes facial muscle movements into action units, recognizes expressions through combination, then processes the images using a convolutional neural network, extracts expression features, and then makes an emotion judgment; The physiological information monitoring module infers emotions by collecting heart rate, skin conductance, and electroencephalogram physiological signals through wearable devices, extracts parameters of heart rate variability and skin conductance response, then denoises through wavelet transform and extracts time-frequency features, and uses a support vector machine model to establish a mapping relationship between physiological signals and emotions; The big data information collection module includes a user personality feature collection module and a user physical feature collection module. First, it confirms the user identity information, and then collects information related to the user's personality and physical characteristics through the local server and the cloud server; The emotion interaction dynamic evaluation module makes an emotion dynamic evaluation of the user through the emotion ethics data judgment of the speech emotion analysis module, the text semantic parsing module, the visual emotion recognition module, and the physiological information monitoring module, plus the data collection of the big data information collection module.

2. The dynamic evaluation system for the ethical boundary of robot emotional interaction according to claim 1, characterized in that: The user personality feature collection module includes a recent personality feature module of a person, a long-term personality feature module of a person, and a real-time personality feature module of a person. The recent personality feature module of a person collects information on the user's recent social media dynamics, consumption record graph, and motion sensing data through the cloud server as recent personality feature data of a person; the long-term personality feature module of a person collects information on the user's living area, birth area, and family characteristics through the cloud server as long-term personality feature data of a person; The real-time personality feature module of a person collects the real-time action state, real-time medical record, and current environmental state through the cloud server as real-time personality feature data of a person.

3. The dynamic evaluation system for the ethical boundary of robot emotional interaction according to claim 2, wherein: In the real-time personality feature module of a person, the cloud server collects the real-time action state of a person by calling the networked devices around the address where the person is located, the cloud server queries the recent medical records of the person by calling the local hospital public server, and the cloud server monitors the environment where the person is located by calling the surrounding networked monitoring devices at the address where the person is located.

4. The dynamic evaluation system for the ethical boundaries of robot emotional interaction according to claim 1, characterized in that: The user physical characteristic acquisition module includes a recent physical characteristic module of the person, a long-term physical characteristic module of the person, and a real-time physical characteristic module of the person. The recent physical characteristic module of the person collects the user's diet and catering information, the person's work information, and the climate and environment information through the cloud server as the recent physical characteristic module of the person. The long-term physical characteristic module of the person collects the user's age characteristic information, genetic gene information, and living schedule information through the cloud server as the long-term physical characteristic module of the person; The real-time physical characteristic module of the person collects the person's real-time medical records, real-time environmental conditions, and real-time stress status through the cloud server.

5. A method for dynamically evaluating the ethical boundaries of robot emotional interaction according to any one of claims 1-4, characterized in that: It includes the following steps: Step 1: Speech emotion analysis. The speech emotion analysis module extracts fundamental frequency, energy, formant, speech rate acoustic parameters, combines with Mel-frequency cepstral coefficients to extract high-level features, and then constructs an end-to-end classification system through a recurrent neural network to preliminarily judge the emotion of the user at that time; Step 2: Text semantic parsing. The text semantic parsing module analyzes the emotional tendency and implicit emotion in the text through natural language processing technology, and at the same time uses the BERT pre-trained model for context emotion reasoning; Step 3: Visual emotion recognition. The visual emotion recognition module captures facial expressions and limb movement visual information through a camera to recognize emotions, decomposes the facial muscle movement into action units, combines to recognize expressions, then uses a convolutional neural network to process images, extracts expression features, and then makes an emotion judgment; Step 4: Physiological information monitoring. The physiological information monitoring module infers emotions by collecting heart rate, skin conductance, and electroencephalogram physiological signals through wearable devices, extracts parameters of heart rate variability and skin conductance response, then denoises through wavelet transform and extracts time-frequency features, and uses a support vector machine model to establish a mapping relationship between physiological signals and emotions; Step 5: Big data information collection. The user personality characteristic acquisition module and the user physical characteristic acquisition module first confirm the user's identity information, and then collect the user's personality and physical related information through the local server and the cloud server; Step 6: Emotional interaction dynamic evaluation. The emotional interaction dynamic evaluation module makes an emotional dynamic evaluation of the user through the emotional ethics data judgment of the speech emotion analysis module, the text semantic parsing module, the visual emotion recognition module, and the physiological information monitoring module, plus the data collection of the big data information collection module.

6. The method for dynamically evaluating the ethical boundary of robot emotional interaction according to claim 5, characterized in that: The evaluation algorithm of the said Step 6 is as follows: Where: S is the speech emotion analysis value, and the value range is [0,1]; T is the text semantic parsing value, and the value range is [-1,1]; V is the visual emotion recognition value, and the value range is [0,2π); P is the speech emotion analysis value, and the value range is [0,100]; It is the basic weight vector, with a value range of [0, 4] and an initial value of [0.3, 0.25, 0.2, 0.25]. It satisfies Σw = 1 through L1 regularization constraint; A m (x m ) is the weight modulation function for the influence quantity A (the calculation result of the data collected by the user personality characteristic collection module); B m (x m ) is the weight modulation function of the influence quantity B (the calculation result of the data collected by the user physical characteristic collection module); f m (x m ) is a parameter normalization function.

7. The method for dynamically evaluating the ethical boundary of robot emotional interaction according to claim 6, wherein: The interval value of the said speech emotion analysis value S corresponds to different influence quantity A weights and influence quantity B weights.

8. The method for dynamically evaluating the ethical boundary of robot emotional interaction according to claim 6, characterized in that: The quantization calculation formula of the said influence quantity B is as follows: Among them, Automatically discovered through density clustering Determined through ridge regression optimization 9. The method for dynamically evaluating the ethical boundary of robot emotional interaction according to claim 6, characterized in that: The mathematical expression of the said influence quantity A adopts a piecewise linear modulation model based on neuroscience.

Images