Interaction method and device for skin management robot and medium

Abstract

The invention provides an interaction method and device for a skin management robot and a medium, and the method comprises the steps: receiving the voice information of a user, the voice information at least comprises the skin demand characteristics of the user and the feedback of the user, and uploading the voice information to a cloud server; receiving response information generated by the cloud server based on the voice information, the response information being generated by the following steps: converting the voice information into text information, analyzing the text information, and generating response information matched with the voice information; and converting the received response information into an audio signal and playing the audio signal. Voice information of a user can be accurately analyzed and matched response information can be generated by introducing a large language model, so that the intelligent level of interaction is greatly improved; according to the method, the historical interaction data of the user is fully utilized, the accuracy value is calculated, and the content of the skin management knowledge base is dynamically adjusted, so that continuous optimization of the knowledge base is realized, and the interaction accuracy is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of artificial intelligence, and particularly relates to an interaction method, device and medium for a skin management robot. Background Art

[0002] With the rapid development of artificial intelligence and Internet of Things technologies, the intelligent demand in the skin management field is increasing day by day. The traditional skin management methods mainly rely on users to consult materials by themselves or consult professionals, which is inefficient and lacks personalized services. In recent years, the maturity of artificial intelligence technology has provided new ideas for solving this problem. However, most of the current skin management devices on the market have single functions and are difficult to meet the diverse needs of users. For example, some devices can only provide basic skin care knowledge queries and cannot dynamically adjust the content according to the actual needs of users; they cannot make full use of the user's historical interaction data and understand the user's intention, resulting in inaccurate answer results and being unable to attract users, leaving a large room for improvement. Summary of the Invention

[0003] In order to solve the problems existing in the above-mentioned prior art, the present invention provides an interaction method for a skin management robot, including the following steps: Receiving the voice information of the user, the voice information at least includes the user's skin demand characteristics and user feedback, and uploading the voice information to the cloud server; Receiving the response information generated by the cloud server based on the voice information, and the response information is generated through the following steps: Converting the voice information into text information; Analyzing the text information based on a large language model and a skin management knowledge base, and generating response information matching the voice information; Converting the received response information into an audio signal and playing it.

[0004] On the basis of the above solution, the method further includes: When generating the response information, the large language model is not only based on the text information converted from the current user's voice information, but also based on the user's historical interaction data, and the historical interaction data at least includes multiple processed text information, response information and interaction duration, and the interaction duration is the time used for each completed interaction recorded.

[0005] On the basis of the above solution, it further includes: Sending an accuracy value to the large language model, and the accuracy value is calculated based on the historical interaction data, and the calculation formula is: , wherein, U is the accuracy,t i is the duration of playing response information for each i interaction, T is the total duration of the interaction, x i is the weight of each response, de is the user satisfaction of the interaction, i is a positive integer, n is the total number of all response information played in this interaction.

[0006] Based on the above solution, based on text information, response information, user feedback, and accuracy, the large language model dynamically adjusts the content of the skin management knowledge base, which at least includes skin physiological knowledge, skin problems, skin diseases, skin care knowledge, and skin care products.

[0007] Based on the above solution, the method further includes: Collect data of multiple user interactions and extract preference feature vectors from the data; Based on the preference feature vectors, construct a user behavior model through a clustering algorithm. The model associates the user's behavior pattern with skin demand characteristics and predicts the user's needs according to the behavior model; the user's behavior pattern refers to a series of regular or habitual behaviors shown by the user during multiple interactions, and the skin demand characteristics refer to specific needs or problems raised by the user in skin management; Predict the user's needs through the behavior model, convert the prediction result into an audio signal and play it.

[0008] The present invention also provides a skin management robot, which is characterized by including: A first receiving module, which receives the user's voice information, and the voice information at least includes the user's skin demand characteristics and user feedback; A sending module, which uploads the voice information to the cloud server; A second receiving module, which receives the response information generated by the cloud server based on the voice information. The response information is generated through the following steps: Convert the voice information into text information; Based on the large language model and the skin management knowledge base, analyze the text information and generate response information matching the voice information; A voice module, which converts the received response information into an audio signal and plays it.

[0009] On the basis of the above solution, it further includes an accuracy calculation module and a data transmission module. The accuracy calculation module is used to calculate an accuracy value based on historical interaction data, and the data transmission module is used to send the accuracy value to the large language model in the cloud server.

[0010] On the basis of the above solution, the sending module includes an array microphone and a microprocessor. The array microphone is used to collect the user's voice information, and the microprocessor preprocesses the voice information collected by the array microphone. The preprocessing at least includes noise reduction and signal enhancement.

[0011] On the basis of the above solution, the second receiving module includes an Internet of Things transmission module and a voice player. The Internet of Things transmission module transmits the text information to the microprocessor, the microprocessor converts the text information into an audio signal, and the voice player plays the audio signal.

[0012] The present invention also provides a computer-readable storage medium, which has a computer program. When the computer program is executed by a processor, it implements the steps of an interaction method for a skin management robot as described above.

[0013] Compared with the prior art, the present invention has the following beneficial effects: By introducing a large language model and a skin management knowledge base, it can accurately analyze the user's voice information and generate a matching response information, greatly improving the intelligent level of interaction; This method makes full use of the user's historical interaction data, including text information, response information, interaction duration, etc., calculates the accuracy value and dynamically adjusts the content of the knowledge base, so as to realize the continuous optimization of the knowledge base, and the matching degree between the answer scheme and the user's needs is higher; Through long-term interaction, the robot can continuously adapt to the user's needs according to the user's personalized habits and characteristic tendencies, become the user's exclusive skin management expert, and provide accurate and user-habit-compliant solutions in real time. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 It is a flowchart of the interaction method of this application; Figure 2 It is a schematic diagram of module interaction of this application; DETAILED DESCRIPTION OF THE EMBODIMENTS Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Instead, these embodiments are provided to more thoroughly and completely understand the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes and are not used to limit the protection scope of the present disclosure.

[0015] In the description of the embodiments of the present disclosure, the term "including" and its similar terms should be understood as open inclusion, that is, "including but not limited to". The term "based on" should be understood as "at least partially based on". The term "one embodiment" or "the embodiment" should be understood as "at least one embodiment". The terms "first", "second", etc. may refer to different or the same objects. There may also be other explicit and implicit definitions hereinafter.

[0016] The present invention provides an interaction method for a skin management robot, which generates personalized response information by combining user input, historical interaction data, and a large language model, and dynamically optimizes the content of the knowledge base to improve the accuracy and personalization of the interaction. As Figure 1 and Figure 2 shown, it specifically includes the following steps: Receive the voice information input by the user and upload the voice information to the cloud server; The robot receives the user's voice through a voice device and processes the voice information such as noise reduction and signal enhancement through a microprocessor to generate high-quality voice information data; The voice information includes the user's skin demand characteristics, and the user asks questions to the robot according to their own skin needs; the voice information uploads the voice data to the cloud server through the Internet of Things transmission module.

[0017] Receive the response information generated by the cloud server based on the voice information, and the response information is generated through the following steps: Convert the voice information into text information. Specifically, the cloud server calls a voice recognition engine to convert the voice into text information; based on the large language model and the skin management knowledge base, analyze the text information and generate a response information that matches the voice information.

[0018] Based on the response information, the voice information also includes the user's feedback on the response information. The robot answers the questions asked by the user according to the response information returned by the large language model; the robot converts the received response information into an audio signal and plays it to have a conversation with the user.

[0019] Based on the above method, first, it is necessary to build a skin management knowledge base in the large language model accessed by the cloud server, including but not limited to common skin problems, skin physiological knowledge, skin problem and disease knowledge, skin care product knowledge, beauty instrument knowledge, skin management operation process knowledge, customer management knowledge, industry trends and regulations knowledge, etc.

[0020] After the user wakes up the robot, the robot transmits the received user's skin demand problem to the large language model of the cloud server. The user's skin demand problem at least includes: basic user characteristics, such as age, height, weight and other information, the skin part of the user, the user's feelings, the current performance state of the skin, the duration of the skin state, the user's needs, and whether the user's situation is serious, recommending that they seek medical treatment in time, etc.

[0021] The large language model matches the user's skin demand problem with the skin management knowledge base, extracts the content with a high matching degree and transmits it to the robot. When generating the response information, the large language model is not only based on the text information converted from the current user's voice information, but also based on the user's historical interaction data. The historical interaction data at least includes multiple processed text information, response information and interaction duration. The interaction duration is the time used for each completed interaction recorded.

[0022] It has been proved by practice that the interaction realized by the above method already has a high degree of accuracy. In order to provide better services for users, this application introduces accuracy, and through the relearning of accuracy by the large model, the method has a higher accuracy, improving the user experience.

[0023] Specifically, the accuracy value is sent to the large language model. The accuracy value is calculated based on the historical interaction data, and the calculation formula is: , Among them, U is the accuracy, t i is the duration of each response information played in the i-th interaction, T is the total interaction duration, x i is the weight of each response, de is the user satisfaction of the interaction, i is a positive integer, n is the total number of all response information played in this interaction.

[0024] According to one of the embodiments: The user asks the robot: "My skin has been very dry recently", receives the user's voice message, and uploads it to the cloud server; after analysis, the cloud server returns a suggestion: "It is recommended to use a moisturizing lotion morning and evening every day and drink plenty of water to keep the body hydrated"; convert the suggestion into voice and play it to the user; after the user receives the suggestion, they ask: "Which moisturizing lotion with what ingredients is better to use", the cloud server analyzes and responds: "Lotions containing hyaluronic acid are better. Do you need me to recommend products for you?", and conduct subsequent interactions according to the user's answer. If necessary, comprehensive recommendations can be made based on the stored age and skin condition of the user.

[0025] After ending this interaction, the user gives a satisfaction level. According to this satisfaction level, t1 is the duration of the first played response message, t2 is the duration of the second played response message, T is the total duration of this interaction, x1 is the weight of the first response, and x2 is the weight of the second response. Then, according to this interaction, the accuracy can be calculated as follows: 。

[0026] Transfer the calculated accuracy value to the large language model. The large language model comprehensively learns data such as historical interaction data and accuracy, further learns, and updates the skin management knowledge base to improve the accuracy of subsequent interactions. At the same time, summarize the historical interaction data, and summarize and analyze the personalized demand characteristics of users. Therefore, in the continuous interaction process with this user, a more accurate customized skin management robot can be provided for the user, providing a more accurate and personalized solution. In actual use, it is proved that the learning of accuracy can further improve the accuracy of answering questions.

[0027] Based on the text information, response information, user feedback, and accuracy, the large language model dynamically adjusts the content of the skin management knowledge base. Through multiple interactions, a specific skin management knowledge base for the user's needs can be formed, effectively improving the user experience.

[0028] The method also includes constructing a user behavior model. Based on the user's behavior patterns and skin demand characteristics, predict their potential needs and provide relevant suggestions in advance, specifically including: Collect data from multiple user interactions and extract preference feature vectors from the data; Based on the preference feature vectors, construct a user behavior model through a clustering algorithm. The model associates the user's behavior patterns with skin demand characteristics and predicts the user's needs according to the behavior model; the user's behavior patterns refer to a series of regular or habitual behaviors shown by the user during multiple interactions, and the skin demand characteristics refer to the specific needs or questions raised by the user in skin management; Predict the user's needs through the behavior model, convert the prediction result into an audio signal and play it.

[0029] The preference feature vector is a multi-dimensional vector that represents the preferences and needs of users in skin management. The vector dimensions include but are not limited to: the user's concerns (such as dryness, skin spots, sensitivity, etc.); the user's skin care habits (such as skin care in the morning and evening, regular exfoliation, etc.); the user's product preferences (such as brand, ingredients, price range, etc.); the user's emotional state (such as anxiety, happiness, etc.).

[0030] By using machine learning algorithms such as PCA principal component analysis or TF-IDF to extract key features from user interaction data, combined with the user's historical records and real-time input, the preference feature vector is dynamically updated.

[0031] Use clustering algorithms such as K-means clustering, hierarchical clustering or DBSCAN to group user data. The goal of clustering is to group users with similar behavior patterns and demand characteristics into the same category; associate the user's behavior patterns (such as question frequency, focus of attention, etc.) with skin demand characteristics (such as dryness, skin spots, etc.).

[0032] The model output result represents the potential demand corresponding to a specific behavior pattern. Regularly update the model parameters to ensure that it always fits the changes in user needs. Based on the user's behavior pattern and preference feature vector, combined with the constructed behavior model, predict the user's potential demand. A feedback mechanism can also be introduced to adjust the model weights according to the user's satisfaction score for the suggestions.

[0033] The present invention also provides a skin management robot, including: A first receiving module, the first receiving module receives the voice information of the user, and the voice information at least includes the user's skin demand characteristics and user feedback; A sending module, the sending module uploads the voice information to the cloud server; A second receiving module, the second receiving module receives the response information generated by the cloud server based on the voice information, and the response information is generated through the following steps: Convert the voice information into text information; Based on the large language model and the skin management knowledge base, analyze the text information and generate response information that matches the voice information; A voice module, the voice module converts the received response information into an audio signal and plays it.

[0034] The skin management robot further includes an accuracy calculation module and a data transmission module. The accuracy calculation module is used to calculate the accuracy value based on historical interaction data, and the data transmission module is used to send the accuracy value to the large language model in the cloud server.

[0035] The sending module includes an array microphone and a microprocessor. The array microphone is used to collect the user's voice information, and the microprocessor preprocesses the voice information collected by the array microphone. The preprocessing at least includes noise reduction and signal enhancement.

[0036] The second receiving module includes an Internet of Things transmission module and a voice player. The Internet of Things transmission module transmits the text information to the microprocessor. The microprocessor converts the text information into an audio signal, and the voice player plays the audio signal. Specifically, the Internet of Things transmission module can use wireless communication modules such as WIFI, cellular communication, and Bluetooth.

[0037] According to another embodiment, the user says "My cheeks are dry and peeling". The microphone picks up the voice, and the microprocessor reduces the noise and uploads it to the cloud via Wi-Fi. The cloud calls the "moisturizing and repairing" solution in the knowledge base, combines the user's historical data (such as choosing "hyaluronic acid essence" in the past), and generates a response: "It is recommended to use a hyaluronic acid mask for three consecutive days and avoid excessive cleaning." The response text is converted into voice. After the user feedbacks "I have used the mask and the effect is good", the accuracy module calculates (high satisfaction), and the "hyaluronic acid" solution in the knowledge base is marked for subsequent analysis and question answering.

[0038] The skin management robot of the present application further includes a mirror and a charging interface. The mirror can be a light-supplementing or non-light-supplementing mirror, and the charging interface is a common charging interface and can be charged with a 5V power supply.

[0039] Based on the same inventive concept, the present invention also provides a computer-readable storage medium. The computer-readable storage medium has a computer program, and when the computer program is executed by a processor, it implements the steps of an interaction method for a skin management robot as described above.

[0040] The above are only the preferred embodiments of the present application and are not used to limit the present application. For those skilled in the art, the present application can have various changes and modifications. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present application shall be included within the protection scope of the present application.

[0041] Although the specific implementation manners of the present invention are described above, it is not a limitation to the protection scope of the present invention. Those skilled in the art should understand that on the basis of the technical solution of the present invention, various modifications or deformations that can be made without creative labor by those skilled in the art are still within the protection scope of the present invention.

Claims

1. An interactive method for a skin care robot, characterized in that: The following steps are involved: Receive voice information from a user, the voice information at least including user skin demand characteristics and user feedback, and upload the voice information to a cloud server; Receive response information generated by the cloud server based on the voice information, wherein the response information is generated by the following steps: Converting the voice information into text information; Analyze the text information based on a large language model and a skin management knowledge base, and generate response information matching the voice information; The received response information is converted into an audio signal and played.

2. The interactive method for a skin care robot according to claim 1, characterized in that: The method further comprises: When generating response information, the large language model is not only based on the text information converted from the current user's voice information, but also based on the user's historical interaction data. The historical interaction data at least includes multiple processed text messages, response information and interaction duration. The interaction duration is the recorded time used for each completed interaction.

3. An interactive method for a skin care robot according to claim 1 or 2, characterized in that: Also includes: The accuracy value is sent to the large language model. The accuracy value is calculated based on the historical interaction data. The calculation formula is: ， in, U For accuracy, t i For the i The duration of each response message played during each interaction. T is the total duration of the interaction, x i is the weight of each response, de For user satisfaction with the interaction, i is a positive integer, n The total number of times all response messages are played in this interaction.

4. The interactive method for a skin care robot according to claim 3, characterized in that: Based on text information, response information, user feedback and accuracy, the large language model dynamically adjusts the content of the skin management knowledge base, and the skin management knowledge base at least includes skin physiology knowledge, skin problems, skin diseases, skin care knowledge and skin care products.

5. The voice interaction method for a skin care robot according to claim 1, characterized in that: The method further comprises: Collecting data of multiple user interactions and extracting preference feature vectors from the data; Based on the preference feature vector, a user behavior model is constructed by a clustering algorithm, the model associates the user's behavior pattern with the skin demand characteristics, and predicts the user's needs according to the behavior model; the user's behavior pattern refers to a series of regular or habitual behaviors exhibited by the user during multiple interactions, and the skin demand characteristics refer to the specific needs or problems raised by the user in terms of skin management; Predict user needs through behavioral models, convert the prediction results into audio signals and play them.

6. A skin care robot, characterized in that: include: A first receiving module, the first receiving module receives the user's voice information, the voice information at least including the user's skin demand characteristics and user feedback; A sending module, wherein the sending module uploads the voice information to a cloud server; A second receiving module, wherein the second receiving module receives response information generated by the cloud server based on the voice information, wherein the response information is generated by the following steps: Converting the voice information into text information; Analyze the text information based on a large language model and a skin management knowledge base, and generate response information matching the voice information; A voice module converts the received response information into an audio signal and plays the audio signal.

7. A skin care robot according to claim 6, characterized in that: It also includes an accuracy calculation module and a data transmission module. The accuracy calculation module is used to calculate the accuracy value based on historical interaction data, and the data transmission module is used to send the accuracy value to the large language model in the cloud server.

8. The skin care robot according to claim 6, characterized in that: The sending module includes an array microphone and a microprocessor. The array microphone is used to collect voice information of the user. The microprocessor pre-processes the voice information collected by the array microphone. The pre-processing includes at least noise reduction and signal enhancement.

9. A skin care robot according to claim 8, characterized in that: The second receiving module includes an Internet of Things transmission module and a voice player. The Internet of Things transmission module transmits text information to a microprocessor, the microprocessor converts the text information into an audio signal, and the voice player plays the audio signal.

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

Images