CN112036208A

CN112036208A - Artificial intelligence sprite based on interactive learning system

Info

Publication number: CN112036208A
Application number: CN201910414576.5A
Authority: CN
Inventors: 刘军
Original assignee: Shenzhen Scope Co ltd
Current assignee: Shenzhen Scope Co ltd
Priority date: 2019-05-17
Filing date: 2019-05-17
Publication date: 2020-12-04
Anticipated expiration: 2039-05-17
Also published as: CN112036208B

Abstract

The invention provides an artificial intelligence sprite based on an interactive learning system, the interactive learning system comprises a server and a learning terminal connected with the server through a network, the learning terminal and the server both comprise a plurality of processors, the artificial intelligence sprite is a software system running on the processors, and a program module forming the artificial intelligence sprite comprises: a voice detection unit; an expression feature analysis unit; a convolutional neural network voice analysis unit; a semantic judgment unit; a psychological characteristic analysis unit; and the sprite image is a direct medium for interaction between the user and the learning interaction system and is displayed on a display screen of the learning terminal, and the user can carry out interface operation and information calling on each program module through voice interaction with the sprite image. And bidirectional question and answer can be supported, and the user experience is improved.

Description

Artificial intelligence sprite based on interactive learning system

Technical Field

The invention relates to an artificial intelligence interaction technology and a learning system, in particular to artificial intelligence based on an interactive learning system.

Background

With the development of the internet and computers, the way students learn is also changing. With the increasing maturity of the related technology of artificial intelligence, students can change from the traditional school classroom to a one-to-one learning mode of one person and one machine by means of a home computer or an intelligent display terminal.

The interactive learning system is a bidirectional communication type learning mode which is automatically carried out by learners by multimedia courseware or online resources by utilizing a multimedia computer technology and a network technology, and is also an interactive learning mode, and plays a role on two parties participating in learning in an interactive relationship through dynamic (mainly speaking) interaction so as to achieve the purpose of learning.

The current interactive learning system mainly uses a robot with a fixed form or a fixed medium as a carrier, for example: learning robots, intelligent sound equipment and the like, and performing interactive learning of question and answer with users; for the smart display, for example: learning computers, mobile phones and APPs (application programs) carried by the learning computers and the mobile phones learn by watching videos, pictures, characters and the like on an interface and perform interactive question answering with the learning computers and the mobile phones, and the intelligent display end usually interacts with users through virtual robots or virtual character images.

The current interactive learning system has the following common problems: on the one hand, the learning system can only be asked questions one way by the user; on the other hand, the interactive robot or device has a single image, even if the virtual AI is carried on the display end, the image is also a fixed expression and action, the interactive experience is poor, the attraction to the user is very limited, the user cannot be asked about the problems individually, and the virtual robot cannot complete a diversified interactive mode.

Disclosure of Invention

The technical problem to be solved by the invention is to provide an artificial intelligence sprite based on an interactive learning system, which can support two-way question and answer and improve user experience, aiming at the defects in the prior art.

The technical scheme adopted by the invention for solving the technical problems is as follows: the utility model provides an artificial intelligence sprite based on interactive learning system, this interactive learning system includes server and the learning terminal with this server network connection, this learning terminal and this server all contain a plurality of processors, this artificial intelligence sprite is a software system of operation on these processors, and the program module who constitutes this artificial intelligence sprite includes:

the voice detection unit is used for collecting and preprocessing the questions and the replied voice and providing a voice signal in the process of using the interactive learning system to learn by the user;

the expression characteristic analysis unit is used for carrying out parameterization processing and identification on expressions and action behaviors in the question and answer process of the user and generating original data in the background of the interactive learning system;

the convolutional neural network voice analysis unit is used for carrying out voice size and key word feature processing on the voice signal provided by the voice detection unit;

the semantic judgment unit is used for carrying out semantic judgment on key words in the voice result processed by the convolutional neural network voice analysis unit;

the psychological characteristic analysis unit is used for matching the voice size and the mood characteristic words output by the convolutional neural network voice analysis unit with the result of the expression characteristic analysis unit to obtain the psychological characteristic category of the user in the process of questioning and answering; and

the elfin image is a direct medium for interaction between the user and the learning interactive system, is displayed on the display screen of the learning terminal, and the user can carry out interface operation and information calling on each program module through voice interaction with the elfin image.

In some embodiments, the semantic determination unit is configured with a counter for counting valid user questions and answers, and the sum of the counting results is used as a statistical data: a growth value.

In some embodiments, the artificial intelligence sprite is provided with a plurality of thresholds, the sprite image evolving through feeding of the growth value each time the growth value reaches a threshold level.

In some embodiments, the artificial intelligence sprite is provided with a reward module, the image of the sprite is evolved to the highest level, corresponding reward is carried out, and zero clearing and restoring are carried out.

In some embodiments, the artificial intelligence sprite is provided with a database for storing standard model data of psychological characteristics and grade model data of a sprite figure.

In some embodiments, the standard model data of the psychological characteristics includes an expression model, a psychological model and a color model corresponding to joy, anger, sadness and happiness.

In some embodiments, the database is also used for storing data of the users asking and answering during the learning process of the users through the learning interactive system.

In some embodiments, big data analysis is performed on data generated by the questions and answers of the users in the database, the matching criteria of the questions and answers of the users are iterated, and the iterated result is used as the matching criteria of the subsequent users when questions related to the questions and answers are asked and answered.

In some embodiments, based on the analysis of the big data of the user by the system in the database, the questions and answers with small probability and large probability are fed back regularly, so that active questioning of knowledge points which are of little interest or much interest to the user is realized.

In some embodiments, the sprite image is matched with different feature colors for different psychological feature categories of the user, the feature colors being combined with animations and actions of the sprite image.

The invention has the advantages that the two-way question and answer can be supported and the user experience is improved through the ingenious matching among the voice detection unit, the expression characteristic analysis unit, the convolutional neural network voice analysis unit, the semantic judgment unit, the psychological characteristic analysis unit and the elf.

Drawings

The invention will be further described with reference to the accompanying drawings and examples, in which:

FIG. 1 illustrates the framework structure of an artificial intelligence sprite based interactive learning system of the present invention;

figure 2 illustrates a flow diagram of the image feeding of sprites in the present invention.

Wherein the reference numerals are as follows: 100. the system comprises an artificial intelligence sprite 10, a user 20, a server 21, a database 22, question and answer feedback information 30, a learning terminal 41, a voice detection unit 42, an expression characteristic analysis unit 43, a convolutional neural network voice analysis unit 431, a tone size and mood characteristic word 44, a semantic judgment unit 441, a counter 442, a growth value 45, a psychological characteristic analysis unit 451, a psychological characteristic category 452, a characteristic color 46 and a sprite image.

Detailed Description

The preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

Referring to fig. 1 and 2, fig. 1 illustrates a framework structure of an artificial intelligence sprite based on an interactive learning system according to the present invention. Figure 2 illustrates a flow diagram of the image feeding of sprites in the present invention. The invention provides an artificial intelligence sprite 100 based on an interactive learning system, wherein the interactive learning system comprises a server 20 (such as a cloud server) and a learning terminal 30 (such as a smart phone) connected with the server 20 through a network, and the learning terminal 30 and the server 20 both comprise a plurality of processors. Artificial intelligence sprite 100 is a software system running on these processors. User 10 (student) can perform interactive learning well by applying artificial intelligence sprite 100. The program modules constituting the artificial intelligence sprite 100 include: a voice detection unit 41, an expression feature analysis unit 42, a convolutional neural network voice analysis unit 43, a semantic judgment unit 44, a psychological feature analysis unit 45, and a sprite image 46. These program modules may be flexibly configured to run on the processor of the server 20 and/or the learning terminal 30 according to the needs of the actual application. Wherein the sprite figure 46 is shown on the display screen of the learning terminal 30.

And the voice detection unit 41 is used for collecting and preprocessing the questions and the replied voice of the user 10 in the process of learning by using the interactive learning system and providing a sound signal.

And the expression characteristic analysis unit 42 is used for carrying out parameterization processing and recognition on expressions and action behaviors in the question and answer process of the user 10 and generating original data in the background of the interactive learning system. Specifically, a high-definition camera or the like in a hardware device is used, for example: integrally disposed in the learning terminal 30, or independently disposed from the learning terminal 30.

And a convolutional neural network speech analysis unit 43, configured to extract high-level features of the speech signal provided by the speech detection unit 42 simply and efficiently, and perform feature processing on the speech size and the keyword for the speech collected by the user through the plurality of convolutional layers and corresponding convolutional kernels. The processing methods involved here are currently relatively sophisticated technologies, such as: patent document CN109272990A discloses a speech recognition method based on a convolutional neural network, and is not repeated here. It is worth mentioning that the prior art does not disclose what key features are recognized by speech and how to effectively apply and process the collected feature parameters.

And the convolutional neural network voice analysis unit 43 is configured to perform convolutional neural network calculation on the user voice to obtain the voice size and the mood feature word 431 in the user voice. For example, these mood-related feature words include: exclamation words, visceral words, etc. Further, the key words obtained by processing the user voice are stored in a storage module of the interactive learning system as the result parameters of the preprocessing.

And the semantic judgment unit 44 is used for performing semantic judgment on key words in the voice result processed by the convolutional neural network voice analysis unit 43, and the judgment result is further used by the sprite image 46. For example, if the determination is successful, it indicates that the question and answer of the user 10 are valid, and the big data is captured and matched by linking the system database and the internet cloud database, and is fed back to the user for reply.

And a counter 441, configured in the semantic determination unit 44, for counting the valid user questions and answers. Specifically, after the system (i.e., artificial intelligence sprite 100) determines that the question and answer of the user are valid, counter 441 starts counting, and the total value of the counting results is used as a statistical data: a growth value 442 for further use by the sprite image 46.

For the growth value 442, the system may set a number of thresholds, and the sprite image 46 "evolves" one level by "feeding" the growth value 16 each time the growth value 442 reaches a threshold gear, which is shown on the interface of the learning terminal 30, with different levels of sprite animation being displayed.

Specifically, when the determination is no, a-level sprite animation is displayed, and when the determination is yes, further, when the determination is no, B-level sprite animation is displayed; if the determination is yes, further, a determination is made as to whether the growth value 442 is equal to or larger than C, and if the determination is no, a C-level sprite animation is displayed.

The psychological characteristic analyzing unit 45 is configured to match the information such as the speech size and the mood characteristic words 431 output by the convolutional neural network speech analyzing unit 43 with the result of the expression characteristic analyzing unit 42, so as to calculate and determine the psychological characteristic category 451 such as joy, anger, sade and the like during the process of asking and answering the question of the user. Different character colors 452 are matched to sprite image 46 for different psychological characteristic categories 451 of the user, and character colors 452 do not conflict with animations and actions of sprite image 46.

A sprite image 46, which is a direct medium for the user 10 to interact with the learning interactive system, includes a cartoon character, animation character or object.

It will be appreciated that artificial intelligence sprite 100 is a virtual (or software) robot and sprite image 46 is the medium (interface, portal) through which the virtual robot interacts with user 10. Other program modules of artificial intelligence sprite 100 are the internal structure of this virtual robot. User 10 may interface various program modules of the system with the retrieval of information through voice interaction with sprite 46.

The actions of the sprite figure 46 specifically include: wait action P0, ask action P1, answer action P2, and exception action P3. In the interaction process, in order to increase the interaction experience of the user 10 and the attraction of the sprite image 46, the system is respectively configured with different actions for the question and answer of the user, and for the question or answer of the user, the system is in an abnormal condition when the effective result and the key words cannot be traversed in the database 21 through the server 20; in order not to affect the user's learning attention when the user is not interacting with the system during the learning process, unusual actions and waiting actions are set for the sprite figure 46.

As the number of questions and answers for user 10 accumulates, the greater the rating of the sprite figure 46 "feeding". The system is provided with a reward module for rewarding the sprite image 46 of the user 10 to evolve to the highest level, performing corresponding reward, and performing zero clearing reduction, thereby enabling the sprite image 46 to interact with the user 10 in a diversified manner.

And the server 20 is used for linking the system to the database of each program module, temporarily storing data generated in real time and background linking each program module of the system. The server 20 is further configured to perform keyword search and information retrieval in the cloud of the external network after the semantic determination unit 44 operates when the user 10 performs question and answer, that is, reply to the question and answer of the user based on analysis of big data and the like in the shared network. It is worth mentioning that such a path is one of the functions of the system, and is not personalized, and is only a way to improve the accuracy of information feedback.

A database 21 for storing standard model data of psychological characteristics, such as expression models, psychological models and color models corresponding to joy, anger, sadness and happiness; and animation of sprite image 46, motion level model data such as an ABC level sprite animation. These model data are system pre-stored features.

And the database 21 is also used for storing data of questions and answers of the user 10 during the learning process of the user 10 through the artificial intelligence sprite 100, and the data belong to personalized and independent data of the user 10.

The system analyzes big data of the data generated by the question and answer of the user in the database 21, and iterates the matching standard of the question and answer of the user, wherein the iterated data belongs to the data independently generated by the user, so that when the subsequent user asks and answers the relevant questions, the standard matched by the system is the iterative result in the personalized learning process of the user, namely: the result of the iteration is used as a matching criterion for the subsequent users when asking and answering the relevant questions.

It is understood that the database 21 may be configured on the server 20, on the learning terminal 30, or on another device independent from the server 20 and the learning terminal 30, and is invoked by the server 20 and the learning terminal 30.

Based on the analysis of the big data of the user by the system in the database 21, the questions and answers with small probability and large probability are fed back at regular time (namely, the question and answer feedback information 22), the probability threshold value is preset by the system, so that active questioning is performed on knowledge points which are concerned by the user rarely or much, and the questioning period and time can be set individually according to the user.

The learning terminal 30 may be a mobile phone terminal or a PC terminal supporting the artificial intelligence sprite 100, or may be an intelligent learning device with a screen, a camera, a microphone, a sound system, or the like.

The invention is further described in detail with reference to specific embodiments in the following, with reference to fig. 1 and 2:

firstly, when the user 10 learns at the learning terminal 30, a question Q1 is presented, and the action matching of the sprite image 46 is a question action P0; assuming that at this time, the growth value 442 of the user 10 is a-1, the sprite image 46 (corresponding to the level A sprite animation) remains in a static state with no motion;

secondly, the convolutional neural network voice analysis unit 43 calculates and outputs the keyword [ X ] of the Q1 question, the sprite 46 traverses all the sentences related to [ X ] which are answered and provided by the user 10 in the past in the database 21 through the server 20 for matching, and simultaneously traverses and matches the data in the cloud of the extranet, and feeds back the matching result in time.

Thirdly, the sprite character 46 can be matched with the answer of the question Q1 in the system, then the question Q1 is judged to be valid, in the counter 441 +1, the growth value 442 of the user 10 is a, and then the sprite character 46 of the user 10 is "evolved" from the level a sprite animation to the level B sprite animation.

Fourthly, the elfin image 46 is set to 'normal', and no mood characteristic voice, words and expressions are detected in the system; the sprite image 46 is set to be 'happy', and the mood characteristic 'take a care of' voice and smiling expression are detected in the system; the sprite 46 is set to "fidgety" and detects the high pitch intonation in the system along with the words "fast click", "how", etc.; elfin character 46 is set to "sad" and the relieved speech is detected at the system with words such as "sad" and so on; the sprite piece 46 is set to "anger", words such as rereaded voice accompanied by a network word, etc. are detected at the sprite piece 46. For the five psychological characteristics, the system presets corresponding expression characteristics respectively for the expression characteristic analysis unit 41 to match.

For example, the above five psychological characteristics correspond to colors of "green, yellow, orange, purple, red", etc., and the sprite 46 is identified by combining its expression, and when it is judged that the user 10 is "fidgety" in the process of interactive questioning, the image of the sprite 46 becomes "yellow", and the presentation of the questioning action P1 is performed.

Fifthly, the system presets that the small probability threshold is 20 percent, the large probability threshold is 80 percent, the period is one week, the sprite image 46 carries out statistical calculation on all the key words of the question and answer in the last week of the user in the database 21 every other week, and actively asks the key words of which the number is lower than 20 percent and higher than 80 percent.

Compared with the prior art, the artificial intelligence sprite 100 based on the interactive learning system of the present invention can bring the following advantages:

1. with the help of the convolutional neural network speech analysis unit 43, the speech asked and answered by the user can be subjected to high-level feature analysis through the convolutional neural network speech analysis model, and feature parameters such as speech size, mood feature words and semantic keywords are extracted.

2. By matching the analysis of the speech size, pitch and mood characteristics with the sprite images without color and image levels during the expression and question-answering process of the user, different psychological characteristics of the user can be captured and different sprite images 46 can be generated.

3. By analyzing the database 21 of the question and answer of the user, the knowledge points related to the large probability and the small probability of the user are output to actively ask the user in a timing and qualitative mode.

4. The sprite image 46 specifically adopts images such as cartoons, characters, real objects and the like, has a mode of growing and feeding, and enables the sprite image 46 to grow and evolve in the process of interacting with a user by setting a growth value 442 and sprite animation and actions of different levels and a mode of mutual transformation and matching, so that the diversity and interestingness of interactive experience are increased.

It should be understood that the above embodiments are only used for illustrating the technical solutions of the present invention, and not for limiting the same, and those skilled in the art can modify the technical solutions described in the above embodiments, or make equivalent substitutions for some technical features; and such modifications and substitutions are intended to be included within the scope of the appended claims.

Claims

1. The utility model provides an artificial intelligence sprite based on interactive learning system, this interactive learning system includes server and the learning terminal with this server network connection, this learning terminal all contains a plurality of processors with this server, this artificial intelligence sprite is a software system of operation on these processors, its characterized in that, and the program module who constitutes this artificial intelligence sprite includes:

2. An interactive learning system based artificial intelligence sprite as claimed in claim 1 wherein: the semantic judgment unit is provided with a counter for counting the effective user questions and answers, and the sum of the counting results is used as statistical data: a growth value.

3. An interactive learning system based artificial intelligence sprite as claimed in claim 2 wherein: the artificial intelligence sprite is provided with a plurality of threshold values, and the sprite image evolves by feeding of the growth value each time the growth value reaches a threshold level.

4. An interactive learning system based artificial intelligence sprite as claimed in claim 3 wherein: the artificial intelligence sprite is provided with a reward module, the image of the sprite is evolved to the highest level, corresponding reward is carried out, and zero clearing reduction is carried out.

5. An interactive learning system based artificial intelligence sprite as claimed in claim 1 wherein: the artificial intelligence sprite is provided with a database for storing standard model data of psychological characteristics and grade model data of a sprite image.

6. An interactive learning system based artificial intelligence sprite as claimed in claim 5 wherein: the standard model data of the psychological characteristics comprises an expression model, a psychological model and a color model corresponding to joy, anger, sadness and happiness.

7. An interactive learning system based artificial intelligence sprite as claimed in claim 5 wherein: the database is also used for storing the data of the users' own question and answer, namely the question and answer feedback information, in the process of learning through the interactive learning system.

8. An interactive learning system based artificial intelligence sprite as claimed in claim 7 wherein: and performing big data analysis on data generated by the question and answer of the user in the database, iterating the matching standard of the question and answer of the user, and taking the iteration result as the matching standard of the follow-up user when the question and answer is related to the question.

9. An interactive learning system based artificial intelligence sprite as claimed in claim 8 wherein: based on the analysis of the system to the big data of the user in the database, the questions and answers with small probability and large probability are fed back regularly, so that the knowledge points which are concerned by the user rarely or much are asked actively.

10. An interactive learning system based artificial intelligence sprite as claimed in claim 1 wherein: for different psychological characteristic categories of the user, different characteristic colors are matched for the sprite image, and the characteristic colors are combined with the animation and the action of the sprite image.