Hybrid Physical Education Teaching and Curriculum Design Based on a Voice Interactive Artiﬁcial Intelligence Educational Robot

: In order to promote the development of individualized, accurate and intelligent physical education teaching, combined with artiﬁcial intelligence technology, the current physical education teaching mode has been improved. Through the establishment of an artiﬁcial intelligence educational robot based on voice interaction, a hybrid physical education teaching mode is constructed to realize personalized education for students. First, the speech recognition system is designed from three aspects of speech recognition, interaction management and speech synthesis, and the accuracy of recognition is improved by algorithm. Second, a new mode of hybrid physical education teaching is constructed. Through intelligent information technology, the advantages of traditional physical education teaching are combined to improve the classroom e ﬃ ciency of physical education teaching and personalized education ability for students. Finally, the relevant experimental scheme and questionnaire are designed, and the actual situation of an educational robot introduced into physical education teaching is investigated and evaluated. The results show that the recognition accuracy of the artiﬁcial intelligence speech recognition system can reach more than 90%. It can communicate well with students and answer students’ questions. An educational robot is introduced into physical education teaching, and students’ learning attitude and interest are evaluated. The results show that before and after the introduction of an educational robot in physical education teaching, the average score of students’ learning interest increases by 21 points, and the average score of learning attitude increases by 9.8 points. Therefore, the introduction of an artiﬁcial intelligence educational robot based on voice interaction in physical education teaching can help to improve the classroom e ﬃ ciency of physical education teaching and students’ interest. This study provides a reference for the development of artiﬁcial intelligence teaching and promoting the development of artiﬁcial intelligence. educational robot into a physical education classroom is investigated by questionnaire. The results show that the recognition accuracy of the designed speech interaction system based on artiﬁcial intelligence is more than 90%. It can recognize di ﬀ erent genders


Introduction
Learning motivation, learning interest and learning attitude are the key factors affecting the development of students' mental health. In the process of school education, schools and teachers should understand the characteristics of students' learning motivation, attitude and interest, and the formation mode and development law, so as to better carry out ideological education and teaching work for students. Learning motivation is the driving force to cause, maintain and promote students' action; learning attitude and enthusiasm are the direct manifestation of students' learning motivation, so stimulating students' learning motivation is the basic condition for forming correct learning attitude;  The following describes the voice interactive educational robot model: (1) Speech recognition module. It consists of two parts: speech acquisition and recognition. Through a small microphone array, directional pickup is formed for the speaker in a certain range to reduce the impact of external noise. Speech recognition realizes the transformation from speech to text through the processing of speech signal. A special vocabulary library for the vocabulary in the professional field is established to store the professional vocabulary of different disciplines, and it is used in the training of speech model [19,20].
Speech recognition consists of three parts. First, the input speech is transformed into speech signal through a microphone, which is the reception and processing of speech signal. The second is the training of an acoustic model and language model, and the establishment of a pronunciation dictionary. Finally, the input speech signal is decoded, and the speech model to be recognized is searched and matched with the trained model to find out the optimal speech template. The output layer is used to identify results.
Among them, the acoustic model is an important part of speech recognition. The training process is to establish the model parameters for each voice of the voice data in the speech database, and compare with the voice parameters in the acoustic model, calculate the distance between them, and find the most similar results. In the acoustic model, the speech units have different characteristics. Therefore, the acoustic model designed according to the characteristics of speech pronunciation can improve the recognition rate of speech recognition. The modeling unit of the acoustic model includes phoneme, syllable and word. In speech recognition with a large number of words, phoneme is generally used as the modeling unit of the acoustic model. Although the acoustic model can record each supported speech feature parameter, it still has a low recognition rate for some homophones and near syllable words. Therefore, it is necessary to restrict the recognition system from grammar rules to improve the recognition rate. Language model can describe the internal relationship and transfer relationship between different speech units, eliminate the fuzziness of words and improve the discrimination of recognition. The common modeling methods of the language model include rule-based model and statistical-based model. The performance of the statistical language model is related to the content of the training model, so the recognition rate is very low when the recognition content is different from the training content. Therefore, for speech recognition in a specific field, it is necessary to establish a speech database of professional vocabulary, combine it with the relevant grammar and semantic rules to improve the ability and performance of speech recognition in specific The following describes the voice interactive educational robot model: (1) Speech recognition module. It consists of two parts: speech acquisition and recognition. Through a small microphone array, directional pickup is formed for the speaker in a certain range to reduce the impact of external noise. Speech recognition realizes the transformation from speech to text through the processing of speech signal. A special vocabulary library for the vocabulary in the professional field is established to store the professional vocabulary of different disciplines, and it is used in the training of speech model [19,20].
Speech recognition consists of three parts. First, the input speech is transformed into speech signal through a microphone, which is the reception and processing of speech signal. The second is the training of an acoustic model and language model, and the establishment of a pronunciation dictionary. Finally, the input speech signal is decoded, and the speech model to be recognized is searched and matched with the trained model to find out the optimal speech template. The output layer is used to identify results.
Among them, the acoustic model is an important part of speech recognition. The training process is to establish the model parameters for each voice of the voice data in the speech database, and compare with the voice parameters in the acoustic model, calculate the distance between them, and find the most similar results. In the acoustic model, the speech units have different characteristics. Therefore, the acoustic model designed according to the characteristics of speech pronunciation can improve the recognition rate of speech recognition. The modeling unit of the acoustic model includes phoneme, syllable and word. In speech recognition with a large number of words, phoneme is generally used as the modeling unit of the acoustic model. Although the acoustic model can record each supported speech feature parameter, it still has a low recognition rate for some homophones and near syllable words. Therefore, it is necessary to restrict the recognition system from grammar rules to improve the recognition rate. Language model can describe the internal relationship and transfer relationship between different speech units, eliminate the fuzziness of words and improve the discrimination of recognition. The common modeling methods of the language model include rule-based model and statistical-based model. The performance of the statistical language model is related to the content of the training model, so the recognition rate is very low when the recognition content is different from the training content. Therefore, for speech recognition in a specific field, it is necessary to establish a Sustainability 2020, 12, 8000 5 of 14 speech database of professional vocabulary, combine it with the relevant grammar and semantic rules to improve the ability and performance of speech recognition in specific fields through a language model, and reduce the search content of speech recognition. Through the pronunciation dictionary, the mapping relationship between words and phonemes is established, and the relationship between acoustic model and language model is established. The phonemes of the corresponding words can help to match the corresponding words in the acoustic model [21][22][23][24][25].
(2) Interaction management module. The results given by the speech recognition module may have deviation, so it is necessary to match the recognition results with the rules in the interaction management module, judge whether the user's results are correct according to the matching results and select the next step process according to different answers, and broadcast the corresponding answers of the process with a speech synthesis module. In the teaching environment of an educational robot, the robot needs to answer students' questions about the physical education course, but the performance of the speech recognition module will be affected by noise and other factors, which will lead to the robot's wrong judgment. Therefore, it is necessary to set the rules of voice interaction, judge students' intention and give reasonable answers [26,27].
(3) Speech synthesis module. It includes two parts: speech broadcast and speech synthesis. It uses prosodic analysis and synthesis algorithm to transform text into speech waveform, and then converts electrical signal into sound signal to realize voice broadcast. Chinese speech synthesis is realized by waveform splicing. First of all, the natural language is divided and the pronunciation of a single word is extracted as a phonetic unit. These units are combined into a pronunciation database. In the process of synthesis, the monosyllabic of the pronunciation database is adjusted, and the splicing algorithm is used to splice the complete sentences, so as to achieve the function of speech output [28][29][30][31].
The designed educational robot is used to solve the problems encountered by students in their daily physical education. It can complete roll call, read text, assign homework and other basic teaching activities, and push corresponding teaching resources to students according to their age, gender, interest and other information. Through the analysis of students' learning data, students' current learning attitude and interest are judged, and then whether further deep learning and extended learning are needed. Therefore, in order to reflect its characteristics in teaching and explore the voice interaction between students and robots, the actual application of voice interaction is tested. An educational robot is divided into three parts, including motion hardware layer, hardware management layer and microcomputer. The motion layer is used to realize the walking and speed control of the educational robot; the motion management layer can control the hardware such as level meter and tachometer, and control the overall motion state of the robot; and the microcomputer contains voice and action recognition system, which can communicate with students.
The voice interactive education robot is tested by question and answer, and Figure 2

Construction of Hybrid Teaching Mode in Physical Education
Teaching mode refers to the fixed framework and procedure of teaching activities established under the guidance of teaching theory. It standardizes the whole teaching activity and the internal relations among various elements from a macro perspective, and highlights the orderliness and operability of the teaching mode. The traditional physical education classroom teaching is that the teacher demonstrates and explains the technical movements, and the students complete the learning of motor skills through imitation after the teacher's demonstration. However, many movements have continuity and space, which cannot be taught by decomposition. It is difficult for students to understand the essentials of movements in a short period of time only through eye observation. Hybrid teaching can solve this problem well. Using modern educational resources, teachers record movement essentials into videos, and mark the technical essentials in the process of movement with slow motion or words to help students learn movement essentials. The establishment of a hybrid physical education teaching mode is based on the theory of hybrid teaching. Through the analysis of the characteristics of physical education teaching and the development of information technology, the advantages of traditional classroom and online classroom are combined [32][33][34][35], and a hybrid physical education teaching mode as shown in Figure 3 is constructed.

Construction of Hybrid Teaching Mode in Physical Education
Teaching mode refers to the fixed framework and procedure of teaching activities established under the guidance of teaching theory. It standardizes the whole teaching activity and the internal relations among various elements from a macro perspective, and highlights the orderliness and operability of the teaching mode. The traditional physical education classroom teaching is that the teacher demonstrates and explains the technical movements, and the students complete the learning of motor skills through imitation after the teacher's demonstration. However, many movements have continuity and space, which cannot be taught by decomposition. It is difficult for students to understand the essentials of movements in a short period of time only through eye observation. Hybrid teaching can solve this problem well. Using modern educational resources, teachers record movement essentials into videos, and mark the technical essentials in the process of movement with slow motion or words to help students learn movement essentials. The establishment of a hybrid physical education teaching mode is based on the theory of hybrid teaching. Through the analysis of the characteristics of physical education teaching and the development of information technology, the advantages of traditional classroom and online classroom are combined [32][33][34][35], and a hybrid physical education teaching mode as shown in Figure 3 is constructed.
In the hybrid physical education teaching mode, the teaching process is divided into three interrelated parts, namely, autonomous learning before class, practical learning in class and consolidated extended learning after class. The traditional classroom teaching mode of "teaching before learning" has been changed into the new classroom teaching mode of "learning before teaching". In the whole teaching process, teachers guide students to learn independently by designing teaching activities, organize classroom teaching content according to students' self-study situation, and guide students to consolidate the learning content after class.
In the process of learning independently before class, according to the course content and the analysis of the learning situation, the teacher records the relevant physical education teaching video, makes the teaching courseware, sets the related questions, defines the learning goal, and transmits them to the teaching platform. Through the teaching platform, students can find solutions to the problems assigned by teachers through self-learning, watch teaching courseware and teaching videos, conduct group discussion, and seek teachers' answers to the difficulties encountered online. In the process of autonomous learning, teachers need to guide and supervise students.  In the hybrid physical education teaching mode, the teaching process is divided into three interrelated parts, namely, autonomous learning before class, practical learning in class and consolidated extended learning after class. The traditional classroom teaching mode of "teaching before learning" has been changed into the new classroom teaching mode of "learning before teaching". In the whole teaching process, teachers guide students to learn independently by designing teaching activities, organize classroom teaching content according to students' self-study situation, and guide students to consolidate the learning content after class.
In the process of learning independently before class, according to the course content and the analysis of the learning situation, the teacher records the relevant physical education teaching video, makes the teaching courseware, sets the related questions, defines the learning goal, and transmits them to the teaching platform. Through the teaching platform, students can find solutions to the problems assigned by teachers through self-learning, watch teaching courseware and teaching videos, conduct group discussion, and seek teachers' answers to the difficulties encountered online. In the process of autonomous learning, teachers need to guide and supervise students.
In the classroom practice teaching, students will combine the content of autonomous learning with their own understanding, and display it in the form of groups. The teacher and other students make comments after watching. According to the feedback of the class, the teacher answers the students' questions and explains and demonstrates the teaching contents. Students learn the content of the course in groups, and further master the content of the course. In the guidance of students' learning interaction, the robot can follow the students and communicate with the students. By observing the students' performance in the course, the robot judges the students' learning state, identifies each action of the students, compares them with the standard actions in the system library, analyzes whether the students' actions are standardized, and guides them at the appropriate time to help them complete the teaching contents of physical education courses. When students have learning difficulties, the robot encourages students to adjust their learning attitude and weakens their fear and anxiety. Through the robot record of each student's learning state, learning progress, as well as course mastery, teachers can carry out personalized teaching guidance for each student.
During the consolidation after class, students record the contents of physical education courses according to their own learning situation and upload them to the course platform. The teacher comments on the students' videos and points out the existing problems. The students reflect and summarize according to the teacher's comments and improve the existing problems. In addition, teachers can upload the collected course materials to the teaching platform, allowing students to selectively watch according to their own learning situation, so as to realize the extension of the course In the classroom practice teaching, students will combine the content of autonomous learning with their own understanding, and display it in the form of groups. The teacher and other students make comments after watching. According to the feedback of the class, the teacher answers the students' questions and explains and demonstrates the teaching contents. Students learn the content of the course in groups, and further master the content of the course. In the guidance of students' learning interaction, the robot can follow the students and communicate with the students. By observing the students' performance in the course, the robot judges the students' learning state, identifies each action of the students, compares them with the standard actions in the system library, analyzes whether the students' actions are standardized, and guides them at the appropriate time to help them complete the teaching contents of physical education courses. When students have learning difficulties, the robot encourages students to adjust their learning attitude and weakens their fear and anxiety. Through the robot record of each student's learning state, learning progress, as well as course mastery, teachers can carry out personalized teaching guidance for each student.
During the consolidation after class, students record the contents of physical education courses according to their own learning situation and upload them to the course platform. The teacher comments on the students' videos and points out the existing problems. The students reflect and summarize according to the teacher's comments and improve the existing problems. In addition, teachers can upload the collected course materials to the teaching platform, allowing students to selectively watch according to their own learning situation, so as to realize the extension of the course [36][37][38][39][40][41].
In the process of physical education classroom teaching, students only need to describe the sports items learned by robots. The robots will analyze and introduce the action essentials and difficulties involved in the course in detail, and use the pre-loaded sports teaching video to assist with the explanation. The robot answers the problems encountered by the students, makes the originally boring teaching process easy to understand and improves the students' learning efficiency [42]. For example, for wushu exercise or gymnastics and other sports with many action standard essentials, the robot will play the video of action to be taught in the course, analyze the main points by slow motion, and correct the students' actions. For group sports, robots will tell students how to avoid personal Sustainability 2020, 12, 8000 8 of 14 mistakes that affect the overall performance of team games. In the future, with the development of artificial intelligence robots, robots will have the ability to surpass senior teachers and complete students' personalized teaching contents independently [43,44].
Hybrid physical education teaching can break through the limitation of teaching hours, allow students to take the initiative to learn, and let teachers carry out individualized teaching and guidance for students, so as to realize individualized teaching of students according to their aptitude and truly achieve individualized teaching with students as the main body. Moreover, the hybrid teaching can enable teachers to mainly solve the problems encountered by students, and enable students to study the conventional classroom warming up, physical knowledge and skills through video or courseware, making teaching more targeted, facilitating students to consolidate learning after class, and improving the classroom efficiency.

Research on Hybrid Teaching Mode of Physical Education Course Based on a Voice Interactive Robot
In order to study the actual situation of an educational robot as an auxiliary tool to guide students to learn physical education courses, two classes of grade five in a primary school are selected for teaching experiments. A total of 80 students from the two classes are divided into two groups with 40 students in each group. The two groups have four physical education classes every week, and a teacher uses the physical education teaching method of an educational robot and the traditional teaching method respectively for three months. Before and after the experiment, the two groups are given the Evaluation Scale Of Primary School Students' Interest In Physical Education (27 questions in total, with a score of 1-5 for each question) and the Evaluation Scale Of Primary School Students' Learning Attitude Towards Physical Education (12 questions in total, with a score of 1-5 for each question) compiled by Professor Wang Xiaozan. Students' interest in physical education and learning attitude before and after the course are scored, and the average score is calculated by adding the scores. Finally, four teachers are asked to grade the students' performance of wushu exercise in terms of the degree of completion and expression (total score 100 points). SPSS is used to analyze the collected questionnaire data. Confirmatory factor analysis is used to test the structural validity, AVE method is used to test the discrimination validity, and Cronbach's α coefficient is used to test the reliability of the scale. The distribution and recovery of the questionnaire meet the needs of scientific research.

Validity and Reliability Analysis of the Scale
In order to test the construct validity of the scale, the KMO value needs to be detected. The KMO value of this study is 0.967, and the KMO values of the other secondary potential variables are all above 0.7, which indicates that the test effect of the scale is good, and it is suitable for factor analysis. AVE method is used to analyze the discrimination validity of the scale. The computed results show that the AVE values of each dimension are greater than the correlation coefficient between the dimensions, that is, the square of standardized correlation. Therefore, the dimensions of the scale used have discriminant validity.
Cronbach's α coefficient is used to process the collected data, and the reliability of the total scale and each dimension of the scale is tested. The Cronbach's α coefficient is calculated to be 0.975, the reliability test value of each secondary potential variable dimension is above 0.8, and the overall α coefficient of each dimension is greater than the judgment standard of 0.7. This shows that the scale data used has high internal consistency, good stability and good overall reliability, which can meet the needs of scientific research.

Performance Test of Educational Robot Based on Voice Interaction
The designed educational robot is used in teachers' daily teaching, so in order to reflect its characteristics in teaching and explore the voice interaction between students and robots, the actual Sustainability 2020, 12, 8000 9 of 14 application of voice interaction is explored. The educational robot is divided into three parts, including motion hardware layer, hardware management layer and microcomputer. The motion hardware layer is used to realize the walking and speed control of the educational robot; the hardware management layer can control the hardware such as level meter and tachometer to control the overall motion state of the robot; and the microcomputer contains a speech recognition system, which can communicate with students.
An educational robot in the physical education teaching environment for students' speech recognition is studied. The commonly used 200 sentences of PE teaching commands are tested on 10 students, including five boys (the first five groups) and five girls (the last five groups). Each time, 40 sentences are randomly selected for recognition test. Figure 4 shows the specific recognition results. The recognition accuracy rate is the percentage of correct teaching commands identified in 40 teaching commands. data used has high internal consistency, good stability and good overall reliability, which can meet the needs of scientific research.

Performance Test of Educational Robot Based on Voice Interaction
The designed educational robot is used in teachers' daily teaching, so in order to reflect its characteristics in teaching and explore the voice interaction between students and robots, the actual application of voice interaction is explored. The educational robot is divided into three parts, including motion hardware layer, hardware management layer and microcomputer. The motion hardware layer is used to realize the walking and speed control of the educational robot; the hardware management layer can control the hardware such as level meter and tachometer to control the overall motion state of the robot; and the microcomputer contains a speech recognition system, which can communicate with students.
An educational robot in the physical education teaching environment for students' speech recognition is studied. The commonly used 200 sentences of PE teaching commands are tested on 10 students, including five boys (the first five groups) and five girls (the last five groups). Each time, 40 sentences are randomly selected for recognition test. Figure 4 shows the specific recognition results. The recognition accuracy rate is the percentage of correct teaching commands identified in 40 teaching commands.   Figure 4 shows that 36 out of every 40 physical education commands can be identified, and the overall recognition accuracy is more than 90%. In addition, the recognition of different genders and different sound characteristics can achieve the recognition effect of speaker-independent. However, the recognition effect of an educational robot will be affected when the environment is noisy and the standard Mandarin is not standard. This shows that its anti-interference ability and speech recognition ability needs to be further improved. Voice interaction technology is used to answer students' questions, which will deepen students' understanding of the problems and increase the interest of classroom teaching. Therefore, the application of voice interaction technology in the design of an educational robot is very meaningful.

The Influence of Different Physical Education Teaching Methods on Students
Students' interest in physical education represents the students' understanding of the physical education classroom and the positive degree of learning. Therefore, the cultivation of students' interest in sports is the key to physical education teaching. The changes of students' interest in  Figure 4 shows that 36 out of every 40 physical education commands can be identified, and the overall recognition accuracy is more than 90%. In addition, the recognition of different genders and different sound characteristics can achieve the recognition effect of speaker-independent. However, the recognition effect of an educational robot will be affected when the environment is noisy and the standard Mandarin is not standard. This shows that its anti-interference ability and speech recognition ability needs to be further improved. Voice interaction technology is used to answer students' questions, which will deepen students' understanding of the problems and increase the interest of classroom teaching. Therefore, the application of voice interaction technology in the design of an educational robot is very meaningful.

The Influence of Different Physical Education Teaching Methods on Students
Students' interest in physical education represents the students' understanding of the physical education classroom and the positive degree of learning. Therefore, the cultivation of students' interest in sports is the key to physical education teaching. The changes of students' interest in physical education before and after using different teaching methods are compared between the two groups. Figure 5 shows the results.
Sustainability 2020, 12, x FOR PEER REVIEW 10 of 14 physical education before and after using different teaching methods are compared between the two groups. Figure 5 shows the results.  Figure 5 shows that the average score of the robot participating teaching group before and after the experiment is 84 points and 105 points, an increase of 21 points; the average score of the traditional teaching method group is 84.5 points before the experiment, and the average score after the experiment is 98 points, an increase of 13.5 points. This shows that the participation of an educational robot can significantly increase students' interest in sports learning, and the effect is better than that of traditional teaching methods.
Students' learning attitude is the overall evaluation and internal tendency of physical education learning, and its score is related to students' academic performance. Therefore, cultivating a good learning attitude is the key to physical education teaching. Figure 6 shows the results of the two groups of students' sports learning attitude before and after the experiment.  Figure 5 shows that the average score of the robot participating teaching group before and after the experiment is 84 points and 105 points, an increase of 21 points; the average score of the traditional teaching method group is 84.5 points before the experiment, and the average score after the experiment is 98 points, an increase of 13.5 points. This shows that the participation of an educational robot can significantly increase students' interest in sports learning, and the effect is better than that of traditional teaching methods.
Students' learning attitude is the overall evaluation and internal tendency of physical education learning, and its score is related to students' academic performance. Therefore, cultivating a good learning attitude is the key to physical education teaching. Figure 6 shows the results of the two groups of students' sports learning attitude before and after the experiment.
groups. Figure 5 shows the results.  Figure 5 shows that the average score of the robot participating teaching group before and after the experiment is 84 points and 105 points, an increase of 21 points; the average score of the traditional teaching method group is 84.5 points before the experiment, and the average score after the experiment is 98 points, an increase of 13.5 points. This shows that the participation of an educational robot can significantly increase students' interest in sports learning, and the effect is better than that of traditional teaching methods.
Students' learning attitude is the overall evaluation and internal tendency of physical education learning, and its score is related to students' academic performance. Therefore, cultivating a good learning attitude is the key to physical education teaching. Figure 6 shows the results of the two groups of students' sports learning attitude before and after the experiment.   Figure 6 shows that the average score of learning attitude of the robot participation group before the experiment is 44.5, and the average score after the experiment is 54.3, an increase of 9.8 points; the average score of the traditional teaching method group is 45.3 before the experiment, and the average score after the experiment is 50, an increase of 4.7 points. This shows that robot participation in teaching can improve students' attitude towards sports learning, and robot participation in teaching can enhance students' learning motivation.
The teacher scores the two groups of students on the degree of completion and performance of wushu exercise, and Figure 7 shows the results. Figure 6 shows that the average score of learning attitude of the robot participation group before the experiment is 44.5, and the average score after the experiment is 54.3, an increase of 9.8 points; the average score of the traditional teaching method group is 45.3 before the experiment, and the average score after the experiment is 50, an increase of 4.7 points. This shows that robot participation in teaching can improve students' attitude towards sports learning, and robot participation in teaching can enhance students' learning motivation.
The teacher scores the two groups of students on the degree of completion and performance of wushu exercise, and Figure 7 shows the results.  Figure 7 shows that the average score of the robot participation group is 86 points, and the average score of the traditional teaching group is 81.83. There is a big difference between the two classes in the completion degree of wushu exercise. The average score of the robot participation group is 85 points, while that of the traditional teaching group is 84.5 points. There is no significant difference between the two groups. This shows that the introduction of an educational robot in physical education teaching can better improve the learning effect on students. However, to achieve a higher quality teaching effect and improve the expression of wushu exercise, teachers need to teach in person.
To sum up, the participation of robots in physical education teaching can enhance students' interest and attitude in sports learning and help students to memorize classroom knowledge. Moreover, robot participation in teaching is an attempt to integrate artificial intelligence technology into the physical education teaching method, which proves that the development of science and technology in the field of physical education teaching can promote students' learning enthusiasm, improve their interest in sports learning and sports learning attitude. However, robot teaching cannot improve students' performance.

Conclusions
In order to study the application of an AI educational robot based on voice interaction in physical education, and promote the development of physical education to intelligent and individualized education, first, the voice interaction system based on AI is designed from three aspects of speech  Figure 7 shows that the average score of the robot participation group is 86 points, and the average score of the traditional teaching group is 81.83. There is a big difference between the two classes in the completion degree of wushu exercise. The average score of the robot participation group is 85 points, while that of the traditional teaching group is 84.5 points. There is no significant difference between the two groups. This shows that the introduction of an educational robot in physical education teaching can better improve the learning effect on students. However, to achieve a higher quality teaching effect and improve the expression of wushu exercise, teachers need to teach in person.
To sum up, the participation of robots in physical education teaching can enhance students' interest and attitude in sports learning and help students to memorize classroom knowledge. Moreover, robot participation in teaching is an attempt to integrate artificial intelligence technology into the physical education teaching method, which proves that the development of science and technology in the field of physical education teaching can promote students' learning enthusiasm, improve their interest in sports learning and sports learning attitude. However, robot teaching cannot improve students' performance.

Conclusions
In order to study the application of an AI educational robot based on voice interaction in physical education, and promote the development of physical education to intelligent and individualized education, first, the voice interaction system based on AI is designed from three aspects of speech recognition, interaction management and speech synthesis, so as to improve the accuracy of speech recognition and enhance the interactive experience. Then, the curriculum mode of hybrid physical education teaching is constructed. Combined with the advantages of traditional physical education and intelligent information technology, the individualized education ability of physical education teaching is improved.
Finally, the speech recognition accuracy of the designed speech interaction system is tested, and the actual effect of introducing an educational robot into a physical education classroom is investigated by questionnaire. The results show that the recognition accuracy of the designed speech interaction system based on artificial intelligence is more than 90%. It can recognize different genders and different voice characteristics, and can be used in a physical education classroom. The results of the questionnaire survey show that the introduction of an educational robot can significantly improve students' learning attitude and interest in physical education. Therefore, the introduction of an artificial intelligence education robot based on voice interaction in physical education teaching can help to improve the classroom efficiency of physical education teaching and students' interest. Finally, the speech recognition accuracy of the designed voice interaction system is tested, and the actual effect of introducing an educational robot into a physical education classroom is investigated by questionnaire. The results show that the recognition accuracy of the designed voice interaction system based on artificial intelligence is more than 90%. It can recognize different gender and different voice characteristics, achieve the recognition effect of a non-specific person, and can be used in a physical education classroom. The results of the questionnaire survey show that the introduction of an educational robot can significantly improve students' learning attitude and interest in physical education. Therefore, the introduction of an artificial intelligence education robot based on voice interaction in physical education teaching can help to improve the classroom efficiency of physical education teaching and students' interest.
The research reveals that students' interest in physical education is affected by many aspects. Therefore, in the further study of physical education participation, the influence of moral and social behavior on students should be considered. The significance of a physical education curriculum is not only to improve students' interest in sports and physical health, but also to strengthen students' awareness of fair competition, cultivate students' ability in teamwork, temper their willpower and cultivate their healthy mind.
However, there are still some deficiencies. The anti-interference ability of the designed speech recognition system is weak, which will affect the recognition accuracy in a noisy environment. In addition, its recognition accuracy for dialects is also low. Moreover, due to the small sample size, the conclusion of this study does not prove that the use of educational robots in the physical education classroom can enhance everyone's interest in sports, knowledge learning and physical health. Therefore, in the follow-up study, it is necessary to study a wider range of people, and prove the relationship between the variables by means of statistics or control group.