Professional Development Paths and Teacher Competence Enhancement of Vocational School Teachers under the Integration of Intelligent Technology

This section aims to recognize students’ emotional states in the classroom through the observation of facial expressions. A dual-attention fusion of facial landmark features for Vision Transformer expression recognition method called DualAtt-ViT is proposed.

For the research of classroom emotion recognition of students based on behavioral gestures, a behavioral recognition method of R(2+1)D-Transformer with fusion of auxiliary loss, named R(2+1)D-Transformer-AL, is proposed.

Focusing on the identification of students’ affective states in the classroom environment, this study constructs a comprehensive analytical framework for classroom affective identification by analyzing two key dimensions: students’ facial expressions and behavioral gestures. To quantify emotional states, the PAD Emotion Scale was used to assign values and quantitatively assess students’ facial expressions and behaviors.

Considering that facial expressions and behaviors may have different importance in expressing emotions, their influence is adjusted by designing a fully connected network and training the weights in the decision fusion process. Therefore, different behaviors and expressions may have different weights when calculating the overall emotion value. Ultimately, the overall average emotion value of students is calculated using the following formula: (22) EOverall=∑ wi×Ei∑ wi \[{{E}_{Overall}}=\frac{\sum{{{w}_{i}}}\times {{E}_{i}}}{\sum{{{w}_{i}}}}\] where E_i represents the affective value of a single behavior or facial expression, w_i represents the corresponding weight, and the overall affective value E_Overall represents the students’ classroom affective state.

The application object is chosen to be a class of a vocational college with a total of 48 students, of which 22 are female and 26 are male, and the class is divided into four groups on average.

Hardware facilities for data collection are deployed in the class, so the experimental environment is in a real teaching scenario (recording classroom), with two HD surveillance cameras in the front and back of the recording classroom as the global filming tools, one at the front of the classroom to collect the students’ expression and gesture data from the frontal perspective of the students, and the other at the back of the classroom as the auxiliary of the front-end camera to collect the students’ body gesture data and the The other one is at the back of the classroom, as the auxiliary of the front-end camera, to collect the students’ body posture data and the expression data that cannot be captured by the front-end camera, thus improving the accuracy of the recognition and avoiding the occurrence of omission and misdetection. In addition, in order to further analyze the mood and learning state of each group, this study added a mobile filming device at the right rear end of the classroom for real-time filming and recording.

At the end of the course, the classroom video captured by the HD camera is imported into a data acquisition and processing server, which converts the video into an image sequence and uses a facial expression and behavioral gesture fusion algorithm based on intelligent technology to realize the intelligent recognition of students’ emotions, which is provided to the teacher for post-course reflection, introspection, and evaluation.

The teaching process in vocational schools is divided into five stages: problem introduction, activity exploration, work presentation, outcome evaluation and program adjustment. Based on the emotional recognition of students’ facial expressions and body postures, we use data visualization methods, such as “emotion columns”, to present students’ emotional experiences during the problem solving process, and thus analyze the students’ emotional changes and learning status during the problem solving process.

1)

Results of learning emotion detection

The distribution of the percentage of students’ emotions in each stage is shown in Figure 2, and the trend of the seven emotions in each stage is shown in Figure 3. In each stage of the learning process students’ emotional bias is different, in the initial stage of the classroom teacher for the introduction of the problem, students show more average emotions, mainly neutral (43.4%), that is, the normal natural state of listening to the teacher’s lectures, the student’s emotions are more often shown as happy (29.2%) and surprised (31.5%), indicating that the teacher’s introduction of the content of the teacher to cause curiosity and interest in the students, and thus engaged in listening to the lectures. Thus, they were engaged in listening to the lecture.

Figure 2.

The proportion of students’ emotions at all stages

Figure 3.

The changes in the seven emotions at various stages

In the second stage of activity exploration, students’ positive emotions increased (32.4%) and negative emotions decreased (23.4%), and more emotions appeared as neutral (24.9%), happy (33.5%), and angry (17.1%), which showed that students needed to divide the task and discuss with their group members to design the model initially in activity exploration, and in this process, there might be a difference of opinion between students, as well as an interest in completing the task. Differences of opinion, but also full of fighting spirit to complete the task, emotions are high, so in this stage students’ learning emotions are the most abundant.

The third and fourth stages of student emotions show a negative bias, negative emotions are more than 38%, specifically to frustration, surprise and neutral, combined with the actual teaching situation found that this stage of the students will be completed works on display, the teacher and the other groups to evaluate it, for the questioning of others, students are prone to produce a sense of frustration, emotionally depressed, for the praise of others, students are prone to Students are prone to be surprised and happy when they are praised by others.

In the fifth stage, students need to design and readjust their works according to the suggestions given by others, and the teacher also provides and explains more optimized problem-solving tools to help students improve their tasks, in which the percentage of students’ negative emotions decreases significantly (28.5%) and the neutral emotions increase (41.2%), which indicates that students are once again engaged in the design activities, and the happy emotions increase significantly.

Specifically, in terms of the seven categorical mood dimensions, neutral and happy moods had the highest percentages, averaging about 23% and 22%, respectively, reflecting the fact that most of the students were not in a high or low mood, and most of the time were in a smoother learning state.

2)

Statistical analysis of learning status

According to the correspondence between emotion and learning state, the average value of learning state of students in each group in each teaching session was counted, and Figure 4 shows the trend of students’ learning state in each group. From an overall point of view, the learning state of the whole class shows a relatively smooth trend of change, with a lower learning state score (7.92) during independent inquiry learning and a significant increase (8.00) in the learning state when the work is displayed for teacher’s evaluation, which indicates that appropriate participation and intervention by the teacher can effectively improve the learning state of students and mobilize their motivation for learning.

Figure 4.

Changes in the learning state of students in each group

The overall learning state of Group 1 was poor, and the group fell into a low state in the second stage of the activity exploration session (7.63).Group 2 had a greater fluctuation in the change of learning state, and presented a high learning state in the process of completing the task on their own and sharing with their peers (8.33), which can be deduced that the students in this group were more motivated in their independent learning, and their learning was more effective.Group 3 tended to flatten out during the entire Group 4 showed a downward trend from the initial state of good, and then slightly improved. It is found that students in different groups have different learning characteristics in each learning stage. Teachers can find out the learning styles of students according to the learning status of students in each course, and then adjust the subsequent teaching strategies and contents in a targeted way to improve the professional development of teachers.

A potentially fundamental premise of factor analysis is that there should be strong correlations between the original variables, otherwise it is not possible to construct fewer common factors reflecting the common characteristics of certain variables. The Bartlett’s test of sphericity and the KMO method are usually used to test whether the original variables are amenable to factor analysis.

Among the methods for solving factor loading matrix, principal component analysis based on principal component model is the most widely used, which transforms the original variables into another set of uncorrelated variables, i.e. principal components, by making linear changes through coordinate transformation.

Its mathematical model is: (25) { z1=μ11x1+μ12x2+μ13x3+…+μ1pxpz2=μ21x1+μ22x2+μ23x3+…+μ2pxpz3=μ31x1+μ32x2+μ33x3+…+μ3pxp⋮ zk=μp1x1+μp2x2+μp3x3+…+μppxp \[\begin{array}{*{35}{l}} \left\{ \begin{array}{*{35}{l}} {{z}_{1}}={{\mu }_{11}}{{x}_{1}}+{{\mu }_{12}}{{x}_{2}}+{{\mu }_{13}}{{x}_{3}}+\ldots +{{\mu }_{1p}}{{x}_{p}} \\ {{z}_{2}}={{\mu }_{21}}{{x}_{1}}+{{\mu }_{22}}{{x}_{2}}+{{\mu }_{23}}{{x}_{3}}+\ldots +{{\mu }_{2p}}{{x}_{p}} \\ {{z}_{3}}={{\mu }_{31}}{{x}_{1}}+{{\mu }_{32}}{{x}_{2}}+{{\mu }_{33}}{{x}_{3}}+\ldots +{{\mu }_{3p}}{{x}_{p}} \\ \vdots \\ \end{array} \right. \\ {{z}_{k}}={{\mu }_{p1}}{{x}_{1}}+{{\mu }_{p2}}{{x}_{2}}+{{\mu }_{p3}}{{x}_{3}}+\ldots +{{\mu }_{pp}}{{x}_{p}} \\ \end{array}\]

Of these, μi12+μi22+μi32+…+μip2=1(i=1,2,3,…p) . zi\[\mu _{i1}^{2}+\mu _{i2}^{2}+\mu _{i3}^{2}+\ldots +\mu _{ip}^{2}=1\left( i=1,2,3,\ldots p \right)\text{ }.\text{ }{{z}_{i}}\] and z_j(i ≠ j;i,j = 1,2,3,⋯p) are independent of each other. z₁ is the most human in variance among all linear combinations of x₁,x₂,x_i,⋯x_p and has the strongest ability to synthesize the original variables, and z₂,z₃,⋯z_k the ability to synthesize the original variables decreases in that order.

The factor loading matrix is the p eigenroots λ₁,λ₂,λ₃,⋯λ_p of the matrix of correlation coefficients in the process of solving the coefficients of the principal component mathematical model. and the corresponding eigenvectors of the standard orthogonal.

In order to reduce the number of variables, the number of factors in the mathematical model of factor analysis k should be smaller than the original number of variables p. Therefore, when calculating the factor loading matrix, only the first k eigenvalues and eigenvectors are selected.

To determine the number of common factors k: i. Select the common factor corresponding to the eigenroot with an eigenroot value greater than 1: ii. Select the number of eigenroots with a cumulative variance contribution greater than 85% as the number of factors k by calculating the variance contribution (which is equal to the sum of squares of the elements of the jth column in the factor loading matrix Z) and cumulative variance contribution of the common factor z_j. Cumulative variance contribution of the first k factors is defined as: (28) ak=∑i=1kSi2p=∑i=1kλi∑i=1Nλi \[{{a}_{k}}=\frac{\sum\limits_{i=1}^{k}{S_{i}^{2}}}{p}=\frac{\sum\limits_{i=1}^{k}{{{\lambda }_{i}}}}{\sum\limits_{i=1}^{N}{{{\lambda }_{i}}}}\]

The public factor variables obtained after principal component analysis are new variables obtained after linear transformation and their meaning needs to be defined.

A row in loading matrix Z indicates the correlation between the original variable Xi and the factor variables. A column in the loading matrix Z indicates the amount of information that a factor variable can explain about the original variable Xi. Sometimes the factor loading matrix is not well interpreted, and factor rotation is often needed to make the original factor variables more interpretable.

Factor rotation is the right-multiplication of the factor loading matrix Z by an orthogonal matrix to obtain a new matrix B. Without affecting the covariance hi2\[h_{i}^{2}\] of the variables Xi, the variance contribution Si2\[S_{i}^{2}\] of the factors is changed. The factor loadings polarize either close to 0 or 1, thus making the original factor variables more interpretable.

After determining the factor variables, what needs to be obtained are the specific data values for the different common factors, which are the specific values of each factor on each sample. The score function for the jst factor is: (29) Fj=ϖμ1x1+ϖj2x2+ϖj3x3+⋯+ϖjpxp(j=1,2,3…k) \[{{F}_{j}}={{\varpi }_{\mu 1}}{{x}_{1}}+{{\varpi }_{j2}}{{x}_{2}}+{{\varpi }_{j3}}{{x}_{3}}+\cdots +{{\varpi }_{jp}}{{x}_{p}}\left( j=1,2,3\ldots k \right)\]

Comprehensive evaluation is usually done using a factor-weighted total score, which is calculated using the formula: (30) F=a1F1+a2F2+a3F3+⋯+akFk \[F={{a}_{1}}{{F}_{1}}+{{a}_{2}}{{F}_{2}}+{{a}_{3}}{{F}_{3}}+\cdots +{{a}_{k}}{{F}_{k}}\] where a₁,a₂,a₃,…a_k is the weight of the corresponding factor.

The results of the recovered valid questionnaires were entered into SPSS for factor analysis to obtain the descriptive statistics of the data. The statistical results of the sample teachers’ competency profile elements are shown in Fig. 5. The average value of the importance of the 25 teachers’ competency profile elements fluctuates between 3.25 and 4.30, and the importance of all the elements is above the general level, from which it can be seen that these 25 teachers’ competency profile elements derived through the literature screening method as well as the questionnaire survey method are scientifically sound and reasonable. In addition, the standard deviation of the 25 elements of teacher competence fluctuates between 0.30 and 0.79, and the standard deviation of each element of competence is not very large and roughly comparable, so it can be seen that there is consistency in the scoring of the same element by the subjects, which strengthens the credibility of the results of the study.

Figure 5.

The statistical results of the ability characteristics of the sample teachers

In order to explore the intrinsic structure of these vocational ability characteristic elements, this study adopts principal component analysis to extract the main factors, orthogonal rotation by maximum variance method to extract the factors with initial eigenvalue greater than 1, and the final number of extracted public factors is determined to be 5. The eigenvalues and variance contributions of the public factors are shown in Table 2. The rotated factor structure is effectively simplified: the total variance explained by public factors F1~F5 is 35.1%, 26.4%, 12.5%, 8.4% and 4.3%, and the cumulative variance contribution rate reaches 86.7%.

The factor component matrix reflects the loading of the factor on the original variables, and Table 3 shows the component matrix of the teacher competence factors. In the first factor, information updating ability, educational research ability, social service ability, combination of theory and practice, self-study ability, organization and management ability, and academic communication ability, which are variables mainly reflecting the research and professional development ability of teachers in vocational schools, have high loadings on the first factor (0.798~0.961), and can be named as research and professional development ability.

The variables of teaching design ability, teaching method ability, teaching management ability, advanced vocational teaching concept, language expression ability, and practical teaching skills have loadings greater than 0.834 on the second factor, which mainly reflect the vocational teaching ability of teachers in vocational schools, and are named as vocational teaching ability. Similarly, the third, fourth and fifth factors were named as professional practice ability, personal traits, and professional quality, respectively.

Based on the results of the above factor analysis, this study divided the 25 teacher competency profile elements into five dimensions, and the total variable scores of the corresponding common factors were obtained based on the mean scores of each element derived from the SPSS descriptive analysis. Finally, the weights of each teacher competency profile element were calculated separately based on the average of the scores of the competency profile elements corresponding to each teacher competency profile element dimension. The weights of the teacher competency profile elements are shown in Table 4. The weights of scientific research and professional development ability, personal attributes, vocational teaching ability, professional quality and professional practice ability are 19.98%, 20.56%, 20.93%, 18.39% and 20.14%, respectively, of which the vocational teaching ability has the greatest influence on the improvement of teachers’ competence in vocational schools.

The government and enterprises should play a leading role in establishing a big data platform to monitor teachers’ professional practice ability and provide data support for the assessment of teachers’ ability in vocational education. Artificial intelligence experts should be organized to assess the actual situation of teachers’ competence in vocational colleges and introduce corresponding improvement measures. Schools regularly send professional teachers and backbone teachers to work with employees at the production line of enterprises as a way to understand industry dynamics, collect typical work cases, and improve practical teaching ability. Backbone teachers can also make use of the policies provided by the government and vocational institutions to set up on-campus master teacher studios.

In terms of cultivating and improving teachers’ intelligent vocational teaching ability, vocational colleges and universities should formulate a strategic plan for the in-depth integration of artificial intelligence and vocational education, establish a perfect teacher training system, enrich the opportunities for developing teachers’ professional skills, and urge school teachers to participate in more of such relevant national or provincial training programs. Secondly, strengthen the training of teachers’ information technology skills, and develop a targeted and generalized training program for professional teachers’ information technology skills. Teachers should make use of the dynamic, interactive and multimedia features of “Internet +” to actively participate in the teaching of online courses. In addition, they should combine teaching and research, adopt interdisciplinary teaching methods, and give full play to teachers’ intelligent and innovative abilities.

Firstly, teachers themselves should formulate career development plans to improve their professional abilities and promote professional growth in a timely manner according to the school’s training plan and implementation program. Secondly, teachers can improve their basic teaching skills by observing public classes of experienced teachers or online classes of famous teachers. They can also improve their professional theoretical and practical skills by participating in skills competitions. Finally, teachers should utilize their spare time to expand their knowledge, conduct comprehensive research on teaching theories and educational methods, and enrich their deep understanding of vocational education.