English teaching classroom interaction mode analysis method based on big data

Abstract

The invention relates to the technical field of education analysis, and particularly discloses an English teaching classroom interaction mode analysis method based on big data, and the method comprises the steps: data collection and preprocessing, behavior feature extraction, behavior mode mining, classroom interaction mode analysis and application decision. According to the scheme, data of multiple modes of videos, audios and texts are combined, behavior characteristics of teachers and students in an English teaching classroom are automatically extracted and clustered, data changes are flexibly adapted, more accurate behavior modes are mined through hierarchical clustering and dynamic cutting, and the analysis efficiency is improved; the method comprises the following steps: analyzing an interaction mode of an English teaching classroom by using an HMM model, calculating a posterior probability of each moment, describing uncertainty of a state by using probability distribution, and capturing dynamic change of a behavior mode through modeling time dependence, so as to more accurately describe time sequence characteristics of classroom behaviors.

Description

Technical Field

[0001] The present invention relates to the field of educational analysis technology, and specifically to a method for analyzing English teaching classroom interaction patterns based on big data.

[0002] This application claims priority. The application number of the prior application is: 2024118949185, title: A method for analyzing English teaching classroom interaction patterns based on big data, priority date: December 21, 2024. Background Art

[0003] The big data-based English teaching classroom interaction pattern analysis method is a technology that optimizes teaching methods and strategies by collecting and analyzing various interaction data between students and teachers in English classrooms. Traditional English teaching classroom data lacks effective processing of multimodal data and cannot fully combine the discrete degree and distribution of behavioral characteristics for analysis, resulting in poor automatic capture of complex behavioral patterns. General behavioral pattern analysis methods usually only consider independently distributed data points and cannot handle the dynamic changes of behavioral patterns over time in classroom interactions. Moreover, due to the uncertainty of the state of classroom interaction patterns, the influence of noise and error is easily ignored in modeling and analysis. Summary of the Invention

[0004] In response to the above situation, in order to overcome the shortcomings of the existing technology, the present invention provides an English teaching classroom interaction pattern analysis method based on big data. Traditional English teaching classroom data lacks effective processing of multimodal data and cannot fully combine the discreteness and distribution of behavioral characteristics for analysis, resulting in poor automatic capture of complex behavioral patterns. This solution combines data from multiple modalities such as video, audio, and text to automatically extract and cluster the behavioral characteristics of teachers and students in English teaching classrooms, flexibly adapt to data changes, and mine more accurate behavioral patterns through hierarchical clustering and dynamic cutting, thereby improving analysis efficiency. General behavioral pattern analysis methods usually only consider independently distributed data points and cannot handle the dynamic changes of behavioral patterns over time in classroom interactions. Due to the uncertainty of the state of classroom interaction patterns, the influence of noise and error is easily ignored in modeling and analysis. This solution uses the HMM (Hidden Markov Model) to analyze the interaction pattern of English teaching classrooms, calculates the posterior probability of each moment, and uses probability distribution to describe the uncertainty of the state. By modeling time dependency, the dynamic changes of behavioral patterns are captured, thereby more accurately describing the temporal characteristics of classroom behavior.

[0005] The technical solution adopted by the present invention is as follows: The present invention provides a method for analyzing English teaching classroom interaction patterns based on big data, the method comprising the following steps:

[0006] Step S1: Data collection and preprocessing: Collect interactive data between teachers and students in English classrooms, including video data, audio data, and text data. Preprocess the interactive data and construct multimodal data according to class hours. Aggregate the multimodal data of all class hours and divide them into training and test sets.

[0007] Step S2: Behavioral feature extraction: extracting the behavioral features of teachers and students from the training set. The behavioral features are composed of action feature vectors, audio feature vectors, and semantic feature vectors.

[0008] Step S3: Behavioral pattern mining: using a hierarchical clustering algorithm to group all behavioral features extracted from the training set, and constructing behavioral patterns based on the grouping of behavioral features;

[0009] Step S4: Analyze the classroom interaction pattern, build an HMM model to analyze the correlation between the behavioral features in each behavior pattern, create a hidden state to represent the interaction pattern of the English teaching classroom, and output the state transition probability matrix and observation probability matrix;

[0010] Step S5: Apply the decision to predict the change of the behavior pattern based on the state transition probability matrix and the observation probability matrix, and output the future state of the teacher-student interaction in the English teaching classroom.

[0011] Furthermore, in step S2, the behavior feature extraction specifically includes the following steps:

[0012] Step S21: using the target detection model to detect the target object in the video data, identify the target object's motion changes, obtain the time period corresponding to each motion change, and construct a motion feature vector for the target object based on the time period corresponding to each motion change;

[0013] Step S22: Using a speech recognition model to analyze the speech duration, speaking rate, and emotional state of each target subject in the audio data, mapping the speech duration of the target subject to the time period of its action feature vector, and constructing the speaking rate and emotional state into the audio feature vector of the target subject;

[0014] Step S23: Use the pre-trained language model to analyze the content and semantics of the text data, extract classroom interaction topics and keywords, introduce the sentiment dictionary, analyze the emotional tendency of the target object by matching the sentiment vocabulary, and construct a semantic feature vector corresponding to the time period of its action feature vector;

[0015] Step S24: Check whether the action feature vector, audio feature vector, and semantic feature vector are aligned, and construct a feature matrix to represent the behavior features of the target object in each time period.

[0016] Furthermore, in step S3, the behavior pattern mining specifically includes the following steps:

[0017] Step S31: similarity calculation, using a distance metric to evaluate the similarity between all behavioral features in the training set;

[0018] Step S32: Hierarchical clustering, starting from a single behavioral feature, merging the most similar behavioral features from bottom to top, generating a new cluster each time the features are merged, and generating a cluster tree after all behavioral features are merged;

[0019] Step S33: Calculate the intra-cluster variance using the following formula: ;

[0020] Where, represents a cluster, represents the index of the cluster, Represents a cluster The variance within Represents a cluster The behavioral characteristics, Represents a cluster The center point of all behavioral features within Represents the Euclidean distance between the behavioral feature and the center point;

[0021] Step S34: Select the number of clusters and calculate the total variance of the cluster tree based on the number of clusters. When the total variance of the cluster tree is minimized, the number of clusters reaches the optimal number. The cluster tree is cut into subclusters based on the optimal number of clusters. The formula used is as follows: ;

[0022] Where, represents the total variance of the cluster tree, represents the number of clusters;

[0023] Step S35: constructing a behavior pattern, calculating the mean of the behavior features contained in the same level in each sub-cluster, and using the mean of the behavior features of all levels to construct the behavior pattern of the sub-cluster;

[0024] Step S36: Evaluate the effectiveness of the model. Calculate the average value of the silhouette coefficients of all subclusters and set a silhouette coefficient threshold. When the average value of the silhouette coefficients reaches the threshold, the behavior model is valid. If the average value of the silhouette coefficients does not reach the threshold, adjust the number of clusters and re-cut the cluster tree. The formula used is as follows: ;

[0025] Where, represents a subcluster, Represents a subcluster The silhouette coefficient, Represents a subcluster The average distance between behavioral features within a Represents a subcluster The average distance to other subclusters, Represents the maximum function.

[0026] Furthermore, in step S4, the classroom interaction mode analysis specifically includes the following steps:

[0027] Step S41: Initialize parameters and build an HMM model, including hidden states, a set of behavioral features in the behavioral pattern, a state transition probability matrix, and an observation probability matrix. Define the hidden states of the classroom interaction pattern based on prior knowledge and initialize the probability of each hidden state.

[0028] Step S42: Counting the hidden state sequence of each behavior pattern in the training set and establishing a state transition probability matrix;

[0029] Step S43: Calculate the observation probability of specific behavioral features in each hidden state and establish an observation probability matrix;

[0030] Step S44: Obtain the time period of the behavioral features contained in each behavioral pattern, correspond to the specific time of the English teaching class, and use the forward-backward algorithm on the training set to calculate the probability distribution of the hidden state at all times. The formula used is as follows: ;

[0031] Where, express Hidden state at all times The forward probability of express Hidden state at all times The backward probability of express Hidden state at all times The posterior probability of Represents the total probability of a behavior pattern;

[0032] Step S45: Update the state transition probability matrix and observation probability matrix according to the hidden state probability distribution at all times;

[0033] Step S46: Test and optimize the state transition probability matrix and the observation probability matrix on the test set to verify the accuracy of the HMM model in analyzing the interaction pattern in the actual English teaching classroom.

[0034] The beneficial effects achieved by the present invention using the above scheme are as follows:

[0035] (1) Traditional English teaching classroom data lacks effective processing of multimodal data and cannot fully combine the discreteness and distribution of behavioral characteristics for analysis, resulting in poor automatic capture of complex behavioral patterns. This solution combines video, audio, and text data in multiple modalities to automatically extract and cluster the behavioral characteristics of teachers and students in English teaching classrooms, flexibly adapt to data changes, and mine more accurate behavioral patterns through hierarchical clustering and dynamic cutting, thereby improving analysis efficiency.

[0036] (2) General behavior pattern analysis methods usually only consider independently distributed data points and cannot handle the dynamic changes of behavior patterns over time in classroom interactions. In addition, due to the uncertainty of the state of classroom interaction patterns, it is easy to ignore the impact of noise and error in modeling analysis. This scheme uses the HMM model to analyze the interaction pattern of English teaching classes, calculates the posterior probability at each moment, and uses probability distribution to describe the uncertainty of the state. By modeling time dependence, the dynamic changes of behavior patterns are captured, thereby more accurately describing the temporal characteristics of classroom behavior. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 This is a flow chart of a method for analyzing interactive patterns in English teaching classes based on big data proposed by the present invention;

[0038] Figure 2 Schematic diagram of the process of step S2.

[0039] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention. DETAILED DESCRIPTION

[0040] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0041] Example 1, see Figure 1 The present invention provides a method for analyzing English teaching classroom interaction patterns based on big data, which includes the following steps:

[0042] Step S1: Data collection and preprocessing: Use IoT teaching devices to collect interactive data between teachers and students in English classrooms, including video data, audio data, and text data. The interactive data is cleaned and structured, and multimodal data is constructed according to class hours. The multimodal data of all class hours are pooled, and 60% of them are randomly selected as the training set and 40% as the test set.

[0043] Step S2: Behavioral feature extraction: extracting the behavioral features of teachers and students from the training set. The behavioral features are composed of action feature vectors, audio feature vectors, and semantic feature vectors.

[0044] Step S3: Behavioral pattern mining: using a hierarchical clustering algorithm to group all behavioral features extracted from the training set, and constructing behavioral patterns based on the grouping of behavioral features;

[0045] Step S4: Analyze the classroom interaction pattern, build an HMM model to analyze the correlation between the behavioral features in each behavior pattern, create a hidden state to represent the interaction pattern of the English teaching classroom, and output the state transition probability matrix and observation probability matrix;

[0046] Step S5: Apply the decision to predict the change of the behavior pattern based on the state transition probability matrix and the observation probability matrix, and output the future state of the teacher-student interaction in the English teaching classroom.

[0047] Example 2, see Figure 1 This embodiment is based on the above embodiment. In step S1, video data is obtained through a camera to record the actions of teachers and students; audio data is obtained through an audio sensor to record the voice interaction between teachers and students; text data is obtained through the Internet of Things teaching platform, including classroom teaching content, student answers, and teacher-student dialogue texts.

[0048] Example 3, see Figure 1 This embodiment is based on the above embodiment. In step S2, behavior feature extraction specifically includes the following steps:

[0049] Step S21: Use the YOLO target detection model to detect the target object in the video data, identify the target object's movement changes, obtain the time period corresponding to each movement change, the movement changes include gesture changes, facial expression changes, and movement path changes, and construct a movement feature vector for the target object based on the time period corresponding to each movement change;

[0050] Step S22: Use the DeepSpeech speech recognition model to analyze the speech duration, speaking rate, and emotional state of each target subject in the audio data, map the speech duration of the target subject to the time period of its action feature vector, and construct the speaking rate and emotional state into the audio feature vector of the target subject;

[0051] Step S23: Use the pre-trained language model to analyze the content and semantics of the text data, extract classroom interaction topics and keywords, introduce the sentiment dictionary, analyze the emotional tendency of the target object by matching the sentiment vocabulary, and construct a semantic feature vector corresponding to the time period of its action feature vector;

[0052] Step S24: Check whether the action feature vector, audio feature vector, and semantic feature vector are aligned, and construct a feature matrix to represent the behavior features of the target object in each time period.

[0053] Example 4, see Figure 1 and Figure 2 This embodiment is based on the above embodiment. In step S3, behavior pattern mining specifically includes the following steps:

[0054] Step S31: Similarity calculation, using Manhattan distance to calculate the similarity between all behavioral features in the training set;

[0055] Step S32: Hierarchical clustering, starting from a single behavioral feature, gradually merges the two behavioral features with the smallest Manhattan distance from bottom to top, generating a new cluster each time the features are merged, and generating a cluster tree after all behavioral features are merged;

[0056] Step S33: Calculate the intra-cluster variance using the following formula: ;

[0057] Where, represents a cluster, represents the index of the cluster, Represents a cluster The variance within Represents a cluster The behavioral characteristics, Represents a cluster The center point of all behavioral features within Represents the Euclidean distance between the behavioral feature and the center point;

[0058] Step S34: Select the number of clusters and calculate the total variance of the cluster tree based on the number of clusters. When the total variance of the cluster tree is minimized, the number of clusters reaches the optimal number. The cluster tree is cut into subclusters based on the optimal number of clusters. The formula used is as follows: ;

[0059] Where, represents the total variance of the cluster tree, represents the number of clusters;

[0060] Step S35: constructing a behavior pattern, calculating the mean of the behavior features contained in the same level in each sub-cluster, and using the mean of the behavior features of all levels to construct the behavior pattern of the sub-cluster;

[0061] Step S36: Evaluate the effectiveness of the model. Calculate the average value of the silhouette coefficients of all subclusters and set the silhouette coefficient threshold to 0.6. When the average value of the silhouette coefficients reaches the threshold, the behavior model is valid. If the average value of the silhouette coefficients does not reach the threshold, adjust the number of clusters and re-cut the cluster tree. The formula used is as follows: ;

[0062] Where, represents a subcluster, Represents a subcluster The silhouette coefficient, Represents a subcluster The average distance between behavioral features within a Represents a subcluster The average distance to other subclusters, Represents the maximum function.

[0063] By performing the above operations, traditional English teaching classroom data lacks effective processing of multimodal data and cannot fully combine the discreteness and distribution of behavioral characteristics for analysis, resulting in poor automatic capture of complex behavioral patterns. This solution combines data from multiple modalities such as video, audio, and text to automatically extract and cluster the behavioral characteristics of teachers and students in English teaching classrooms, flexibly adapt to data changes, and mine more accurate behavioral patterns through hierarchical clustering and dynamic shearing, thereby improving analysis efficiency.

[0064] Example 5, see Figure 1 This embodiment is based on the above embodiment. In step S4, classroom interaction mode analysis specifically includes the following steps:

[0065] Step S41: Parameter initialization, constructing an HMM model, including hidden states, a set of behavioral features in the behavioral pattern, a state transition probability matrix, and an observation probability matrix. Based on prior knowledge, the hidden states of the classroom interaction pattern are defined, and the probability of each hidden state is initialized. Three hidden states are set: oral interaction, independent practice, and cooperative interaction.

[0066] Step S42: Count the hidden state sequence of each behavior pattern in the training set and establish a state transition probability matrix. The formula used is as follows: ;

[0067] Where, represents the state transition probability matrix, 、 、 Represents three different hidden states, represents the state transition probability;

[0068] Step S43: Calculate the observation probability of the specific behavior feature under each hidden state and establish an observation probability matrix. The formula used is as follows: ;

[0069] Where, represents the observation probability matrix, 、 、 Indicates behavioral characteristics, Represents the observation probability of specific behavioral features in the hidden state;

[0070] Step S44: Obtain the time period of the behavioral features contained in each behavioral pattern, correspond to the specific time of the English teaching class, and use the forward-backward algorithm on the training set to calculate the probability distribution of the hidden state at all times;

[0071] Step S45: Update the state transition probability matrix and observation probability matrix according to the hidden state probability distribution at all times;

[0072] Step S46: Test and optimize the state transition probability matrix and the observation probability matrix on the test set to verify the accuracy of the HMM model in analyzing the interaction pattern in the actual English teaching classroom.

[0073] By performing the above operations, general behavior pattern analysis methods usually only consider independently distributed data points and cannot handle the dynamic changes of behavior patterns over time in classroom interactions. In addition, due to the uncertainty of the state of classroom interaction patterns, the influence of noise and error is easily ignored in modeling analysis. This solution uses the HMM model to analyze the interaction pattern of English teaching classes, calculates the posterior probability at each moment, and uses probability distribution to describe the uncertainty of the state. By modeling time dependency, the dynamic changes of behavior patterns are captured, thereby more accurately describing the temporal characteristics of classroom behavior.

[0074] Example 6, see Figure 1 This embodiment is based on the above embodiment. In step S44, the probability distribution of hidden states at all times is calculated, specifically including the following steps:

[0075] Step S441: Calculate the forward probability of each hidden state at the initial moment. The formula used is as follows: ;

[0076] Where, represents the initial hidden state The forward probability of Represents the hidden state The initial probability of Represents the hidden state Downward behavioral characteristics The probability of observation;

[0077] Step S442: Calculate the time from the initial moment to The cumulative forward probability of all hidden states at the moment is as follows: ;

[0078] Where, express Hidden state at all times The forward probability of represents the total number of hidden states, Represents the hidden state To hidden state The transition probability, In the hidden state Downward behavioral characteristics The probability of observation;

[0079] Step S443: Define the backward probability of each hidden state at the end time. The formula used is as follows: ;

[0080] Where, Represents the hidden state The backward probability at the end time, Indicates the end time;

[0081] Step S444: Recursively calculate from the end time to The cumulative backward probability of all hidden states at the moment is as follows: ;

[0082] Where, express Hidden state at all times The backward probability of In the hidden state Downward behavioral characteristics The observation probability of express Hidden state at all times The backward probability of

[0083] Step S445: Combine the forward probability and the backward probability to obtain the probability distribution of the hidden state at all times. The formula used is as follows: ;

[0084] Where, express Hidden state at all times The posterior probability of Represents the sum of the observation probabilities of behavioral features in all hidden states.

[0085] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0086] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

[0087] The present invention and its embodiments are described above. This description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if a person skilled in the art is inspired by this and, without departing from the purpose of the present invention, designs structures and embodiments similar to this technical solution without inventiveness, they shall fall within the scope of protection of the present invention.

Claims

1. A method for analyzing interactive patterns in English teaching classes based on big data, characterized by: The method comprises the following steps: Step S1: Data collection and preprocessing: Collect interactive data between teachers and students in English classrooms, including video data, audio data, and text data. Preprocess the interactive data and construct multimodal data according to class hours. Aggregate the multimodal data of all class hours and divide them into training and test sets. Step S2: Behavioral feature extraction: extracting the behavioral features of teachers and students from the training set. The behavioral features are composed of action feature vectors, audio feature vectors, and semantic feature vectors. Step S3: Behavioral pattern mining: using a hierarchical clustering algorithm to group all behavioral features extracted from the training set, and constructing behavioral patterns based on the grouping of behavioral features; Step S4: Analyze the classroom interaction pattern, build an HMM model to analyze the correlation between the behavioral features in each behavior pattern, create a hidden state to represent the interaction pattern of the English teaching classroom, and output the state transition probability matrix and observation probability matrix; Step S5: Apply the decision to predict the change of the behavior pattern based on the state transition probability matrix and the observation probability matrix, and output the future state of the teacher-student interaction in the English teaching classroom.

2. The method for analyzing interactive patterns in English teaching classes based on big data according to claim 1, characterized in that: In step S2, the behavior feature extraction includes the following steps: Step S21: using the target detection model to detect the target object in the video data, identify the target object's motion changes, obtain the time period corresponding to each motion change, and construct a motion feature vector for the target object based on the time period corresponding to each motion change; Step S22: Using a speech recognition model to analyze the speech duration, speaking rate, and emotional state of each target subject in the audio data, mapping the speech duration of the target subject to the time period of its action feature vector, and constructing the speaking rate and emotional state into the audio feature vector of the target subject; Step S23: Use the pre-trained language model to analyze the content and semantics of the text data, extract classroom interaction topics and keywords, introduce the sentiment dictionary, analyze the emotional tendency of the target object by matching the sentiment vocabulary, and construct a semantic feature vector corresponding to the time period of its action feature vector; Step S24: Check whether the action feature vector, audio feature vector, and semantic feature vector are aligned, and construct a feature matrix to represent the behavior features of the target object in each time period.

3. The method for analyzing interactive patterns in English teaching classes based on big data according to claim 1, characterized in that: In step S3, the behavior pattern mining includes the following steps: Step S31: similarity calculation, using a distance metric to evaluate the similarity between all behavioral features in the training set; Step S32: Hierarchical clustering, starting from a single behavioral feature, merging the most similar behavioral features from bottom to top, generating a new cluster each time the features are merged, and generating a cluster tree after all behavioral features are merged; Step S33: Calculate the intra-cluster variance using the following formula: ； Where, represents a cluster, represents the index of the cluster, Represents a cluster The variance within Represents a cluster The behavioral characteristics, Represents a cluster The center point of all behavioral features within Represents the Euclidean distance between the behavioral feature and the center point; Step S34: Select the number of clusters and calculate the total variance of the cluster tree based on the number of clusters. When the total variance of the cluster tree is minimized, the number of clusters reaches the optimal number. The cluster tree is cut into subclusters based on the optimal number of clusters. The formula used is as follows: ； Where, represents the total variance of the cluster tree, represents the number of clusters; Step S35: constructing a behavior pattern, calculating the mean of the behavior features contained in the same level in each sub-cluster, and using the mean of the behavior features of all levels to construct the behavior pattern of the sub-cluster; Step S36: Evaluate the effectiveness of the model. Calculate the average value of the silhouette coefficients of all subclusters and set a silhouette coefficient threshold. When the average value of the silhouette coefficients reaches the threshold, the behavior model is valid. If the average value of the silhouette coefficients does not reach the threshold, adjust the number of clusters and re-cut the cluster tree. The formula used is as follows: ； Where, represents a subcluster, Represents a subcluster The silhouette coefficient, Represents a subcluster The average distance between behavioral features within a Represents a subcluster The average distance to other subclusters, Represents the maximum function.

4. The method for analyzing interactive patterns in English teaching classes based on big data according to claim 1, characterized in that: In step S4, the classroom interaction pattern analysis includes the following steps: Step S41: Initialize parameters and build an HMM model, including hidden states, a set of behavioral features in the behavioral pattern, a state transition probability matrix, and an observation probability matrix. Define the hidden states of the classroom interaction pattern based on prior knowledge and initialize the probability of each hidden state. Step S42: Counting the hidden state sequence of each behavior pattern in the training set and establishing a state transition probability matrix; Step S43: Calculate the observation probability of specific behavioral features in each hidden state and establish an observation probability matrix; Step S44: Obtain the time period of the behavioral features contained in each behavioral pattern, correspond to the specific time of the English teaching class, and use the forward-backward algorithm on the training set to calculate the probability distribution of the hidden state at all times. The formula used is as follows: ； Where, express Hidden state at all times The forward probability of express Hidden state at all times The backward probability of express Hidden state at all times The posterior probability of Represents the total probability of a behavior pattern; Step S45: Update the state transition probability matrix and observation probability matrix according to the hidden state probability distribution at all times; Step S46: Test and optimize the state transition probability matrix and the observation probability matrix on the test set to verify the accuracy of the HMM model in analyzing the interaction pattern in the actual English teaching classroom.

Images