An adaptive teaching quality evaluation system based on machine learning

Abstract

Disclosed is an adaptive teaching quality evaluation system based on machine learning. First, the training operation monitoring video and the training operation reference video are divided into video segments, and then feature extraction is performed by a training operation semantic feature extractor to obtain a sequence of training operation local time domain semantic coding feature maps and a sequence of training operation reference local time domain semantic coding feature maps. Then, the training operation semantic difference measurement is performed on each group of corresponding training operation local time domain semantic coding feature maps and training operation reference local time domain semantic coding feature maps in the sequence of training operation local time domain semantic coding feature maps and the sequence of training operation reference local time domain semantic coding feature maps to obtain training operation difference semantic measurement features. Finally, based on the training operation difference semantic measurement features, it is determined whether the training operation meets the predetermined standard. In this way, the teaching quality and efficiency can be improved.

Description

Technical Field

[0001] The present application relates to the field of teaching quality assessment, and more specifically, to an adaptive teaching quality assessment system based on machine learning. Background Art

[0002] Teaching quality evaluation is an important topic in the field of education, which is related to the improvement of teaching effectiveness and the optimal allocation of teaching resources. The purpose of teaching quality evaluation is to give teachers and students timely and effective feedback to adjust teaching methods and improve learning strategies.

[0003] However, the evaluation of teaching quality in the prior art is usually carried out by testing students' mastery of theoretical knowledge, such as evaluating students' learning outcomes through test scores or homework completion. The teaching process requires not only teaching theoretical knowledge, but also practical operation teaching. For teaching that requires practice and operation, students' mastery of theoretical knowledge cannot represent the level of practical operation. In other words, it is impossible to fully and accurately evaluate students' practical operation level and teaching quality by simply testing theoretical knowledge.

[0004] Therefore, an optimized adaptive teaching quality assessment system is desired. Summary of the invention

[0005] In order to solve the above technical problems, this application is proposed. The embodiment of this application provides an adaptive teaching quality evaluation system based on machine learning, which can help teachers better understand students' actual operation ability, optimize the teaching process, improve teaching quality and efficiency, and promote the cultivation of students' practical ability and innovation ability.

[0006] According to one aspect of the present application, there is provided an adaptive teaching quality assessment system based on machine learning, which includes:

[0007] A training operation monitoring video acquisition module is used to obtain the training operation monitoring video collected by the camera;

[0008] A practical training operation reference video acquisition module is used to extract practical training operation reference videos from a database;

[0009] A video segmentation module, used to segment the practical training operation monitoring video and the practical training operation reference video into video segments to obtain a sequence of practical training operation monitoring video segments and a sequence of practical training operation reference video segments;

[0010] A video clip semantic feature extraction module, used for performing feature extraction on the sequence of the training operation monitoring video clips and the sequence of the training operation reference video clips respectively through a training operation semantic feature extractor based on a deep neural network model to obtain a sequence of training operation local time domain semantic coding feature maps and a sequence of training operation reference local time domain semantic coding feature maps;

[0011] A training operation semantic difference measurement module is used to perform training operation semantic difference measurement on each group of corresponding training operation local time domain semantic coding feature maps and training operation reference local time domain semantic coding feature maps in the sequence of the training operation local time domain semantic coding feature maps and the sequence of the training operation reference local time domain semantic coding feature maps to obtain training operation difference semantic measurement features; and

[0012] The training operation standardization detection module is used to determine whether the training operation meets the predetermined standards based on the difference semantic measurement characteristics of the training operation.

[0013] Compared with the prior art, the adaptive teaching quality evaluation system based on machine learning provided by the present application first divides the practical operation monitoring video and the practical operation reference video into video segments, and then extracts features through the practical operation semantic feature extractor to obtain a sequence of practical operation local time domain semantic coding feature maps and a sequence of practical operation reference local time domain semantic coding feature maps, then performs practical operation semantic difference measurement on each group of corresponding practical operation local time domain semantic coding feature maps and practical operation reference local time domain semantic coding feature maps in the sequence of practical operation local time domain semantic coding feature maps and the sequence of practical operation reference local time domain semantic coding feature maps to obtain practical operation difference semantic measurement features, and finally, based on the practical operation difference semantic measurement features, determines whether the practical operation meets the predetermined standard. In this way, the teaching quality and efficiency can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. The following drawings are not intentionally scaled to the actual size, and the focus is on illustrating the main purpose of the present application.

[0015] Figure 1 It is a block diagram of an adaptive teaching quality assessment system based on machine learning according to an embodiment of the present application.

[0016] Figure 2 Flow chart of an adaptive teaching quality assessment method based on machine learning according to an embodiment of the present application.

[0017] Figure 3Schematic diagram of the system architecture of the adaptive teaching quality assessment method based on machine learning according to an embodiment of the present application.

[0018] Figure 4 This is an application scenario diagram of an adaptive teaching quality assessment system based on machine learning according to an embodiment of the present application. DETAILED DESCRIPTION

[0019] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, not all embodiments. Based on the embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without creative work also fall within the scope of protection of the present application.

[0020] As shown in this application and claims, unless the context clearly indicates an exception, the words "a", "an", "an" and / or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0021] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules can be used and run on the user terminal and / or server. The modules are only illustrative, and different aspects of the system and method can use different modules.

[0022] Flowcharts are used in the present application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps may be processed in reverse order or simultaneously as required. Meanwhile, other operations may also be added to these processes, or a certain step or several steps of operations may be removed from these processes.

[0023] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0024] In view of the above technical problems, the technical concept of the present application is to collect students' practical operation monitoring videos in real time through cameras for teaching practical operations, and introduce data processing and analysis algorithms based on machine learning at the back end to perform video semantic difference measurement analysis on the practical operation monitoring videos and standard practical operation reference videos, so as to judge whether the practical operation in the practical operation monitoring videos meets the predetermined standards, so as to perform adaptive teaching quality evaluation. In this way, it is possible to evaluate students' performance in practical operations more comprehensively and meticulously, so as to improve the accuracy of teaching quality evaluation, so as to facilitate the provision of personalized feedback and guidance based on the evaluation results. In this way, it can help teachers better understand students' practical operation capabilities, optimize the teaching process, improve teaching quality and efficiency, and promote the cultivation of students' practical and innovative capabilities.

[0025] Figure 1 FIG. 1 is a block diagram of an adaptive teaching quality evaluation system based on machine learning according to an embodiment of the present application. Figure 1 As shown, according to the embodiment of the present application, the adaptive teaching quality assessment system 100 based on machine learning includes: a practical training operation monitoring video acquisition module 110, which is used to obtain the practical training operation monitoring video collected by the camera; a practical training operation reference video acquisition module 120, which is used to extract the practical training operation reference video from the database; a video segmentation module 130, which is used to segment the practical training operation monitoring video and the practical training operation reference video into video segments to obtain a sequence of practical training operation monitoring video segments and a sequence of practical training operation reference video segments; a video segment semantic feature extraction module 140, which is used to extract the sequence of practical training operation monitoring video segments and the sequence of practical training operation reference video segments respectively through a practical training operation semantic feature extractor based on a deep neural network model. The sequence of training operation reference video clips is subjected to feature extraction to obtain a sequence of training operation local time-domain semantic coding feature maps and a sequence of training operation reference local time-domain semantic coding feature maps; a training operation semantic difference measurement module 150 is used to perform training operation semantic difference measurement on each group of corresponding training operation local time-domain semantic coding feature maps and training operation reference local time-domain semantic coding feature maps in the sequence of training operation local time-domain semantic coding feature maps and the sequence of training operation reference local time-domain semantic coding feature maps to obtain training operation difference semantic measurement features; and a training operation standardization detection module 160 is used to determine whether the training operation meets the predetermined standard based on the training operation difference semantic measurement features.

[0026] Specifically, in the technical solution of the present application, first, a practical training operation monitoring video captured by a camera is obtained, and a practical training operation reference video is extracted from a database. It should be understood that in the practical training operation monitoring video and the practical training operation reference video, a longer video may contain multiple different practical training operation steps or scenes, and each step or scene may have different semantic features. Therefore, in order to be able to extract the semantic features of each training segment more accurately and fine-grainedly, so as to capture the details of each training step or scene, in the technical solution of the present application, it is necessary to segment the practical training operation monitoring video and the practical training operation reference video into video segments to obtain a sequence of practical training operation monitoring video segments and a sequence of practical training operation reference video segments. In particular, by segmenting the practical training operation monitoring video and the practical training operation reference video into segments, their corresponding segments can be compared, thereby more accurately measuring the differences in practical training operations.

[0027] Then, the sequence of the practical operation monitoring video clips and the sequence of the practical operation reference video clips are respectively subjected to feature extraction in a practical operation semantic feature extractor based on a three-dimensional convolutional neural network model, so as to respectively extract the local time domain semantic feature information about the practical operation in each of the practical operation monitoring video clips and each of the practical operation reference video clips, thereby obtaining a sequence of local time domain semantic coding feature maps of the practical operation and a sequence of local time domain semantic coding feature maps of the practical operation reference.

[0028] Correspondingly, in the video segment semantic feature extraction module 140, the deep neural network model is a three-dimensional convolutional neural network model. It is worth mentioning that the three-dimensional convolutional neural network (3D CNN) is a deep learning model for processing video data and time series data. Unlike the traditional two-dimensional convolutional neural network (2D CNN), the three-dimensional convolutional neural network can consider time and space information at the same time. The three-dimensional convolutional neural network is very useful in video processing and can capture the time series features and motion information in the video. It models each frame in the video and the relationship between them by applying a three-dimensional convolution operation in the time dimension. In this way, the dynamic changes and time series relationships of the video can be taken into account when extracting features, so as to better understand the video content. In the adaptive teaching quality evaluation system based on machine learning, using the three-dimensional convolutional neural network as a deep neural network model, feature extraction can be performed on the practical operation monitoring video segment and the practical operation reference video segment. By learning the time series information and motion pattern in the video, the three-dimensional convolutional neural network can extract the local time domain semantic coding feature map, which is then used to measure the semantic differences of the practical operation and detect the standardization of the practical operation.

[0029] Furthermore, since each of the local time domain semantic coding feature maps of the practical operation respectively represents the semantic features of the student's practical operation in each local time domain, and each of the local time domain semantic coding feature maps of the practical operation reference respectively represents the semantic features of the standard practical operation in each local time domain. Therefore, in order to be able to more fully and meticulously measure the degree of semantic difference between the student's actual operation and the reference operation, in the technical solution of the present application, the semantic difference measurement coefficients of the practical operation of each group of corresponding local time domain semantic coding feature maps of the practical operation and the practical operation reference local time domain semantic coding feature maps in the sequence of the practical operation local time domain semantic coding feature maps and the sequence of the practical operation reference local time domain semantic coding feature maps are further calculated to obtain a practical operation difference semantic measurement feature vector composed of multiple practical operation semantic difference measurement coefficients. By calculating the semantic difference measurement coefficients of the practical operation of each group of corresponding local time domain semantic coding feature maps of the practical operation and the practical operation reference local time domain semantic coding feature maps, the degree of semantic difference of the local time domain practical operation between the actual operation and the reference operation can be quantified. Such measurement methods can help evaluate whether students' actual operations meet the predetermined standards more fully and comprehensively.

[0030] Correspondingly, the training operation semantic difference measurement module 150 is used to respectively calculate the training operation semantic difference measurement coefficients of each group of corresponding training operation local time domain semantic coding feature maps and training operation reference local time domain semantic coding feature maps in the sequence of the training operation local time domain semantic coding feature maps and the sequence of the training operation reference local time domain semantic coding feature maps to obtain a training operation difference semantic measurement feature vector composed of multiple training operation semantic difference measurement coefficients as the training operation difference semantic measurement feature.

[0031] It should be understood that the training operation semantic difference measurement coefficient refers to a numerical value used to measure the difference between the local time domain semantic coding feature map of the training operation and the reference local time domain semantic coding feature map of the training operation. It can be used to quantify the degree of semantic difference of the training operation, thereby evaluating the quality and standardization of the training operation. The calculation method of the training operation semantic difference measurement coefficient can be determined according to specific needs and scenarios. Generally, some distance or similarity measurement methods can be used to calculate the difference between the local time domain semantic coding feature map of the training operation and the reference local time domain semantic coding feature map of the training operation. Commonly used measurement methods include Euclidean distance, cosine similarity, correlation coefficient, etc. The larger the value of the training operation semantic difference measurement coefficient, the greater the semantic difference of the training operation; the smaller the value, the smaller the semantic difference of the training operation. By calculating the training operation semantic difference measurement coefficient, the degree of difference between each group of training operation local time domain semantic coding feature maps and the training operation reference local time domain semantic coding feature maps can be quantified, and then the training operation difference semantic measurement feature vector composed of multiple training operation semantic difference measurement coefficients can be obtained. This feature vector can be used to evaluate and compare the quality of the training operation, so as to determine whether the training operation meets the predetermined standards.

[0032] Specifically, the training operation semantic difference measurement module 150 is used to: calculate the training operation semantic difference measurement coefficients of each group of corresponding training operation local time domain semantic coding feature maps and training operation reference local time domain semantic coding feature maps in the sequence of the training operation local time domain semantic coding feature maps and the sequence of the training operation reference local time domain semantic coding feature maps respectively using the following coefficient calculation formula to obtain the training operation difference semantic measurement feature vector composed of multiple training operation semantic difference measurement coefficients; wherein the coefficient calculation formula is:

[0033]

[0034] Among them, p (x,y,z) and q (x,y,z) are the eigenvalues ​​of each position of each group of corresponding training operation local time domain semantic coding feature maps and training operation reference local time domain semantic coding feature maps in the sequence of the training operation local time domain semantic coding feature maps and the sequence of the training operation reference local time domain semantic coding feature maps, respectively; W is the width of each group of corresponding training operation local time domain semantic coding feature maps and training operation reference local time domain semantic coding feature maps; H is the height of each group of corresponding training operation local time domain semantic coding feature maps and training operation reference local time domain semantic coding feature maps; C is the number of channels of each group of corresponding training operation local time domain semantic coding feature maps and training operation reference local time domain semantic coding feature maps; Hp is the eigenvalue of each position in the training operation difference semantic metric feature vector; log represents the logarithmic function with base 2.

[0035] Then, the difference semantic measurement feature vector of the practical training operation is passed through a classifier to obtain a classification result, and the classification result is used to indicate whether the practical training operation meets the predetermined standard. In other words, the semantic difference feature information between the semantic features of the students' actual practical training operation in each local time domain and the semantic features of the standard reference practical training operation is used for classification processing to determine whether the practical training operation in the practical training operation monitoring video meets the predetermined standard, thereby performing an adaptive teaching quality evaluation. In this way, the performance of students in practical operations can be evaluated more comprehensively and meticulously to improve the accuracy of teaching quality evaluation, thereby facilitating the provision of personalized feedback and guidance based on the evaluation results.

[0036] Correspondingly, the training operation standardization detection module 160 is used to: pass the training operation difference semantic measurement feature vector through a classifier to obtain a classification result, and the classification result is used to indicate whether the training operation meets the predetermined standard.

[0037] Specifically, the training operation standardization detection module 160 is further used to: use the fully connected layer of the classifier to fully connect the training operation difference semantic measurement feature vector to obtain an encoded classification feature vector; and input the encoded classification feature vector into the Softmax classification function of the classifier to obtain the classification result.

[0038] That is, in the technical solution of the present application, the label of the classifier includes the practical training operation meeting the predetermined standard (first label), and the practical training operation not meeting the predetermined standard (second label), wherein the classifier determines to which classification label the practical training operation difference semantic measurement feature vector belongs through the soft maximum function. It is worth noting that the first label p1 and the second label p2 here do not contain artificially set concepts. In fact, during the training process, the computer model does not have the concept of "whether the practical training operation meets the predetermined standard". It only has two classification labels and the probability of the output feature under these two classification labels, that is, the sum of p1 and p2 is one. Therefore, the classification result of whether the practical training operation meets the predetermined standard is actually converted into a binary classification class probability distribution that conforms to the laws of nature through the classification label. In essence, the physical meaning of the natural probability distribution of the label is used, rather than the linguistic text meaning of "whether the practical training operation meets the predetermined standard".

[0039] It should be understood that the role of the classifier is to use the given categories and known training data to learn classification rules and classifiers, and then classify (or predict) unknown data. Logistic regression and SVM are often used to solve binary classification problems. For multi-class classification problems, logistic regression or SVM can also be used, but multiple binary classifications are required to form a multi-classification. However, this is prone to errors and is not efficient. Commonly used multi-classification methods include the Softmax classification function.

[0040] Furthermore, in the technical solution of the present application, the adaptive teaching quality assessment system based on machine learning also includes a training module for training the practical operation semantic feature extractor based on the three-dimensional convolutional neural network model and the classifier.

[0041] Accordingly, in one example, the training module includes: a training data acquisition unit, used to acquire training data, wherein the training data includes a training operation monitoring video acquired by a camera, a training operation reference video extracted from a database, and a true value of whether the training operation meets a predetermined standard; a training video segmentation unit, used to segment the training operation monitoring video and the training operation reference video into video segments to obtain a sequence of training operation monitoring video segments and a sequence of training operation reference video segments; a training video segment semantic feature extraction unit, used to perform feature extraction on the sequence of training operation monitoring video segments and the sequence of training operation reference video segments through the training operation semantic feature extractor based on the three-dimensional convolutional neural network model to obtain a sequence of training operation local time domain semantic coding feature maps and a training operation reference local time domain semantic coding feature map. A sequence of training operation semantic difference measurement units, used to respectively calculate the training operation semantic difference measurement coefficients of each group of corresponding training operation local time domain semantic coding feature maps and training operation reference local time domain semantic coding feature maps in the sequence of training operation local time domain semantic coding feature maps and the sequence of training operation reference local time domain semantic coding feature maps to obtain a training operation difference semantic measurement feature vector composed of multiple training operation semantic difference measurement coefficients; a loss function value calculation unit, used to pass the training operation difference semantic measurement feature vector through the classifier to obtain a classification loss function value; and a loss training unit, used to train the training operation semantic feature extractor based on the three-dimensional convolutional neural network model and the classifier with the classification loss function value, wherein in each round of iteration of the training, the training operation difference semantic measurement feature vector is converged and optimized.

[0042] The loss function value calculation unit is used to: the classifier processes the training operation difference semantic measurement feature vector according to the following classification training formula to obtain the training classification result, and the classification training formula is: softmax{(W n ,B n ):…:(W1,B1)|X}, where W1 to W n is the weight matrix, B1 to B n is a bias vector, X is a semantic measurement feature vector of the difference between the training and actual operations; and a cross entropy value between the training classification result and the true value is calculated as the classification loss function value.

[0043] In particular, in the technical solution of the present application, the sequence of the training operation local time domain semantic coding feature graphs and the sequence of the training operation reference local time domain semantic coding feature graphs respectively represent the time series distribution of the semantic features of the video segments of the training operation monitoring video segments based on three-dimensional convolution coding and the time series distribution of the semantic features of the video segments of the training operation reference video segments based on three-dimensional convolution coding. Due to the differences in the video shooting techniques in the video source domain and the morphological differences of the characters in the video of the training operation monitoring video segments and the training operation reference video segments, the calculation accuracy of the training operation semantic difference measurement coefficients of each group of corresponding training operation local time domain semantic coding feature graphs and training operation reference local time domain semantic coding feature graphs in the sequence of the training operation local time domain semantic coding feature graphs and the sequence of the training operation reference local time domain semantic coding feature graphs will have baseline shift and divergence, so that the training operation difference semantic measurement feature vector composed of multiple training operation semantic difference measurement coefficients has information game discretization, thereby affecting the classification effect of the training operation difference semantic measurement feature vector.

[0044] Based on this, the applicant of the present application preferably performs convergence optimization on the training practice operation difference semantic measurement feature vector each time the training practice operation difference semantic measurement feature vector is classified and iterated through the classifier.

[0045] Accordingly, in one example, in each iteration of the training, the training-practice operation difference semantic measurement feature vector is convergently optimized using the following optimization formulation to obtain an optimized training-practice operation difference semantic measurement feature vector; wherein the optimization formulation is:

[0046]

[0047] Among them, v iis the i-th eigenvalue of the training operation difference semantic metric feature vector, and L is a scale hyperparameter, log|·| represents a logarithmic function with base 2, and v' i It is the i-th eigenvalue of the semantic metric feature vector of the difference in training operation after optimization.

[0048] Specifically, when the training and practical operation difference semantic measurement feature vector is classified and iterated by the classifier, and the weight matrix of the classifier acts on the training and practical operation difference semantic measurement feature vector, due to the compactness of the weight matrix itself, the position information game between the eigenvalues ​​of each position of the training and practical operation difference semantic measurement feature vector will be discretized to produce a large-scale information game, resulting in the classification solution being unable to converge to the Nash equilibrium on the basis of the game. In this way, the equivalent convergence of the information game equalization of the training and practical operation difference semantic measurement feature vector based on the vector information self-controlled equilibrium neighborhood of the training and practical operation difference semantic measurement feature vector can be promoted by the self-game of the eigenvalue in the local neighborhood, thereby improving the classification effect of the training and practical operation difference semantic measurement feature vector. In this way, a machine learning algorithm can be used to automatically evaluate the performance of students in practical operations to improve the accuracy of teaching quality evaluation, so as to facilitate the provision of personalized feedback and guidance based on the evaluation results. In this way, teachers can better understand students' practical operation capabilities, optimize the teaching process, improve teaching quality and efficiency, and promote the cultivation of students' practical and innovative capabilities.

[0049] In summary, the adaptive teaching quality assessment system 100 based on machine learning according to the embodiment of the present application is explained, which can improve the teaching quality and efficiency.

[0050] As described above, the adaptive teaching quality assessment system 100 based on machine learning according to the embodiment of the present application can be implemented in various terminal devices, such as a server having an adaptive teaching quality assessment algorithm based on machine learning according to the embodiment of the present application. In one example, the adaptive teaching quality assessment system 100 based on machine learning according to the embodiment of the present application can be integrated into the terminal device as a software module and / or a hardware module. For example, the adaptive teaching quality assessment system 100 based on machine learning according to the embodiment of the present application can be a software module in the operating system of the terminal device, or can be an application developed for the terminal device; of course, the adaptive teaching quality assessment system 100 based on machine learning according to the embodiment of the present application can also be one of the many hardware modules of the terminal device.

[0051] Alternatively, in another example, the adaptive teaching quality assessment system 100 based on machine learning according to the embodiment of the present application and the terminal device may also be separate devices, and the adaptive teaching quality assessment system 100 based on machine learning may be connected to the terminal device via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0052] Figure 2 Flow chart of an adaptive teaching quality assessment method based on machine learning according to an embodiment of the present application. Figure 3 FIG. 1 is a schematic diagram of the system architecture of the adaptive teaching quality evaluation method based on machine learning according to an embodiment of the present application. Figure 2 and Figure 3 As shown, according to the adaptive teaching quality assessment method based on machine learning according to the embodiment of the present application, it includes: S110, obtaining a practical training operation monitoring video collected by a camera; S120, extracting a practical training operation reference video from a database; S130, respectively performing video segmentation on the practical training operation monitoring video and the practical training operation reference video to obtain a sequence of practical training operation monitoring video segments and a sequence of practical training operation reference video segments; S140, respectively performing video segmentation on the sequence of practical training operation monitoring video segments and the sequence of practical training operation reference video segments through a practical training operation semantic feature extractor based on a deep neural network model; Perform feature extraction on the sequence to obtain a sequence of local time-domain semantic coding feature maps of training operations and a sequence of local time-domain semantic coding feature maps of training operation references; S150, perform semantic difference measurement on each group of corresponding local time-domain semantic coding feature maps of training operations and local time-domain semantic coding feature maps of training operation references in the sequence of local time-domain semantic coding feature maps of training operations and the sequence of local time-domain semantic coding feature maps of training operations to obtain difference semantic measurement features of training operations; and, S160, determine whether the training operation meets a predetermined standard based on the difference semantic measurement features of the training operation.

[0053] In a specific example, in the above-mentioned adaptive teaching quality assessment method based on machine learning, the deep neural network model is a three-dimensional convolutional neural network model.

[0054] Here, those skilled in the art can understand that the specific operations of each step in the above-mentioned adaptive teaching quality evaluation method based on machine learning have been referred to above. Figure 1 The description of the adaptive teaching quality evaluation system 100 based on machine learning has been introduced in detail, and therefore, its repeated description will be omitted.

[0055] Figure 4 FIG. 1 is an application scenario diagram of an adaptive teaching quality evaluation system based on machine learning according to an embodiment of the present application. Figure 4As shown, in this application scenario, first, a training operation monitoring video collected by a camera is obtained (for example, Figure 4 D1 as shown in the figure) and extracting practical operation reference videos from the database (for example, Figure 4 Then, the training operation monitoring video and the training operation reference video are input to a server (for example, Figure 4 In S) shown in , the server is able to use the adaptive teaching quality assessment algorithm based on machine learning to process the practical training operation monitoring video and the practical training operation reference video to obtain a classification result indicating whether the practical training operation meets the predetermined standards.

[0056] According to another aspect of the present application, a non-volatile computer-readable storage medium is provided, on which computer-readable instructions are stored. When the instructions are executed by a computer, the above-mentioned method can be executed.

[0057] The program part of the technology can be considered as a "product" or "manufactured product" in the form of executable code and / or related data, which is participated in or realized by computer-readable media. Tangible and permanent storage media can include any memory or storage used by any computer, processor, or similar device or related module. For example, various semiconductor memories, tape drives, disk drives or any similar devices that can provide storage functions for software.

[0058] The present application uses specific words to describe the embodiments of the present application. For example, "first / second embodiment", "one embodiment", and / or "some embodiments" refer to a certain feature, structure or characteristic related to at least one embodiment of the present application. Therefore, it should be emphasized and noted that "one embodiment" or "an embodiment" or "an alternative embodiment" mentioned twice or multiple times in different positions in this specification does not necessarily refer to the same embodiment. In addition, some features, structures or characteristics in one or more embodiments of the present application can be appropriately combined.

[0059] In addition, it will be appreciated by those skilled in the art that various aspects of the present application may be illustrated and described by a number of patentable categories or situations, including any new and useful process, machine, product or combination of substances, or any new and useful improvements thereto. Accordingly, various aspects of the present application may be performed entirely by hardware, entirely by software (including firmware, resident software, microcode, etc.), or by a combination of hardware and software. The above hardware or software may all be referred to as "data blocks", "modules", "engines", "units", "components" or "systems". In addition, various aspects of the present application may be represented as a computer product located in one or more computer-readable media, which includes computer-readable program code.

[0060] Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by those of ordinary skill in the art to which this application belongs. It should also be understood that terms such as those defined in common dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant technology, and should not be interpreted in an idealized or extremely formal sense, unless explicitly defined as such herein.

[0061] The above is an explanation of the present application and should not be considered as a limitation thereof. Although several exemplary embodiments of the present application are described, it will be readily understood by those skilled in the art that many modifications may be made to the exemplary embodiments without departing from the novel teachings and advantages of the present application. Therefore, all of these modifications are intended to be included within the scope of the present application as defined by the claims. It should be understood that the above is an explanation of the present application and should not be considered to be limited to the specific embodiments disclosed, and modifications to the disclosed embodiments and other embodiments are intended to be included within the scope of the appended claims. The present application is defined by the claims and their equivalents.

Claims

1. An adaptive teaching quality evaluation system based on machine learning, characterized in that: include: A training operation monitoring video acquisition module is used to obtain the training operation monitoring video collected by the camera; A practical training operation reference video acquisition module is used to extract practical training operation reference videos from a database; A video segmentation module, used to segment the practical operation monitoring video and the practical operation reference video into video segments to obtain a sequence of practical operation monitoring video segments and a sequence of practical operation reference video segments; A video clip semantic feature extraction module, used for performing feature extraction on the sequence of the training operation monitoring video clips and the sequence of the training operation reference video clips respectively through a training operation semantic feature extractor based on a deep neural network model to obtain a sequence of training operation local time domain semantic coding feature maps and a sequence of training operation reference local time domain semantic coding feature maps; A training operation semantic difference measurement module is used to measure the semantic difference of the training operation on each group of corresponding local time domain semantic coding feature maps of the training operation and the sequence of the local time domain semantic coding feature maps of the training operation reference, so as to obtain the semantic measurement feature of the difference of the training operation; as well as A training operation standardization detection module is used to determine whether the training operation meets the predetermined standard based on the difference semantic measurement characteristics of the training operation; The deep neural network model is a three-dimensional convolutional neural network model; The training operation semantic difference measurement module is used to: Respectively calculating the training operation semantic difference measurement coefficients of each group of corresponding training operation local time domain semantic coding feature maps and training operation reference local time domain semantic coding feature maps in the sequence of the training operation local time domain semantic coding feature maps and the sequence of the training operation reference local time domain semantic coding feature maps to obtain a training operation difference semantic measurement feature vector composed of multiple training operation semantic difference measurement coefficients as the training operation difference semantic measurement feature; The training operation semantic difference measurement module is used to: The training operation semantic difference measurement coefficients of each group of corresponding training operation local time domain semantic coding feature maps and training operation reference local time domain semantic coding feature maps in the sequence of the training operation local time domain semantic coding feature maps and the sequence of the training operation reference local time domain semantic coding feature maps are calculated respectively by the following coefficient calculation formula to obtain the training operation difference semantic measurement feature vector composed of multiple training operation semantic difference measurement coefficients; The coefficient calculation formula is: Among them, p (x,y,z) and q (x,y,z) are the eigenvalues ​​of each position of each group of corresponding training operation local time domain semantic coding feature maps and training operation reference local time domain semantic coding feature maps in the sequence of the training operation local time domain semantic coding feature maps and the sequence of the training operation reference local time domain semantic coding feature maps, respectively; W is the width of each group of corresponding training operation local time domain semantic coding feature maps and training operation reference local time domain semantic coding feature maps; H is the height of each group of corresponding training operation local time domain semantic coding feature maps and training operation reference local time domain semantic coding feature maps; C is the number of channels of each group of corresponding training operation local time domain semantic coding feature maps and training operation reference local time domain semantic coding feature maps; Hp is the eigenvalue of each position in the training operation difference semantic metric feature vector; log represents the logarithmic function with base 2.

2. The adaptive teaching quality evaluation system based on machine learning according to claim 1 is characterized in that: The practical training operation standardization detection module is used to: The difference semantic measurement feature vector of the training operation is passed through a classifier to obtain a classification result, and the classification result is used to indicate whether the training operation meets a predetermined standard.

3. The adaptive teaching quality evaluation system based on machine learning according to claim 2 is characterized in that: It also includes a training module for training a practical operation semantic feature extractor based on a three-dimensional convolutional neural network model and the classifier.

4. The adaptive teaching quality evaluation system based on machine learning according to claim 3 is characterized in that: The training module comprises: A training data acquisition unit, used to acquire training data, wherein the training data includes a monitoring video of a training operation acquired by a camera, a reference video of a training operation extracted from a database, and a true value of whether the training operation meets a predetermined standard; A training video segmentation unit, used to segment the training operation monitoring video and the training operation reference video into video segments to obtain a sequence of training operation monitoring video segments and a sequence of training operation reference video segments; A training video clip semantic feature extraction unit, used for performing feature extraction on the sequence of the training operation monitoring video clips and the sequence of the training operation reference video clips respectively through the training operation semantic feature extractor based on the three-dimensional convolutional neural network model to obtain a sequence of training operation local time domain semantic coding feature maps and a sequence of training operation reference local time domain semantic coding feature maps; A training operation semantic difference measurement unit is used to respectively calculate the training operation semantic difference measurement coefficients of each group of corresponding training operation local time domain semantic coding feature maps and training operation reference local time domain semantic coding feature maps in the sequence of the training operation local time domain semantic coding feature maps and the sequence of the training operation reference local time domain semantic coding feature maps to obtain a training operation difference semantic measurement feature vector composed of a plurality of training operation semantic difference measurement coefficients; A loss function value calculation unit, used for passing the training and actual operation difference semantic measurement feature vector through the classifier to obtain a classification loss function value; and A loss training unit is used to train the training operation semantic feature extractor and the classifier based on the three-dimensional convolutional neural network model with the classification loss function value, wherein in each round of iteration of the training, the training operation difference semantic measurement feature vector is converged and optimized.

5. The adaptive teaching quality evaluation system based on machine learning according to claim 4 is characterized in that: The loss function value calculation unit is used to: The classifier processes the training operation difference semantic measurement feature vector using the following classification training formula to obtain a training classification result. The classification training formula is: softmax{(W n ,B n ):…:(W1,B1)|X}, where W1 to W n is the weight matrix, B1 to B n is a bias vector, X is a semantic metric feature vector of the training operation difference; and A cross entropy value between the training classification result and the true value is calculated as the classification loss function value.

Images