KR102675724B1 - Method for Providing Cognitive Test - Google Patents

Abstract

A cognitive testing method for art therapy is disclosed.
This embodiment collects the data necessary for art therapy, develops a process to learn the collected art therapy data with artificial intelligence, induces machine learning, and determines the test performer's performance based on the results of machine learning. We provide a cognitive testing method for art therapy that allows you to design treatment by interpreting pictures.

Description

Cognitive test method for art therapy {Method for Providing Cognitive Test}

One embodiment of the present invention relates to a cognitive testing method for art therapy.

The content described below simply provides background information related to this embodiment and does not constitute prior art.

Recently, art therapy (a combination of psychology and art) has emerged as a new field in the history of mental health treatment. Art therapy is an unconventional form of therapy used to build self-esteem and awareness, improve cognitive and motor skills, resolve conflict or stress, and encourage patient resilience.

Psychotherapy is carried out between the practitioner and the examiner, and the examiner evaluates the practitioner's psychological state through the act of expressing the inner psychological state through pictures. In Korea's current art therapy, the test taker expresses his or her emotions on a test sheet and designs the treatment based on interpretation through the tester. After the treatment is performed, the treatment method is adjusted in various ways depending on the prognosis.

As for the prior art therapy model, there is a therapy assistance site (Therapist Aid). The therapy assistance site allows examiners to organize an art therapy process for therapists through membership, and build a treatment model according to standards.

Children, adolescents, and the general public can draw a variety of facial expressions when expressing their own facial expressions. Through drawings, the results of art therapy can be derived much more clearly than drawings of simple objects.

Therefore, there is a need to develop a process for art therapy.
The above-mentioned art therapy technology is disclosed in Korean Patent Publication No. 10-2015-0118594.

This embodiment collects the data necessary for art therapy, develops a process to learn the collected art therapy data with artificial intelligence, induces machine learning, and determines the test performer's performance based on the results of machine learning. The purpose is to provide a cognitive testing method for art therapy that allows interpreting pictures to design treatment.

According to one aspect of this embodiment, the process of receiving an ID and password from a connected terminal; When the ID and password entered from the connected terminal are authenticated, logging in to the corresponding account; A process of receiving member age information from the terminal where the login was performed; A process of receiving pictures, emoticons, and icons representing the emotional state of the test subject from the terminal where the login was performed; A process of deriving a result value corresponding to a picture, an emoticon, an icon expressing the emotional state of the test subject, and the age information based on learning results learned based on AI; It provides a cognitive test method for art therapy, comprising: generating treatment plan information according to the results and then transmitting the treatment plan information to a treatment center.

As described above, according to this embodiment, data necessary for art therapy is collected, a process for learning the collected art therapy data with artificial intelligence is developed, machine learning is induced, and machine learning is performed. It has the effect of being able to design treatment by interpreting the test performer's drawings based on the results.

1 is a diagram schematically showing an art therapy device according to this embodiment.
Figure 2 is a flowchart for explaining an art therapy method using machine learning and image classification according to this embodiment.
Figure 3 is a diagram for explaining the supervised learning method according to this embodiment.
Figure 4 is a diagram for explaining ML model training according to this embodiment.
Figure 5 is a diagram for explaining a cloud service according to this embodiment.
Figure 6 is a diagram for explaining the ML cloud design according to this embodiment.
Figure 7 is a diagram to explain building an intelligent application using AI services according to this embodiment.
Figure 8 is a diagram for explaining a custom ML model construction business using a managed ML service according to this embodiment.
Figure 9 is a diagram for explaining a managed ETL service for data processing according to this embodiment.
Figure 10 is a diagram for explaining machine learning on multiple platforms according to this embodiment.
Figure 11 is a diagram for explaining the model distribution approach according to this embodiment.

Hereinafter, this embodiment will be described in detail with reference to the attached drawings.

1 is a diagram schematically showing an art therapy device according to this embodiment.

The art therapy device 100 receives a picture from the examinee and generates a supervised learning model. The art therapy device 100 learns the picture input from the examinee as a standard for recognizing objects using artificial intelligence, and trains it inside the system using supervised machine learning.

The art therapy device 100 can be set as a test reference value when a test subject (performer) performs a cognitive psychological test in the future based on the learning results. The art therapy device 100 can obtain information about the result of a picture input by the examinee by transmitting information to the center after the tester (or therapist) or by requesting an explanation from the examinee (practitioner) on its own.

The art therapy device 100 according to this embodiment includes a login unit 110, an information input unit 120, a learning unit 130, a result derivation unit 140, and a treatment plan transmission unit 150. Components included in the art therapy device 100 are not necessarily limited thereto.

The login unit 110 receives an ID and password from the connected terminal. When the ID and password entered from the connected terminal are authenticated, the login unit 110 logs in with the corresponding account.

The information input unit 120 receives the member's age information from the terminal where the login has been performed. The information input unit 120 receives pictures, emoticons, and icons representing the emotional state of the test subject from the terminal where the login has been performed.

The learning unit 130 classifies the emotional state corresponding to the input picture, emoticon, icon, and age information representing the emotional state of the test subject, and generates a learning result by learning the emotional state.

The learning unit 130 defines the ML stack and ML workload when generating a learning result by learning the result value corresponding to the picture, emoticon, icon, and age information representing the emotional state of the test subject input based on AI. , establish design principles and architecture, and establish a framework.

The learning unit 130 uses API calls to perform supervised learning on result values corresponding to pictures, emoticons, icons, and age information expressing the emotional state of the test subject in order to perform a cloud-based psychological test on the test subject. .

The learning unit 130 uses a cloud service tool to assign labels to pictures, emoticons, icons, and age information expressing the emotional state of the test subject, and builds an ML model based on the labels.

The learning unit 130 uses an ML model to generate learning results by learning result values corresponding to pictures, emoticons, icons, and age information expressing the emotional state of the test subject.

The learning unit 130 uses an ML model to define input values to output learning results obtained by learning results corresponding to pictures, emoticons, icons, and age information representing the emotional state of the test subject.

The learning unit 130 determines classification problems, regression problems, and clustering problems when defining input values, and checks data availability and collects a data lake when collecting data.

When learning result values corresponding to pictures, emoticons, icons, and age information expressing the emotional state of the test subject, the learning unit 130 includes at least one of the learning rate, number of epochs, hidden layer, hidden unit, and activation function. Prevent overfitting and underfitting by adjusting hyperparameters.

If the learning result of learning the result using an ML model is less than a preset standard threshold, the learning unit 130 finely adjusts the result value corresponding to the picture, emoticon, icon, and age information expressing the emotional state of the test subject. Afterwards, perform the study again.

If the learning result of learning the result using the ML model is greater than or equal to a preset reference threshold, the learning unit 130 reflects the learning result in the ML model.

The result deriving unit 140 derives result values corresponding to pictures, emoticons, icons, and age information expressing the emotional state of the test subject based on learning results learned based on AI received from the learning unit 130.

The result deriving unit 140 uses K-Means Clustering technology to classify the emotional state corresponding to the input pictures, emoticons, icons, and age information representing the emotional state of the test subject. The result deriving unit 140 derives the emotional state as a result value.

The treatment plan transmission unit 150 generates treatment plan information according to the result and then transmits the treatment plan information to the treatment center.

Figure 2 is a flowchart for explaining an art therapy method using machine learning and image classification according to this embodiment.

The art therapy device 100 receives an ID and password from the terminal connected to log in to the service (S210). In step S210, when the ID and password input from the connected terminal are authenticated, the art therapy device 100 logs in with the corresponding account.

The art therapy device 100 receives the member's age information from the terminal through which login has been performed (S220). The art therapy device 100 receives pictures, emoticons, and icons that express the emotional state of the test subject from the logged-in terminal (S230).

The art therapy device 100 derives result values corresponding to pictures, emoticons, and icons expressing the emotional state of the test subject based on learning content learned based on AI (S240). The art therapy device 100 designs a treatment plan according to the results corresponding to each emotion and transmits the corresponding information to the treatment center (S250).

In Figure 2, steps S210 to S250 are described as being sequentially executed, but this is not necessarily limited. In other words, the steps shown in FIG. 2 may be applied by modifying them or executing one or more steps in parallel, so FIG. 2 is not limited to a time-series order.

As described above, the art therapy method using machine learning and image classification according to the present embodiment shown in FIG. 2 may be implemented as a program and recorded on a computer-readable recording medium. The computer-readable recording medium on which the program for implementing the art therapy method using machine learning and image classification according to this embodiment is recorded includes all types of recording devices that store data that can be read by a computer system. .

Figure 3 is a diagram for explaining the supervised learning method according to this embodiment.

The art therapy device 100 defines the ML stack and ML workload, sets design principles and reference architecture, and establishes a Well-Architected framework from an ML perspective to enhance the effectiveness of machine learning (ML). Set .

The art therapy device 100 strengthens the psychological test learning model using machine learning (ML) to define the ML stack and ML workload. The art therapy device 100 builds a system to input the emotional data (emotion picture) of the test subject, and uses K-Clustering technology to derive a result indicating what emotional classification state it corresponds to.

The art therapy device 100 performs supervised learning using an AI service. Supervised learning is performed based on the participant’s active data. The art therapy device 100 performs AI-based learning using services basically built within the cloud service. The art therapy device 100 performs supervised learning to perform a psychological test on a test subject on a cloud-based basis using a simple API call.

The art therapy device 100 performs supervised learning using an ML service. Art therapy device 100 uses cloud service tools to label data and build custom ML models. Since the art therapy device 100 performs ML services on a cloud basis, there is no need to build a direct infrastructure for learning.

The art therapy device 100 builds a model using open source frameworks such as Open API and Open CV. The art therapy device 100 is a cloud service that can build its own model using containers and computing infrastructure preloaded with major open source frameworks such as TensorFlow, PyTorch, MXNet, etc.

The art therapy device 100 performs a psychological test using a cloud-based infrastructure. The art therapy device 100 identifies business goals and advises whether key performance indicators (KPIs) should be monitored based on psychological test results. The art therapy device 100 defines an ML model and inputs to provide a desired output.

When defining an input, the art therapy device 100 determines whether it is a classification problem, a regression problem, or a clustering problem. When collecting data, the art therapy device 100 checks data availability and collects a data lake. The art therapy device 100 can establish a treatment plan using various learning models and create a model to increase prognosis.

Figure 4 is a diagram for explaining ML model training according to this embodiment.

The art therapy device 100 uses a cloud service to label and prepare data. Art therapy device 100 uses a fully managed ETL (extract, transform, and load) cloud service to label raw data and create high-quality training datasets.

When learning data is prepared using a cloud service, the art therapy device 100 understands and identifies data patterns in the visualization and analysis stage. The art therapy device 100 hosts a terminal using a cloud service and provides a collection of visualization tools using libraries (seaborn, matplotlib, plotly).

The art therapy device 100 provides a fully managed service that can query a data lake using a cloud service. The art therapy device 100 analyzes pictures, emoticons, and icons expressing the emotional state of the test subject in real time using a fully managed service. The art therapy device 100 provides a BI (Business Intelligence) tool that provides secure sharing and collaboration for storyboards. Art therapy device 100 considers all attributes of the data set and determines their relevance to the model and whether transformation or extraction is needed.

Figure 5 is a diagram for explaining a cloud service according to this embodiment.

When the art therapy device 100 secures a data set for performing learning in connection with a cloud service, it performs a model training step using the data set. The art therapy device 100 selects an ML algorithm appropriate for the business problem to perform model training.

The art therapy device 100 adjusts the hyperparameters of the model to achieve low training errors and prevent overfitting and underfitting. For example, the art therapy apparatus 100 adjusts hyperparameters including at least one of a learning rate, number of epochs, hidden layer, hidden unit, and activation function.

The art therapy device 100 operates based on a cloud service that can call an API to educate data stored in a storage server. The art therapy device 100 can use a cloud service to warn about model training problems, such as a gradient value becoming too large or too small. The art therapy device 100 uses cloud services Deep Learning AMI or Deep Learning Containers depending on the level of expertise and runs open source frameworks such as TensorFlow, PyTorch, MXNet, Horovod, and Keras.

Art therapy device 100 performs model and business evaluations that are used to determine whether the performance and accuracy of the model can achieve business goals.

Art therapy device 100 evaluates the model using historical data or live data. The art therapy device 100 may use the cloud service Deep Learning AMI or the cloud service EMR for model and business evaluation.

If the art therapy device 100 is not satisfied with the model and business evaluation results, it can fine-tune and retrain the model. If the art therapy device 100 is satisfied with the model and business evaluation results, it is distributed to production.

The art therapy device 100 uses the generated model to infer result values corresponding to pictures, emoticons, and icons expressing the emotional state of the test subject.

The art therapy device 100 uses a cloud service that can call an API for inference as an option when distributing a model. The art therapy device 100 may use a user-specifiable cloud inference pipeline.

Art therapy device 100 continuously monitors ML models in production for input data drift and model quality deviations. When using an edge or Internet of Things (IoT) device, the art therapy device 100 performs compilation on the edge device using a cloud service. The art therapy device 100 can use Elastic Inference, a cloud service that can connect any EC2 or GPU inference acceleration in use.

Figure 6 is a diagram for explaining the ML cloud design according to this embodiment.

The art therapy device 100 uses a cloud service for ML processes.

Art therapy device 100 identifies a set of design principles that facilitate ML cloud design. The art therapy device 100 implements agility through the availability of high-quality data. The art therapy device 100 performs model training and evaluation separation from model hosting. The art therapy device 100 detects data drift. Art therapy device 100 automates the training and assessment pipeline. The art therapy device 100 moves to a general scenario and implements a cloud service architecture.

Figure 7 is a diagram to explain building an intelligent application using AI services according to this embodiment.

The art therapy device 100 builds an intelligent application using a cloud service AI service. The art therapy device 100 implements ML services using a cloud service AI service layer. The art therapy device 100 uses a cloud service to analyze pictures, emoticons, and icons that express the emotional state of the test subject.

The art therapy device 100 transmits pictures, emoticons, and icons expressing the emotional state of the test subject to a cloud service in order to analyze image attribute data. The art therapy device 100 transmits data to the cloud service QuickSight for visualization of the data analyzed using the cloud service. When using a cloud service, the art therapy device 100 applies a fully managed cloud AI service that includes a pre-trained model.

The art therapy device 100 can apply the fully managed cloud AI service to text and audio analysis as well as pictures, emoticons, and icons expressing the emotional state of the subject. The art therapy device 100 may use Transcribe (audio analysis) or a cloud service Comprehend (text analysis) to analyze text and audio.

The art therapy device 100 uploads media files and performs video and audio analysis using several cloud service AI services. Art therapy device 100 uses a searchable metadata library. The art therapy device 100 adjusts the media analysis process using the Step Function of the cloud service and searches content using the cloud service ElasticSearch.

Figure 8 is a diagram for explaining a custom ML model construction business using a managed ML service according to this embodiment.

Art therapy device 100 uses a managed ML service to build a custom ML model. The art therapy device 100 builds a unique model for cloud AI services. The art therapy device 100 controls data preparation, data analysis, model training, model evaluation, and model inference steps.

The art therapy device 100 performs all functions end-to-end while being fully managed by a cloud service. The art therapy device 100 uses Lambda, a cloud service, to support an event-based architecture and connect multiple steps of the ML process together.

The art therapy device 100 uses a storage server to function as a data lake for raw, modeled, enhanced, and converted data. The art therapy device 100 uses cloud services to build a unique model and flow from collection to inference.

Figure 9 is a diagram for explaining a managed ETL service for data processing according to this embodiment.

The art therapy device 100 provides a managed ETL service for data processing. Art therapy device 100 sends messages to consumers using email and SMS. The art therapy device 100 uses an ML model to identify the correct customer based on past consumer purchasing patterns. The art therapy device 100 uses a managed service that can send target messages from various channels.

Figure 10 is a diagram for explaining machine learning of multiple platforms according to this embodiment.

The art therapy device 100 performs multi-platform machine learning. The art therapy device 100 supports machine learning with an edge device using a cloud service. The art therapy device 100 uses a cloud service to perform ML inference locally on the device using trained and optimized models created in the cloud.

The art therapy device 100 supports various hardware platforms such as Intel and Nvidia using a cloud service with pre-trained ML models. The art therapy device 100 uses a compiler and runtime to support IoT devices.

Figure 11 is a diagram for explaining the model distribution approach according to this embodiment.

Art therapy device 100 defines a model deployment approach. After building a business problem model, the art therapy device 100 builds a method to infer the model and obtain a prediction. The art therapy device 100 makes an API call to the model using the endpoint of the cloud service.

The above description is merely an illustrative explanation of the technical idea of the present embodiment, and those skilled in the art will be able to make various modifications and variations without departing from the essential characteristics of the present embodiment. Accordingly, the present embodiments are not intended to limit the technical idea of the present embodiment, but rather to explain it, and the scope of the technical idea of the present embodiment is not limited by these examples. The scope of protection of this embodiment should be interpreted in accordance with the claims below, and all technical ideas within the equivalent scope should be interpreted as being included in the scope of rights of this embodiment.

100: Art therapy device
110: Login unit 120: Information input unit
130: Learning unit 140: Result derivation unit
150: Treatment plan transmission unit

Claims (8)

A process of receiving an ID and password from a terminal connected to the information input unit;
When the ID and password input from the connected terminal are authenticated in the login unit, logging in with the corresponding account;
A process of receiving member age information from the terminal where the login was performed at the information input unit;
A process of receiving, at the information input unit, a picture, emoticon, or icon representing the emotional state of the test subject from the terminal where the login was performed;
A test reference value for conducting a cognitive psychological test is set based on the learning results learned using the AI of the learning unit, and a picture or emoticon representing the emotional state of the test subject input from the information input unit based on the set test reference value, Icon, a process of deriving an emotional state corresponding to the age information as a result of a cognitive psychological test in a result derivation unit;
A process of inputting the results from a treatment plan transmission unit into a previously learned learning model and transmitting the output information to a treatment center;
A cognitive testing method for art therapy, comprising: According to paragraph 1,
The learning unit defines an ML stack and ML workload in response to the picture, the emoticon, the icon, and the age information representing the emotional state of the test subject input from the information input unit, and design principles and ML workloads according to the definition. A cognitive test method for art therapy, characterized by setting an architecture and learning the results of a cognitive psychological test based on AI that sets a framework based on the design principles and the architecture. According to paragraph 2,
In order to perform a cloud-based psychological test on a test subject using an API call, supervised learning is performed on the results corresponding to the picture expressing the emotional state of the test subject, the emoticon, the icon, and the age information. Cognitive testing method for art therapy, characterized by: According to paragraph 1,
Using K-Means Clustering technology, the emotional state corresponding to the picture, the emoticon, the icon, and the age information representing the emotional state of the test subject are classified, and the emotional state is classified as the result. A cognitive test method for art therapy characterized by deriving values. According to paragraph 1,
Using a cloud service tool, assign labels to the picture, emoticon, icon, and age information expressing the emotional state of the test subject, and build an emotional state ML model based on each label. do,
Using the emotional state ML model, perform a psychological test corresponding to the picture, the emoticon, the icon, and the age information expressing the emotional state of the test subject, and generate the learning result by learning the results of the psychological test. Cognitive testing method for art therapy, characterized by: According to clause 5,
It is defined as the input value collected by the data lake while data availability is confirmed when information is determined through analysis of the information input from the information input unit according to the classification problem, regression problem, and clustering problem, and data is collected from the determined information. do,
Characterized in outputting the learning result by learning the result using the emotional state ML model with input values defined by the picture, the emoticon, the icon, and the age information expressing the emotional state of the test subject. Cognitive testing methods for art therapy. According to clause 5,
When learning the result value corresponding to the picture, the emoticon, the icon, and the age information expressing the emotional state of the test subject,
A cognitive test method for art therapy, characterized by preventing overfitting and underfitting by adjusting hyperparameters including at least one of learning rate, number of epochs, hidden layer, hidden unit, and activation function. According to clause 5,
If the learning result obtained by learning the result using the ML model is less than a preset reference threshold, the result value corresponding to the picture, the emoticon, the icon, and the age information expressing the emotional state of the test subject is obtained. After adjusting, perform learning again,
A cognitive test method for art therapy, wherein when the learning result of learning the result using the ML model is greater than a preset reference threshold, the learning result is reflected in the ML model.