Artificial intelligence fan and wind supply method of artificial intelligence fan

Abstract

A device for supplying wind for an artificial intelligence fan according to an embodiment comprises: a memory storing at least one command; and a processor performing an operation according to the command, wherein the processor implements a fan control model using learning data for object recognition, wind direction, and wind speed control; the fan control model collects sensor data from a sensor installed in the fan, calculates wind speed and wind direction according to the position and number of human objects recognized from the sensor data, the type of human objects, temperature, ambient temperature, and humidity, and instructs the fan to supply wind according to the calculated wind speed and wind direction, wherein the sensor may include a camera, a temperature sensor, and a humidity sensor, and the sensor data may include a user monitoring image, ambient temperature, user temperature, and ambient humidity.

Description

Artificial intelligence fan and wind supply method of artificial intelligence fan

[0001] The present disclosure relates to an artificial intelligence fan and a method for supplying wind to an artificial intelligence fan, and more specifically, to a smart fan that operates by recognizing the location and temperature of a user object and an area where the fan wind should not reach through an artificial intelligence model, and a method for controlling the smart fan.

[0002] Unless otherwise indicated herein, the materials described in this section are not prior art to the claims of this application, and their inclusion in this section is not intended to be admitted as prior art.

[0003] A fan generates wind by rotating blades attached to a motor's shaft. Traditionally, fans have been controlled by directly controlling the wind speed or direction with buttons.

[0004] However, conventional fan control recognizes not only the human body but also heat-emitting objects, such as gas stoves, as people, and operates accordingly. This causes the fan to direct air to undesirable locations due to object misrecognition. The aforementioned problem is particularly prominent in kitchens during the summer. If a fan is turned on while using a gas stove in the kitchen, the fan may recognize the heat emitted by the stove as a person and direct the air in that direction. As a result, the air is not properly delivered to the person who actually needs cooling, and instead, the air is directed toward the gas stove, hindering effective cooling.

[0005] This problem arises because the sensor technology in electric fans cannot distinguish between heat-emitting objects and human bodies. Conventional fans often mistake heat-emitting objects for people, hindering efficient cooling. Therefore, technological improvements are needed to address this issue.

[0006] An artificial intelligence fan and a method for supplying wind by an artificial intelligence fan according to an embodiment recognize a human object through an artificial intelligence model using a camera located at the center of the fan to determine the location and movement of the human object, and the fan rotates along the human object to supply wind.

[0007] In addition, in the embodiment, the surrounding environment, temperature, humidity, and human body temperature are detected through an artificial intelligence model, and the optimal wind volume is calculated and provided based on the detected temperature, humidity, and body temperature.

[0008] Additionally, the artificial intelligence fan according to the embodiment provides a function to rotate around the human object so that the wind is not supplied directly to the human object.

[0009] Additionally, in the embodiment, when multiple human objects are recognized, the number of human objects and the end portions of the human objects are recognized as contours, so that wind is supplied to the entire area where the human objects are recognized. In particular, rotation is performed up to both edges of the area where multiple human objects exist.

[0010] In addition, in the embodiment, the artificial intelligence fan can be controlled through voice recognition, and when multiple human objects exist, if a specific person's body temperature is higher than the normal range, it can detect this as an abnormality and notify the application of the abnormality.

[0011] In addition, in the embodiment, different wind speeds are supplied according to the body temperature of a human object, and if the body temperature of a specific human object remains high even after supplying wind to a person with a high body temperature for a certain period of time, it is determined that there is a health problem, so the wind speed can be reduced again and the abnormality can be notified through the application.

[0012] However, the problems to be solved according to one embodiment are not limited to those mentioned above.

[0013] A memory storing at least one command for supplying wind of an artificial intelligence fan according to an embodiment; and a processor performing an operation according to the command, wherein the processor implements a fan control model using learning data for object recognition, wind direction, and wind speed control, and the fan control model collects sensor data from a sensor installed in the fan, calculates wind speed and wind direction according to the position and number of human objects recognized from the sensor data, the type of human objects, and temperature and ambient temperature and humidity, and controls the fan to supply wind according to the calculated wind speed and wind direction, and the sensor may include a camera, a temperature sensor, and a humidity sensor, and the sensor data may include a user monitoring image, ambient temperature, and user temperature and ambient humidity.

[0014] In addition, the fan control model can detect heat-emitting objects, including gas stoves and heaters, based on data collected from a temperature detection sensor and learning results of learning data, distinguish between the detected heat-emitting objects and people, and control the fan to provide wind only to the people.

[0015] In addition, the fan control model can classify the recognized person as an infant or an adult, and if the person is an adult, classify the person as male or female, and if the person is classified as a person, control the fan according to a pre-stored control process according to gender and age, thereby providing different wind to each recognized person.

[0016] In addition, the fan control model inputs user-specific data, recognizes the user when the fan is operated, and controls the fan based on the user-specific data matched to the recognized user, thereby providing the user with the desired wind volume and wind speed, and the user-specific data may include authentication data including facial image data, fingerprints, and passwords of each family member.

[0017] In addition, the fan control model can recognize objects around the fan, classify the recognized objects as animals or people, and, if an animal object is recognized, perform a pre-stored control process according to the type of animal to provide wind.

[0018] Additionally, the user-specific data includes the user's preferred control process, and the user's preferred control processor may include the user's preferred fan control options, such as front wind, side wind, wind provision time, wind provision interval, and wind speed.

[0019] Additionally, the fan control model can predict the user's movement path through user motion analysis and control the direction of the fan along the predicted path.

[0020] In addition, the fan control model can recognize a space in which the fan is placed, recognize a human object moving in the recognized space, extract a user preference control process of the recognized human object, and control the fan according to the extracted preference control process.

[0021] Additionally, the fan control model can track a human object for a certain period of time, determine the movement path of the human object in the recognized space, and limit the rotation range of the fan to be included in the movement path.

[0022] In addition, the fan control model can recognize a human object, estimate the body temperature of the recognized human object using an infrared camera installed in the fan, and if the estimated body temperature is outside the normal range, recognize it as an abnormal event and transmit the recognized event to the user terminal.

[0023] The AI ​​fan and its wind supply method described above utilize an AI model to recognize a human object through a camera located at the center of the fan, thereby determining the person's location and movement. This allows the fan to rotate and supply wind following the person, ensuring the user always feels cool in the optimal location.

[0024] Additionally, the embodiment uses an artificial intelligence model to detect the surrounding environment, temperature, humidity, and human body temperature. Based on this detected data, the optimal airflow is calculated and provided, creating a more comfortable environment for the user.

[0025] Additionally, the embodiment provides a function that rotates around the person to prevent direct wind from reaching the person, thereby reducing discomfort caused by direct wind and creating a more comfortable environment through indirect wind supply.

[0026] In addition, when multiple human objects are recognized, the number of human objects and the end portions are recognized as outlines to supply wind evenly to all people, and in particular, in a space with multiple people, the wind can be evenly distributed by rotating from end to end.

[0027] In addition, the fan can be controlled through voice recognition, allowing users to use the fan more conveniently. When multiple human objects exist, if a specific person's body temperature is higher than the normal range, it detects this as an abnormality and notifies the user of the abnormality through the application, allowing the user to quickly recognize and respond to their health status.

[0028] In addition, in the embodiment, different wind speeds are supplied according to the body temperature of a human object, and if the body temperature remains high even after supplying wind to a person with a high body temperature for a certain period of time, the wind speed is reduced again to determine that there is a health problem, and the abnormality is notified through an application, thereby maximizing the convenience and comfort of the user and monitoring the health status.

[0029] The effects of the present invention are not limited to the effects described above, and should be understood to include all effects that can be inferred from the detailed description of the present invention or the composition of the invention described in the claims.

[0030] Figure 1 is a drawing showing an artificial intelligence fan system according to an embodiment.

[0031] Figure 2 is a drawing showing a block diagram of a fan according to an embodiment.

[0032] Figure 3 is a diagram showing the configuration of a command set stored in memory according to an embodiment.

[0033] Figure 4 is a drawing showing an image obtained from a thermal imaging camera in an embodiment.

[0034] Figure 5 is a diagram showing an example of learning data for human object recognition according to an embodiment.

[0035] An artificial intelligence fan according to an embodiment includes a memory storing at least one command for supplying wind; and a processor performing an operation according to the command, wherein the processor implements a fan control model using learning data for object recognition, wind direction, and wind speed control, and the fan control model collects sensor data from a sensor installed in the fan, calculates wind speed and wind direction according to the position and number of human objects recognized from the sensor data, the type of human objects, and temperature and ambient temperature and humidity, and controls the fan to supply wind according to the calculated wind speed and wind direction, and the sensor may include a camera, a temperature sensor, and a humidity sensor, and the sensor data may include a user monitoring image, ambient temperature, and user temperature and ambient humidity.

[0036] An artificial intelligence fan according to an embodiment includes a memory storing at least one command for supplying wind; and a processor performing an operation according to the command, wherein the processor implements a fan control model using learning data for object recognition, wind direction, and wind speed control, and the fan control model collects sensor data from a sensor installed in the fan, calculates wind speed and wind direction according to the position and number of human objects recognized from the sensor data, the type of human objects, and temperature and ambient temperature and humidity, and controls the fan to supply wind according to the calculated wind speed and wind direction, and the sensor may include a camera, a temperature sensor, and a humidity sensor, and the sensor data may include a user monitoring image, ambient temperature, and user temperature and ambient humidity.

[0037] Hereinafter, the embodiments disclosed in this specification will be described in detail with reference to the attached drawings. Regardless of the drawing numbers, identical or similar components will be given the same reference numbers, and redundant descriptions thereof will be omitted. The suffixes "module" and "part" used for components in the following description are assigned or used interchangeably only for the convenience of writing the specification, and do not in themselves have distinct meanings or roles. In addition, when describing the embodiments disclosed in this specification, if it is determined that a specific description of a related known technology may obscure the gist of the embodiments disclosed in this specification, a detailed description thereof will be omitted. In addition, the attached drawings are only intended to facilitate easy understanding of the embodiments disclosed in this specification, and the technical ideas disclosed in this specification are not limited by the attached drawings, and should be understood to include all modifications, equivalents, and substitutes included in the spirit and technical scope of the present invention.

[0038] Terms that include ordinal numbers, such as first, second, etc., may be used to describe various components, but the components are not limited by these terms. These terms are used solely to distinguish one component from another.

[0039] When a component is referred to as being "connected" or "connected" to another component, it should be understood that it may be directly connected or connected to that other component, but that there may be other components intervening. Conversely, when a component is referred to as being "directly connected" or "connected" to another component, it should be understood that there are no other components intervening.

[0040] In this application, terms such as “include” or “have” are intended to specify the presence of a feature, number, step, operation, component, part or combination thereof described in the specification, but should be understood not to exclude in advance the possibility of the presence or addition of one or more other features, numbers, steps, operations, components, parts or combinations thereof.

[0041] In this specification, the term "unit" includes a unit realized by hardware, a unit realized by software, and a unit realized using both. Furthermore, a single unit may be realized using two or more pieces of hardware, and two or more units may be realized by a single piece of hardware.

[0042] Some of the operations or functions described herein as being performed by a terminal, apparatus, or device may instead be performed by a server connected to the terminal, apparatus, or device. Similarly, some of the operations or functions described herein as being performed by a server may also be performed by a terminal, apparatus, or device connected to the server.

[0043] Hereinafter, the present invention will be described in detail with reference to the attached drawings.

[0044] Fig. 1 is a drawing illustrating an artificial intelligence fan system according to an embodiment. Referring to Fig. 1, the artificial intelligence fan system according to the embodiment may be configured to include an artificial intelligence fan (100) and a user terminal.

[0045] In an embodiment, the artificial intelligence fan is a fan that is automatically controlled through an artificial intelligence model, and may include a camera (10), a thermal imaging camera (20), a temperature and humidity sensor (30), an ultrasonic sensor, a lidar sensor, and the like. The camera (10) is installed at the center of the fan to generate a monitoring image of the surroundings of the fan, and recognizes human objects and restricted objects that should not be supplied with wind around the fan. In an embodiment, the restricted objects include, but are not limited to, a gas range, an open flame, a cooking appliance that uses heat, and the like. The thermal imaging camera (20) measures the temperature of objects around the fan. In an embodiment, the thermal imaging camera may be replaced with an infrared camera to measure the temperature of surrounding objects. In addition, the ultrasonic sensor and the lidar sensor measure the distance between the fan and human objects.

[0046] In an embodiment, the artificial intelligence fan can adjust the direction and wind speed of the fan according to the number of human objects recognized by the camera (10) and the temperature of the human objects and the ambient temperature through a learned artificial intelligence-based fan control model. In the artificial intelligence fan and smart fan control method according to the embodiment, a camera located at the center of the fan recognizes human objects through an artificial intelligence model, identifies the location and movement of the human objects, and the fan rotates following the human objects to supply wind.

[0047] The user terminal (200) is a smart terminal linked to an artificial intelligence fan (100), and controls the artificial intelligence fan (100) or receives and outputs abnormal events and operation record data from the artificial intelligence fan (100). In the embodiment, the abnormal events include, but are not limited to, fever of a human object, high temperature generation, and abnormal fan operation events.

[0048] In addition, in the embodiment, the surrounding environment, temperature, humidity, and human body temperature are detected through an artificial intelligence model, and the optimal wind volume is calculated and provided based on the detected temperature, humidity, and body temperature.

[0049] Additionally, the artificial intelligence fan according to the embodiment provides a function to rotate around the human object so that the wind is not supplied directly to the human object.

[0050] Additionally, in the embodiment, when multiple human objects are recognized, the number of human objects and the end portions of the human objects are recognized as contours, so that wind is supplied to the entire area where the human objects are recognized. In particular, rotation is performed up to both edges of the area where multiple human objects exist.

[0051] In addition, in the embodiment, the artificial intelligence fan can be controlled through voice recognition, and when multiple human objects exist, if a specific person's body temperature is higher than the normal range, it can detect this as an abnormality and notify the application of the abnormality.

[0052] In addition, in the embodiment, different wind speeds are supplied according to the body temperature of a human object, and if the body temperature of a specific human object remains high even after supplying wind to a person with a high body temperature for a certain period of time, it is determined that there is a health problem, so the wind speed can be reduced again and the abnormality can be notified through the application.

[0053] Fig. 2 is a drawing showing a block diagram of a fan according to an embodiment.

[0054] The configuration of the artificial intelligence fan (100) illustrated in Fig. 2 is merely a simplified example.

[0055] The communication module (110) can be configured regardless of the communication mode, such as wired or wireless, and can be configured with various communication networks, such as a personal area network (PAN) and a wide area network (WAN). In addition, the communication module (110) can operate based on the well-known World Wide Web (WWW), and can also utilize a wireless transmission technology used for short-distance communication, such as infrared (IrDA: Infrared Data Association) or Bluetooth. For example, the communication module (110) can be responsible for transmitting and receiving data required to perform a technique according to an embodiment of the present disclosure.

[0056] The memory (120) may refer to any type of storage medium. For example, the memory (120) may include at least one type of storage medium among a flash memory type, a hard disk type, a multimedia card micro type, a card type memory (e.g., SD or XD memory, etc.), a RAM (Random Access Memory), a SRAM (Static Random Access Memory), a ROM (Read-Only Memory), an EEPROM (Electrically Erasable Programmable Read-Only Memory), a PROM (Programmable Read-Only Memory), a magnetic memory, a magnetic disk, and an optical disk. Such a memory (120) may also constitute a database as illustrated in FIG. 1.

[0057] The memory (120) can store at least one instruction that can be executed by the processor (130). In addition, the memory (120) can store any type of information generated or determined by the processor (130) and any type of information received by the server (200). For example, the memory (120) stores RM data and RM protocols according to the user, as will be described later. In addition, the memory (120) stores various types of modules, instruction sets, and models.

[0058] The processor (130) may perform technical features according to embodiments of the present disclosure, which will be described later, by executing at least one instruction stored in the memory (120). In one embodiment, the processor (130) may be configured with at least one core and may include a processor for data analysis and / or processing, such as a central processing unit (CPU), a general purpose graphics processing unit (GPGPU), or a tensor processing unit (TPU) of a computer device.

[0059] This processor (130) can train a neural network or model designed using machine learning or deep learning methods. To this end, the processor (130) can perform calculations for neural network training, such as processing input data for training, extracting features from the input data, calculating errors, and updating the weights of the neural network using backpropagation. In addition, the processor (130) can also perform inference for a predetermined purpose using a model implemented using an artificial neural network method.

[0060] FIG. 3 is a diagram showing the configuration of a command set stored in memory according to an embodiment.

[0061] Referring to FIG. 3, the instruction set according to the embodiment may be configured to include a collection unit (121), a preprocessing unit (122), a learning unit (123), and a feedback unit (124). The term 'unit' used in this specification should be interpreted to include software, hardware, or a combination thereof, depending on the context in which the term is used. For example, the software may be machine language, firmware, embedded code, and application software. As another example, the hardware may be a circuit, a processor, a computer, an integrated circuit, an integrated circuit core, a sensor, a MEMS (Micro-Electro-Mechanical System), a passive device, or a combination thereof.

[0062] The collection unit (121) collects learning data for learning a fan control model that controls an artificial intelligence fan and a series of data necessary for controlling the fan. In an embodiment, the learning data includes images for recognizing human objects and animal objects, and information on objects to which the fan wind should not be supplied. Specifically, the learning data includes images for recognizing human objects of various ages and genders, images for recognizing restricted objects such as gas stoves, and voice control command information for voice recognition. In addition, the learning data may include pre-generated control data for optimal control of the fan. In an embodiment, the control data is data on an appropriate wind direction and wind strength preset according to the number of human objects, ambient temperature, and the distance between the fan and human objects. For example, the control data includes, but is not limited to, an optimal strength preset according to the number of human objects and ambient temperature, an optimal strength and an optimal supply range according to the number of human objects, etc.

[0063] The preprocessing unit (122) preprocesses the collected training data to remove biased or discriminatory data from the collected artificial intelligence learning data. In an embodiment, the preprocessing unit (122) preprocesses the collected training data set and processes it into a form suitable for artificial intelligence model learning. For example, the preprocessing unit (122) may perform processes such as noise removal, outlier removal, and missing value processing. In addition, the preprocessing unit (122) may normalize data, remove outliers, or adjust the scale of data through data preprocessing to prevent the model from learning unnecessary patterns.

[0064] The learning unit (123) trains a deep learning neural network with collected learning data to implement a fan control model, which is a deep learning model. In an embodiment, the learning unit (130) can train a model using ambient temperature, the number of human objects, the positions of the human objects, and the temperatures of each human object as input data, and using the appropriate wind speed and direction for the input data as output data.

[0065] In addition, in the embodiment, the learning unit (123) can implement the fan control model by training the model through reinforcement learning. Reinforcement learning (RL) is a field of machine learning in which an agent learns an optimal action (Policy) while interacting with an environment (Environment). Reinforcement learning can be composed of elements such as an agent, which is a subject that makes decisions and performs actions, an environment, which is an external system with which the agent interacts, a state, an action, a reward, and a policy. The environment returns the result of the agent's action. In the embodiment, the environment can include the ambient temperature, the location of a person object, and the location of a restriction object.

[0066] In reinforcement learning, a state is information representing the current state of the environment. In an embodiment, the state may include the temperature, location, distance from the fan, number of human objects, location of constraint objects, etc. of each human object. In an embodiment, the fan control model, which is an agent, determines the fan's behavior based on the aforementioned state.

[0067] An action is an action an agent can take in the environment. An action changes the state of the environment. In an example, an action may include the strength and direction of the wind supplied to the surroundings by the model's fan control process.

[0068] A reward is the feedback the environment provides to an agent as a result of its actions. Rewards represent how well the agent achieved its goals, and agents learn to maximize rewards. In some embodiments, rewards may include lowering the ambient temperature above an appropriate level, providing even airflow to recognized human objects, lowering the body temperature of each human object to an appropriate range, and blocking airflow to restricted objects such as a gas stove. In reinforcement learning, a policy is a strategy that determines which action an agent will take in a given state. Policies can include both deterministic and probabilistic policies.

[0069] In this example, the fan control model, an agent, learns based on the returned rewards and new states. The goal is to learn a policy that maximizes cumulative rewards in the long term. In this example, the fan control model balances exploring new behaviors with exploiting known good behaviors through an exploration-exploitation tradeoff during the process of learning the optimal policy.

[0070] In the embodiment, the learning unit (123) first defines the fan system to be controlled and the goal for learning the fan control model. For example, the learning unit (123) may set the goal of the fan control model to be maintained within a specific temperature range or to optimize energy efficiency, and the environment may be set to the environment in which the fan operates (e.g., the size, temperature, humidity, etc. of the room). Furthermore, in the embodiment, the learning unit (123) learns through the agent's interaction with the environment in reinforcement learning. Furthermore, the learning unit sets the agent environment, action, state, reward, and policy, which are elements of reinforcement learning. For example, the learning unit (123) may set the environment as the size and temperature of the room in which the fan is placed, and the state as the temperature of the recognized person object, the position and number of the person objects, and the ambient temperature. The action may be set to the wind direction and wind force of the fan to change the state of the environment. The reward may be set to lower the temperature around the fan and the temperature of the person object, and the policy may be set to a restricted object. Thereafter, the learning unit (123) selects one of various reinforcement learning algorithms (Q-learning, DQN (Deep Q-Network), PPO (Proximal Policy Optimization)) and trains it. In addition, the learning unit (123) can train the fan control model, which is an agent, in a simulation environment. For example, if the learning unit (123) cannot use an actual environment, it builds a simulation environment and trains the model.

[0071] In addition, the learning unit (123) can implement a fan control model by learning the model through various learning methods as well as reinforcement learning.

[0072] For example, the learning unit (123) can train a model through rule-based control. Rule-based control is a method of controlling a system by defining explicit rules based on specific conditions. This determines the operation of a fan using logical conditional statements and simple rules.

[0073] In the embodiment, the learning unit (123) can set clear rules for controlling the speed and operating status of the fan according to temperature and humidity, and train the model with the set rules.

[0074] Additionally, the learning unit (123) can train the model through PID control (Proportional-Integral-Derivative Control). PID control is a widely used method for controlling continuous variables such as temperature. The PID controller is a control technique that focuses on reducing errors. In the embodiment, the difference (error) between the current state and the target state is calculated, and a control signal is generated based on this error to control the fan.

[0075] Additionally, the learning unit (123) can train the model through machine learning-based models. In an embodiment, the learning unit (123) can learn fan control based on data using a machine learning model. Mainly, a regression model or a classification model is used. For example, the learning unit (123) collects past temperature and humidity data and corresponding fan control data, and trains the model using the collected data. In an embodiment, a sufficient amount of learning data is used to train complex patterns so that the model can effectively control the fan.

[0076] Additionally, in the embodiment, the learning unit (123) can train the model through fuzzy logic control. Fuzzy logic is useful for processing uncertain or ambiguous data. Fuzzy logic systems are similar to rule-based systems, but provide more flexible control methods that are closer to human reasoning. In the embodiment, fuzzy sets are defined for input variables such as temperature and humidity, and fuzzy rules are used to control the fan.

[0077] Referring again to FIG. 3, the processor (130) learns learning data for user object recognition, including human objects and animal objects, and wind direction and wind speed control using the aforementioned method to implement a fan control model.

[0078] In an embodiment, a fan control model collects sensor data from sensors installed in the fan, calculates the position and number of human objects recognized from the sensor data, the type of human objects, and the wind speed and wind direction based on the temperature and ambient temperature and humidity, and controls the fan to supply wind based on the calculated wind speed and wind direction. In the embodiment, the sensors include a camera, a temperature sensor, and a humidity sensor, and the sensor data includes a user monitoring image, ambient temperature, and user temperature and ambient humidity.

[0079] The fan control model then collects user monitoring images, ambient temperature, user temperature, and ambient humidity from cameras, temperature sensors, and humidity sensors installed on the fan. It then calculates the location and number of human objects recognized in the monitoring images, as well as wind speed and direction based on the temperature of the human objects and the ambient temperature and humidity. The fan is then controlled to supply wind based on the calculated wind speed and direction.

[0080] To achieve this, the fan control model collects and integrates sensor data. In an embodiment, the sensor data may include monitoring images collected from a camera, temperature and humidity data, and images from a thermal imaging camera.

[0081] The fan control model then uses computer vision technology to recognize human objects in the monitoring images. This model can utilize deep learning-based object detection algorithms (e.g., YOLO, SSD, Faster R-CNN). Furthermore, it identifies the locations of recognized human objects and tracks the locations and number of people in the room. Furthermore, it compares the ambient temperature with the user's temperature to determine measures to maintain a comfortable temperature. Furthermore, the fan control model assesses the ambient humidity and determines measures to maintain a comfortable humidity level.

[0082] Additionally, the fan control model determines the optimal wind speed based on the temperature of the person object and the ambient temperature. For example, if the user's temperature is high, the fan control model sets a high wind speed. It also sets the wind direction based on the person object's location. For example, the wind can be directed toward the user. In one embodiment, the fan control model continuously updates the control signal based on real-time data to control the fan. The model generates the control signal using a machine learning model (e.g., a regression model or fuzzy logic) or a rule-based system.

[0083] In addition, the fan control model recognizes a human object through monitoring images collected from the fan's camera, identifies the location and movement pattern of the human object, and rotates along the human object to supply wind.

[0084] To achieve this, the fan control model collects real-time monitoring images via a camera attached to the fan. It then continuously captures and processes the camera images. For example, the fan control model recognizes human objects in the images using deep learning-based object detection algorithms (e.g., YOLO, SSD, Faster R-CNN). Furthermore, in one embodiment, the fan control model can build a dataset from various environments containing people, as illustrated in Figure 5, and train an object recognition model based on this dataset. The location of the recognized human object is then tracked frame by frame. This can be achieved by calculating the center coordinates of the human object across multiple frames. The fan control model then records position changes over time and analyzes movement patterns. This allows it to understand how the person moves relative to the fan. In one embodiment, the fan's rotation angle is calculated based on the current location of the human object. Furthermore, the fan can be controlled to rotate in the direction of the person's location based on the camera's field of view. Furthermore, in one embodiment, the fan control model predicts the direction of movement of the person based on the movement pattern and prepares the fan to rotate in that direction in advance.

[0085] For example, a fan control model can periodically analyze camera footage (e.g., every second) and update the person's position to control the fan. Furthermore, a smoothing algorithm can be applied to adjust the fan's rotation speed to ensure smooth rotation in response to sudden movements.

[0086] In addition, when multiple human objects are recognized in the monitoring image, the fan control model recognizes the number of human objects and the end portions of the human objects as contours, supplies wind to the entire area where multiple human objects exist, recognizes both edges of the area where multiple human objects exist, and adjusts the rotation and wind volume of the fan to deliver wind to both edges.

[0087] To achieve this, the fan control model collects real-time monitoring video from a camera attached to the fan. It then uses deep learning-based object detection algorithms (e.g., YOLO, SSD, Faster R-CNN) to recognize human objects in the video. The outlines of the recognized human objects are then extracted, which can be accomplished using libraries such as OpenCV. Furthermore, the left and right ends of the human objects are detected based on the outlines. This can be accomplished by finding the leftmost and rightmost points of each object. The entire area is then defined based on the outlines and edges of the recognized human objects. This area can extend from the leftmost edge to the rightmost edge. Furthermore, the center point of the entire area is calculated to set the basic rotation direction of the fan, and the rotation angle of the fan is adjusted based on the center and edge locations of the entire area. Furthermore, the fan control model adjusts the airflow rate based on the number of recognized human objects and the overall area size. The airflow rate increases with the number of recognized human objects, while the airflow rate decreases with the number of recognized human objects.

[0088] In addition, the fan control model can recognize a human object in a monitoring image, calculate the area and outline of the human object, and control the fan to supply wind to the outline, thereby providing indirect wind.

[0089] To achieve this, the fan control model collects real-time monitoring video from a camera attached to the fan. It then uses deep learning-based object detection algorithms (e.g., YOLO, SSD, Faster R-CNN) to recognize human objects in the video and extract their outlines. This can be accomplished using libraries such as OpenCV. The area of ​​the human object is then calculated based on the extracted outlines. This area information is used to determine the size of the human object.

[0090] Afterwards, the fan control model determines where and how to supply wind based on the extracted outline. In an embodiment, the model can adjust the direction and strength of the wind according to the shape of the outline. In addition, in an embodiment, the fan control model supplies wind around the outline rather than directly, so as to indirectly provide a sense of coolness. This is to prevent the person from being directly exposed to the wind. For example, the fan control model can supply wind to the person object by controlling the fan in a direction that avoids the coordinates of the center point of the person object and avoids the head area of ​​the person object. In addition, the fan control model can supply wind to the space between the outline and the center point of the person object.

[0091] Additionally, the fan control model determines the fan's rotation angle based on the center point of the outline. Furthermore, the model adjusts the airflow based on the area of ​​the human object. Larger areas provide more airflow, while smaller areas provide less airflow.

[0092] In addition, the fan control model recognizes a human object, estimates the body temperature of the recognized human object using an infrared camera installed in the fan, and if the estimated body temperature is outside the normal range, recognizes it as an abnormal event and transmits the recognized event to the user terminal.

[0093] To this end, the fan control model collects temperature data in real time from a thermal imaging camera or infrared camera installed on the fan. As illustrated in Fig. 4, in the embodiment, the temperature of surrounding objects can be determined by analyzing images acquired through the infrared camera or thermal imaging camera installed on the fan. Furthermore, real-time monitoring images are collected through a standard camera. Then, a deep learning-based object detection algorithm (e.g., YOLO, SSD, Faster R-CNN) is used to recognize human objects in the images. Then, the body temperature of the recognized human object is estimated using data from the infrared camera or thermal imaging camera. This can be accomplished by reading the temperature of each pixel in the infrared image and calculating an average. Furthermore, the body temperature data measured by the infrared camera can be corrected by considering environmental temperature and other factors.

[0094] In addition, the fan control model detects heat-emitting objects, including gas stoves and heaters, based on data collected from a temperature detection sensor and the learning results of the learning data, distinguishes between the detected heat-emitting objects and people, and controls the fan to provide wind only to people. To this end, the fan control model collects ambient temperature data from the temperature detection sensor. In an embodiment, the fan control model collects various temperature patterns (e.g., temperature patterns of people, gas stoves, and heaters) and uses them as learning data. Thereafter, temperature and location information detected in the surroundings are collected together to identify the location and intensity of heat-emitting objects. In addition, the processor (130) trains a machine learning model based on the collected data to distinguish the patterns of specific heat-emitting objects (e.g., gas stoves, heaters, etc.). In addition, temperature and location characteristics are trained to distinguish between heat patterns emitted by people and patterns of high-temperature heat-emitting objects such as gas stoves or heaters. The fan control model then detects surrounding objects using real-time sensor data and uses a machine learning model to distinguish between heat-emitting objects and people. For example, if the temperature in a specific area rises above a certain level and originates from a fixed location, it can be identified as a gas stove or heater. On the other hand, people are more likely to move and exhibit relatively consistent heat patterns, so heaters and people can be distinguished based on their movement and heat patterns.

[0095] In this embodiment, when a person is detected, the fan is adjusted to the corresponding location to provide wind. If a heat-emitting object is detected or a high temperature persists for a certain period of time, the object is determined to be non-human and the fan is disabled. Furthermore, the fan control model can intelligently respond to high ambient temperatures or specific environmental conditions. For example, the fan can be adjusted to operate at a low level when a person is present, and automatically turn off when no person is detected or a heat-emitting object is detected.

[0096] Furthermore, the fan control model classifies the recognized person as an infant or an adult. If the person is an adult, it classifies the person as male or female. If the person is classified as a person, it controls the fan according to a pre-stored control process based on the person's gender and age, providing different winds to each recognized person. To achieve this, the fan control model learns training data to distinguish between infants and adults based on the person's height, body type, and movement patterns. Additionally, for adults, it learns features that can distinguish their gender. The fan control model then recognizes the person's location by combining temperature, location, and distance, and supplements the data based on age and gender. The model then distinguishes between infants and adults. For example, if the height and movement patterns are similar to those of an infant, the recognized object is determined to be an infant. In the embodiment, if an object surrounding the fan is recognized as an adult, the model moves to the next step and performs gender classification. In the embodiment, a gender classification model is used to distinguish between male and female objects recognized as adults. At this stage, the gender can be predicted based on the person's body type, movement, height, and other factors. In the example, the predicted gender information is used for subsequent customized wind control.

[0097] The fan control model then sets the preferred wind speed, direction, and frequency for each individual, including infants, adult men, and adult women. For example, for infants, the fan control model may provide weaker wind and intermittently. For adult men, the wind may be set to be relatively stronger and continuous compared to infants. Furthermore, for adult women, the fan control model may set the wind speed to a level intermediate to that provided for infants and men, periodically adjusting the wind speed.

[0098] Afterwards, the fan control model adjusts the fan's angle, speed, and cycle in real time based on the recognized person's position and condition, thereby providing customized wind. In the embodiment, when the person moves or changes position, it automatically recognizes and adjusts the fan's position and speed to provide wind according to the set process. Furthermore, the fan control model considers the difference between infants and adults and sets it so that strong winds are not provided to infants for safety reasons. Furthermore, real-time monitoring continuously recognizes changes in the person's position and adjusts the wind direction and speed appropriately.

[0099] In addition, the fan control model inputs user-specific data, recognizes the user when the fan is turned on, and controls the fan based on the user-specific data matched to the recognized user, thereby providing the user with the desired wind volume and wind speed. In an embodiment, the user-specific data may include authentication data including facial image data, fingerprints, and passwords of each family member. In addition, in an embodiment, the user-specific data includes the user's preferred control process, and the user's preferred control process includes, but is not limited to, the user's preferred fan control options, such as frontal wind, side wind, wind provision time, wind provision interval, and wind speed.

[0100] In the embodiment, the fan control model pre-enters and stores the preferred wind speed (wind volume) and wind volume (wind speed) for each user. This information can be set differently for each user to provide a personalized environment. The fan control model then stores authentication information, such as facial images, fingerprints, and passwords, for each family member. This information is used to accurately recognize users. In the embodiment, a recognition step is performed when the fan model is operated. During this step, the user is identified through facial recognition, fingerprint recognition, password input, etc. The user's authentication information is then matched to confirm the user's identity. In the embodiment, the fan control model loads the stored preference data (desired wind volume and wind speed) for the user once user recognition is complete. In the embodiment, the fan control model pre-configures profiles so that the wind volume and speed can be set according to each user's preferences. For example, User A may prefer level 3 wind volume and medium wind speed, while User B may prefer level 2 wind volume and low wind speed. Afterwards, the fan control model automatically adjusts the fan's wind volume and speed according to the recognized user's preferences. In one embodiment, the fan control model automatically operates to reflect the user's preferred wind speed and wind volume. For example, if the user prefers level 2 wind volume and medium wind speed, the fan operates according to those settings. In addition, the fan control model provides an interface through which the user can change the settings or input new preferences. For example, the wind volume and speed can be adjusted in real time through the fan remote control or an app. In another embodiment, if an immediate adjustment is made based on the user's input, the new settings are saved and automatically applied in the future. For example, the fan control model can perform automatic learning so that the user's preferred environment is automatically applied each time.In an embodiment, the fan control model can learn the user's patterns and operate the fan with the user's preferred settings at specific times or situations.

[0101] Furthermore, the fan control model recognizes objects around the fan, classifies them as animals or humans, and, if an animal is recognized, executes a pre-stored control process based on the animal type to provide wind. To achieve this, the fan control model recognizes objects around the fan using object recognition sensors (e.g., cameras and infrared sensors). Furthermore, the model uses a machine learning model to distinguish between humans and animals. To achieve this, image data of humans and animals are trained and categorized based on specific patterns. If an object is classified as an animal, the model further distinguishes the animal type. For example, a detailed classification model utilizing animal image data is trained to classify the recognized animal object into dogs, cats, birds, etc. In the embodiment, this distinction is used to apply a control process tailored to each animal type. Furthermore, the fan control model configures a pre-set wind control process based on the animal type. For example, dogs can be sensitive to temperature, so the wind control process is configured to provide continuous wind by setting the wind speed to medium. For cats, a wind control process is configured to intermittently provide a gentle breeze, as they may be sensitive to wind. For birds, a wind control process is configured to adjust the wind direction to prevent direct wind contact and provide indirect wind. In one embodiment, the fan control model can be pre-configured to provide optimal wind for each animal based on its characteristics. The fan control model then automatically adjusts the fan's wind volume, direction, and cycle based on the type of animal recognized. Furthermore, the fan control model adjusts the wind speed and direction as set based on the type and location of the animal, providing a customized environment.

[0102] Afterwards, the fan control model recognizes the object in real time when it moves or changes position, and adjusts the control process as needed. For example, if an animal moves away from the fan or moves, the fan can adjust the wind speed or direction to provide the optimal airflow to the desired location. Furthermore, the fan control model provides customization features that allow users to adjust settings for each animal through the user interface. This allows users to adjust the wind speed or direction to suit specific animals. Furthermore, the fan control model can learn and update its established processes, improving the ability to provide increasingly tailored wind control for a variety of animals.

[0103] Additionally, the fan control model predicts the user's movement path through user motion analysis and controls the direction of the fan according to the predicted path. In an embodiment, the user's real-time location and movement data are collected through motion detection sensors (e.g., cameras, infrared sensors) or movement tracking devices. Then, frame-by-frame data or position coordinates that can recognize the user's movement pattern are recorded to obtain a movement trajectory. The fan control model then analyzes the user's movement pattern. In an embodiment, past movement data and real-time data are combined to predict the user's likely movement path. For example, if the user has a habit of moving along a certain path within a room, the fan control model can learn this and predict the next movement location.

[0104] Furthermore, the fan control model uses predictive models (e.g., RNN, LSTM) to predict the user's next movement path based on real-time location data. The model then calculates the predicted location coordinates by considering the user's movement direction, speed, and pattern, and prepares to adjust the fan's direction to that location. The model then automatically adjusts the fan's direction based on the predicted path information. For example, if the model predicts the user will move to the right, it rotates the fan to the right to ensure the user remains immersed in the wind after the movement. Furthermore, the model smoothly adjusts the fan angle to ensure the user remains comfortably immersed in the wind while moving. If the user's actual movement differs from the predicted path, the model updates the predicted path in real time based on the new location. Furthermore, whenever the user's location data is updated, the predictive model recalculates the path and immediately adjusts the fan's direction.

[0105] Furthermore, the fan control model can recognize the space in which the fan is placed, recognize a human object moving within the recognized space, extract the user's preferred control process of the recognized human object, and control the fan according to the extracted preferred control process. To this end, the fan control model utilizes sensors (e.g., cameras, ultrasonic sensors, infrared sensors, etc.) to recognize the structure and size of the space in which the fan is located. Then, the model determines the usable area within the recognized space and the angular range of the fan to define the range within which the fan can provide wind. Furthermore, the fan control model analyzes the characteristics of the space (e.g., room shape, furniture position, etc.) to enable the fan to efficiently provide wind. In one embodiment, the fan control model recognizes a human object moving within the space in real time using a motion detection and object tracking model. In one embodiment, the model tracks the user's location in real time whenever the user moves and prepares to adjust the fan according to the location. Afterwards, the fan control model accurately identifies the user by utilizing user-specific authentication information (e.g., facial recognition, fingerprint, password, etc.) to identify the recognized human object. The fan control model then loads the identified user's preferred control process. The loaded information includes pre-saved, user-specific, customized settings, including the user's preferred wind speed, direction, frequency, and wind pattern.

[0106] Afterwards, the fan control model adjusts the fan's wind volume, wind speed, and direction according to the user's preferred control process. For example, if a certain user prefers a strong wind, the fan volume is set to high, and if another user prefers a soft wind, the fan speed is set to low. In an embodiment, the fan control model adjusts the angle and direction of the fan according to the user's location as he or she moves, so that the desired wind can be provided even in the moving location. Thereafter, the fan control model updates in real time and immediately adjusts the fan settings when the user moves or changes the preferred wind strength or direction. In an embodiment, if the user repeatedly changes the settings in a specific space as needed, the new preferences can be updated and reflected through an automatic learning function.

[0107] In addition, the fan control model tracks a human object for a certain period of time, identifies the movement path of the human object in the recognized space, and limits the rotation range of the fan to be included in the movement path. To this end, the fan control model tracks the location and movement path of the human object in real time for a certain period of time using motion detection and object tracking sensors. In the embodiment, the movement pattern of the human object (e.g., left-right movement, repetitive movement in a specific area) is recorded and analyzed based on the tracking data. Thereafter, the movement path of the human object recorded for a certain period of time is analyzed to identify the main movement range. For example, it determines whether the human object prefers a specific area or an area in which the human object repeatedly moves within the room. Thereafter, the fan control model identifies the start and end points of the movement path, calculates the coordinates of the entire range of movement of the human object, and sets the movement range.

[0108] Afterwards, the fan control model limits the fan's rotation angle range based on the analyzed movement path information. In the embodiment, the fan's rotation range is set to not exceed the person's movement path, depending on the person's movement range. For example, if the person moves from the left to the right end of the room, the fan's rotation range is limited to that section, thereby limiting the rotation angle. Furthermore, the fan control model adjusts the rotation range settings to ensure that the fan can efficiently provide wind.

[0109] In this embodiment, the fan control model updates the movement path in real time when a person moves along a new path or begins moving in a different direction from the original path. If the path changes, the fan control model recalculates the fan's rotation range and resets the rotation angle to match the new path. This ensures that the fan is adjusted to the optimal range to always follow the person as they move.

[0110] Over time, the fan control model learns human movement patterns and can automatically adjust its rotation range to match frequently used travel paths. For example, if a user repeatedly moves along a specific path, the fan's rotation range will learn to remain fixed on that path, automatically optimizing its rotation range.

[0111] Additionally, the fan rotation range can be manually adjusted by the user, allowing the user to customize the rotation range according to their preference. The interface allows the user to limit the rotation angle or adjust the rotation range to suit specific sections.

[0112] In addition, the fan control model recognizes a human object, estimates the body temperature of the recognized human object using an infrared camera installed in the fan, and recognizes an abnormal event when the estimated body temperature is outside the normal range and transmits the recognized event to a user terminal. In an embodiment, the fan control model sets a general body temperature range (e.g., 36.5°C to 37.5°C), and recognizes an abnormal event when the estimated body temperature is outside the set normal range. Thereafter, the fan control model records an abnormal body temperature event and stores the occurrence time, location, and body temperature information of the event. In an embodiment, when an abnormal event is detected, it is transmitted to a user terminal (e.g., a smartphone, a computer, etc.).

[0113] In addition, the fan control model recognizes the ambient temperature and the temperature of each human object in real time, tracks the change in the human object temperature according to the wind supply time, and recognizes it as an abnormal event if there is no change in the human object temperature or if it exceeds the normal temperature even after the wind supply time exceeds a certain period of time.

[0114] To achieve this, the fan control model collects the temperature of human objects and ambient temperature in real time via an infrared camera attached to the fan. It then uses a deep learning-based object detection algorithm to recognize human objects in the image and uses the infrared camera data to estimate the body temperature of the recognized human object. The body temperature data measured by the infrared camera is then corrected to account for environmental temperature and other factors.

[0115] Additionally, the body temperature and ambient temperature of the human object are monitored in real time, and the temperature change of the human object according to the wind supply time is recorded.

[0116] The fan control model then tracks the time the fan has been supplying air, detecting instances where the temperature of the human object remains unchanged or exceeds the normal range after a certain period of time. Furthermore, a typical body temperature range (e.g., 36.5°C to 37.5°C) is set to detect abnormal body temperature: instances where the temperature remains unchanged for a certain period of time after air supply or the body temperature exceeds the normal range are recognized as abnormal events.

[0117] Afterwards, the fan control model records abnormal body temperature events and stores the time, location, and body temperature information of the event. Furthermore, when an abnormal event is detected, the fan control model transmits the time, location, and body temperature information to the user's terminal (smartphone, computer, etc.).

[0118] In addition, the fan control model supplies different wind speeds depending on the body temperature of the human object, and if the body temperature of the human object remains high even after supplying wind to the person with a high body temperature for a certain period of time, the fan control model determines that there is a health problem and reduces the wind speed again.

[0119] In addition, in the embodiment, the fan control model recognizes multiple human objects, and if there is a human object among the recognized human objects whose temperature is higher than the normal range, the location of the human object is transmitted to the registered user terminal. To this end, the fan control model collects the body temperature of the human object in real time from an infrared camera. Furthermore, the ambient temperature is continuously monitored. A deep learning-based object detection algorithm is then used to recognize the human object in the image. The body temperature of the recognized human object is then estimated using data from the infrared camera, and the body temperature data measured by the infrared camera is corrected by considering the environmental temperature and other factors. In the embodiment, a normal body temperature range is set, and cases where the estimated body temperature falls outside the set normal range are detected. For example, a human object whose temperature falls outside the normal range is identified. The fan control model then tracks the location of the recognized human object in real time and records the location of the human object whose temperature falls outside the normal range. Furthermore, the fan control model records abnormal body temperature events and stores the time, location, and body temperature information of the event. If an abnormal event is detected, the fan control model transmits the information to the user terminal.

[0120] Additionally, the fan control model adjusts the fan's wind direction and speed based on the type of space in which the fan is located, the location of the recognized human objects, and the distance between each human object and the fan. To achieve this, the fan control model collects real-time monitoring video via a camera attached to the fan.

[0121] To achieve this, the fan control model uses ultrasonic sensors or LiDAR to measure the distance between a person object and the fan, and analyzes the monitoring video to determine the type of space (e.g., living room, bedroom, office) where the fan is located. Then, a deep learning-based object detection algorithm is used to recognize the person object in the video, build a dataset of various environments containing people, and train an object recognition model based on this data. The fan control model then uses a distance sensor to measure the distance between the recognized person object and the fan in real time. Furthermore, the fan's operating mode is set according to the space type. For example, the fan control model can be set to provide a wider range of air in the living room, and a softer, quieter air in the bedroom.

[0122] Additionally, the fan control model identifies the location of human objects and records the coordinates of each human object. It then adjusts the wind direction and strength based on the distance between each human object and the fan. In one embodiment, the fan control model can provide a weak wind to a nearby person and a strong wind to a distant person. Furthermore, the fan control model adjusts the wind direction of the fan based on the location of the human object. For example, if the person is on the right, the fan rotates to the right.

[0123] Additionally, the airflow is adjusted based on the distance from the person. A weak wind is supplied to a person at a close distance, and a strong wind is supplied to a person at a far distance.

[0124] Referring back to FIG. 3, the feedback unit (124) evaluates the learned artificial neural network model and deep learning model. In an embodiment, the feedback unit (124) may evaluate the artificial neural network model through at least one of accuracy, precision, and recall. Accuracy is an index that measures how well the results predicted by the artificial neural network model match the actual results. Precision is an index that measures the proportion of actual positives among the results predicted as positive. Recall is an index that measures the proportion of actual positives predicted by the model as positive. In an embodiment, the feedback unit (124) may calculate the accuracy, precision, and recall of the artificial neural network model, and evaluate the artificial neural network model based on at least one of the calculated indexes.

[0125] In an embodiment, the feedback unit (124) can measure the accuracy of the artificial neural network model using an evaluation dataset. The evaluation dataset consists of data that the model did not use for training and is used to objectively evaluate the model's performance. In an embodiment, the feedback unit (124) executes the artificial neural network model using the evaluation dataset and compares the artificial neural network model's predicted value for each input data with the actual correct answer value of the corresponding data. Thereafter, the accuracy of the model's predictions can be measured based on the comparison results. For example, the accuracy in the feedback unit (124) can be calculated as the ratio of data correctly predicted by the model among the entire data.

[0126] In addition, the feedback unit (124) can calculate the F1 score, which is an index indicating the balance of precision and recall, which is an index calculated as the harmonic mean of precision and recall, evaluate the artificial neural network model based on the calculated F1 score, generate an AUC-ROC curve, which is an index that visualizes the performance of the classification model in a graph, and evaluate the artificial neural network model based on the generated AUC-ROC curve. In an embodiment, the feedback unit (124) can evaluate that the performance of the model is better as the area under the ROC curve (AUC) is closer to 1.

[0127] In addition, the feedback unit (124) can evaluate the interpretability of the artificial neural network model. In an embodiment, the feedback unit (124) evaluates the interpretability of the artificial neural network model through SHAP (SHapley Additive exPlanations) and LIME (Local Interpretable Model-agnostic Explanations) methods. SHAP (SHapley Additive exPlanations) is a library that provides an interpretation of the results predicted by the model, and the feedback unit (124) extracts SHAP values ​​from the library. In an embodiment, the feedback unit (124) can predict how much the characteristic information input to the model influenced the model prediction through the SHAP value extraction.

[0128] The Local Interpretable Model-agnostic Explanations (LIME) method is a method for explaining model predictions for individual samples. In one embodiment, the feedback unit (124) uses the LIME method to approximate the sample as an interpretable model and calculate the importance of each characteristic. Furthermore, the feedback unit (124) can estimate the influence of each characteristic variable by analyzing the model's internal weights and bias values.

[0129] The feedback unit (124) performs improvement work when the fairness of the artificial neural network model is low or shows discrimination. In an embodiment, the feedback unit (124) collects additional data representing the specific group when the data for the specific group is insufficient by a certain level or more and performs a data preprocessing process. In an embodiment, the feedback unit (124) performs a data preprocessing process including data normalization, outlier removal, and data scaling to prevent the model from learning unnecessary patterns. In addition, in an embodiment, the feedback unit (124) can prevent discrimination or ensure fairness by adding specific conditions to the model learning algorithm.

[0130] In an embodiment, the feedback unit (124) evaluates the performance of the model by comparing the model's predicted results with actual results through confusion matrix analysis to ensure fairness. A confusion matrix is ​​a matrix that evaluates the classification performance of a model in supervised learning. The confusion matrix displays the classification results by comparing the model's predicted results with actual results. In an embodiment, the feedback unit (124) can evaluate the performance of the model by calculating the accuracy and misclassification rate for each class through confusion matrix analysis.

[0131] Additionally, in the embodiment, the feedback unit (124) enables the distribution of data to be confirmed through visual analysis of learning data. For example, in the case of image data, image samples for each class can be visualized to evaluate the diversity and fairness of the data.

[0132] In addition, the feedback unit (124) verifies the fairness and diversity of the learning data and improves the artificial neural network model through fairness verification and evaluation index calculation. In an embodiment, fairness verification is to check whether the artificial neural network model shows discrimination for specific data attributes with respect to the learning information. In an embodiment, the feedback unit (124) can check whether discrimination for specific attributes is present by comparing the number of samples for each attribute or evaluating the classification performance for each attribute.

[0133] Additionally, the feedback unit (124) calculates various indicators to evaluate the performance of the artificial neural network model. For example, model performance can be evaluated by calculating indicators such as accuracy, precision, recall, and F1 score. At this time, indicators for each class can be calculated to evaluate the fairness and diversity of the model.

[0134] In addition, the feedback unit (124) collects feedback on problems that occur when the artificial neural network model is used in an actual environment, and continuously improves the artificial neural network model by reflecting the collected feedback in the artificial neural network model.

[0135] As described above, the AI ​​fan and its wind supply method utilize an AI model to recognize a human object through a camera located at the center of the fan, thereby determining the person's location and movement. This allows the fan to rotate and supply wind following the person, ensuring the user always feels cool in the optimal location.

[0136] Additionally, the embodiment uses an artificial intelligence model to detect the surrounding environment, temperature, humidity, and human body temperature. Based on this detected data, the optimal airflow is calculated and provided, creating a more comfortable environment for the user.

[0137] Additionally, the embodiment provides a function that rotates around the person to prevent direct wind from reaching the person, thereby reducing discomfort caused by direct wind and creating a more comfortable environment through indirect wind supply.

[0138] In addition, when multiple human objects are recognized, the number of human objects and the end portions are recognized as outlines to supply wind evenly to all people, and in particular, in a space with multiple people, the wind can be evenly distributed by rotating from end to end.

[0139] In addition, the fan can be controlled through voice recognition, allowing users to use the fan more conveniently. When multiple human objects exist, if a specific person's body temperature is higher than the normal range, it detects this as an abnormality and notifies the user of the abnormality through the application, allowing the user to quickly recognize and respond to their health status.

[0140] In addition, in the embodiment, different wind speeds are supplied according to the body temperature of a human object, and if the body temperature remains high even after supplying wind to a person with a high body temperature for a certain period of time, the wind speed is reduced again to determine that there is a health problem, and the abnormality is notified through an application, thereby maximizing the convenience and comfort of the user and monitoring the health status.

[0141] A model in this specification may refer to any form of computer program that operates based on a network function, an artificial neural network, and / or a neural network. Throughout this specification, the terms model, neural network, network function, and neural network may be used interchangeably. A neural network is a network in which one or more nodes are interconnected through one or more links to form input node and output node relationships within the neural network. The characteristics of a neural network can be determined based on the number of nodes and links within the neural network, the correlation between the nodes and links, and the weight value assigned to each link. A neural network may be composed of a set of one or more nodes. A subset of the nodes constituting the neural network may constitute a layer.

[0142] A deep neural network (DNN) may refer to a neural network that includes multiple hidden layers in addition to an input layer and an output layer. A deep neural network may include a convolutional neural network (CNN), a recurrent neural network (RNN), an autoencoder, a generative adversarial network (GAN), a restricted boltzmann machine (RBM), a deep belief network (DBN), a Q network, a U network, a Siamese network, a generative adversarial network (GAN), a transformer, and the like. The description of the above-described deep neural network is merely an example, and the present disclosure is not limited thereto.

[0143] Neural networks can learn through at least one of the following methods: supervised learning, unsupervised learning, semi-supervised learning, self-supervised learning, or reinforcement learning. Neural network learning can be the process of applying knowledge to the neural network to perform a specific action.

[0144] Neural networks can be trained to minimize output errors. This process involves repeatedly inputting training data into the neural network, calculating the neural network output and target error for the training data, and backpropagating the neural network error from the output layer to the input layer to update the weights of each node in the neural network to reduce the error. In supervised learning, labeled data is used for each training data, while unsupervised learning uses unlabeled data. The amount of change in the connection weights of each updated node can be determined by the learning rate. The neural network's calculation of input data and backpropagation of errors can constitute a learning cycle (epoch). The learning rate can vary depending on the number of iterations in the neural network's training cycle. Additionally, to prevent overfitting, methods such as increasing the training data, regularization, dropout that disables some nodes, and batch normalization layers can be applied.

[0145] In one embodiment, the model may borrow at least a portion of a transformer. The transformer may be composed of an encoder that encodes embedded data and a decoder that decodes the encoded data. The transformer may have a structure that receives a series of data and outputs a series of data of different types through encoding and decoding steps. In one embodiment, the series of data may be processed into a form operable by the transformer. The process of processing the series of data into a form operable by the transformer may include an embedding process. Expressions such as data tokens, embedding vectors, and embedding tokens may refer to data embedded in a form operable by the transformer.

[0146] To encode and decode a series of data, a transformer can utilize an attention algorithm to process the encoders and decoders within the transformer. An attention algorithm can refer to an algorithm that, for a given query, calculates the similarity for one or more keys, reflects this similarity in the values ​​corresponding to each key, and then weights and sums the values ​​to which the similarity is reflected to calculate an attention value.

[0147] Depending on how the query, key, and value are configured, various types of attention algorithms can be categorized. For example, if attention is obtained by setting the query, key, and value all to the same value, this could be a self-attention algorithm. If attention is obtained by reducing the dimensionality of the embedding vector to process a series of input data in parallel and then generating individual attention heads for each segmented embedding vector, this could be a multi-head attention algorithm.

[0148] In one embodiment, the transformer may be composed of modules that perform multiple multi-head self-attention algorithms or multi-head encoder-decoder algorithms. In one embodiment, the transformer may also include additional components other than attention algorithms, such as embedding, normalization, and softmax. Methods for constructing a transformer using attention algorithms may include methods disclosed in Vaswani et al., Attention Is All You Need, 2017 NIPS, which is incorporated herein by reference.

[0149] A transformer can be applied to various data domains, such as embedded natural language, segmented image data, and audio waveforms, to transform a series of input data into a series of output data. To transform data with various data domains into a series of data that can be input to a transformer, the transformer can embed the data. The transformer can process additional data that expresses the relative positional relationship or phase relationship between the series of input data. Alternatively, vectors expressing the relative positional relationship or phase relationship between the input data can be additionally reflected in the series of input data to embed the series of input data. In one example, the relative positional relationship between the series of input data may include, but is not limited to, word order within a natural language sentence, the relative positional relationship between each segmented image, and the time order of segmented audio waveforms. The process of adding information expressing the relative positional relationship or phase relationship between the series of input data may be referred to as positional encoding.

[0150] In one embodiment, the model may include, but is not limited to, at least one of a Recurrent Neural Network (RNN), a Long Short Term Memory (LSTM) network, a Deep Neural Network (DNN), a Convolutional Neural Network (CNN), and a Bidirectional Recurrent Deep Neural Network (BRDNN).

[0151] In one embodiment, the model may be a model trained using transfer learning. Transfer learning, in this context, refers to a learning method that pre-trains a large amount of unlabeled training data using semi-supervised or self-learning methods to obtain a pre-trained model for a first task, then fine-tunes the pre-trained model to suit a second task, and trains it on labeled training data using supervised learning to implement a target model.

[0152] The disclosed content is merely an example, and various modifications and implementations can be made by a person skilled in the art without departing from the gist of the claims claimed in the patent, so the scope of protection of the disclosed content is not limited to the specific embodiments described above.

[0153] According to the embodiment, the AI ​​fan and its wind supply method utilize an AI model to recognize a human object through a camera located at the center of the fan, thereby determining the person's location and movement. This allows the fan to rotate and supply wind following the person, ensuring the user always feels cool in the optimal location.

[0154] Additionally, the embodiment uses an artificial intelligence model to detect the surrounding environment, temperature, humidity, and human body temperature. Based on this detected data, the optimal airflow is calculated and provided, creating a more comfortable environment for the user.

[0155] Additionally, the embodiment provides a function that rotates around the person to prevent direct wind from reaching the person, thereby reducing discomfort caused by direct wind and creating a more comfortable environment through indirect wind supply.

[0156] In addition, when multiple human objects are recognized, the number of human objects and the end portions are recognized as outlines to supply wind evenly to all people, and in particular, in a space with multiple people, the wind can be evenly distributed by rotating from end to end.

[0157] In addition, the fan can be controlled through voice recognition, allowing users to use the fan more conveniently. When multiple human objects exist, if a specific person's body temperature is higher than the normal range, it detects this as an abnormality and notifies the user of the abnormality through the application, allowing the user to quickly recognize and respond to their health status.

[0158] In addition, in the embodiment, different wind speeds are supplied according to the body temperature of a human object, and if the body temperature remains high even after supplying wind to a person with a high body temperature for a certain period of time, the wind speed is reduced again to determine that there is a health problem, and the abnormality is notified through an application, thereby maximizing the convenience and comfort of the user and monitoring the health status.

[0159] According to the embodiment, the AI ​​fan and its wind supply method utilize an AI model to recognize a human object through a camera located at the center of the fan, thereby determining the person's location and movement. This allows the fan to rotate and supply wind following the person, ensuring the user always feels cool in the optimal location.

[0160] Additionally, the embodiment uses an artificial intelligence model to detect the surrounding environment, temperature, humidity, and human body temperature. Based on this detected data, the optimal airflow is calculated and provided, creating a more comfortable environment for the user.

[0161] Additionally, the embodiment provides a function that rotates around the person to prevent direct wind from reaching the person, thereby reducing discomfort caused by direct wind and creating a more comfortable environment through indirect wind supply.

[0162] In addition, when multiple human objects are recognized, the number of human objects and the end portions are recognized as outlines to supply wind evenly to all people, and in particular, in a space with multiple people, the wind can be evenly distributed by rotating from end to end.

[0163] In addition, the fan can be controlled through voice recognition, allowing users to use the fan more conveniently. When multiple human objects exist, if a specific person's body temperature is higher than the normal range, it detects this as an abnormality and notifies the user of the abnormality through the application, allowing the user to quickly recognize and respond to their health status.

[0164] In addition, in the embodiment, different wind speeds are supplied according to the body temperature of a human object, and if the body temperature remains high even after supplying wind to a person with a high body temperature for a certain period of time, the wind speed is reduced again to determine that there is a health problem, and the abnormality is notified through an application, thereby maximizing the convenience and comfort of the user and monitoring the health status.

[0165] 100: Artificial Intelligence Fan

Claims

1. A memory storing at least one command for supplying wind to an artificial intelligence fan; and A processor comprising: a processor that performs an operation according to the above command; The above processor, We implement a fan control model using learning data for object recognition, wind direction and wind speed control. The above fan control model Collect sensor data from the sensor installed in the above fan, Calculate the location and number of human objects recognized from the above sensor data, the type of human objects, and the wind speed and direction according to the temperature and ambient temperature and humidity, and control the fan to supply wind according to the calculated wind speed and direction. An artificial intelligence fan, wherein the above sensors include a camera, a temperature sensor, and a humidity sensor, and the sensor data includes a user monitoring image, ambient temperature, and user temperature and ambient humidity.

2. In the first paragraph, the fan control model Detect heat-emitting objects, including gas stoves and heaters, based on the data collected from the temperature detection sensor and the learning results of the learning data. An artificial intelligence fan that distinguishes between the detected heat-emitting objects and people and controls the fan to provide wind only to the people.

3. In the second paragraph, the fan control model Distinguish the recognized person as an infant or an adult, and if the person is an adult, Distinguish between male and female An artificial intelligence fan that, when identified as a person, controls the fan according to a pre-stored control process based on gender and age, thereby providing different wind to each recognized person.

4. In the third paragraph, the fan control model Enter user-specific data, recognize the user when the fan is turned on, and control the fan based on the user-specific data matched to the recognized user, thereby providing the desired wind volume and wind speed for the user. The above user-specific data is An artificial intelligence fan that contains authentication data including facial image data, fingerprints, and passwords of each family member.

5. In the third paragraph, the fan control model An artificial intelligence fan that recognizes objects around the fan, classifies the recognized objects as animals or people, and, if an animal object is recognized, performs a pre-stored control process according to the type of animal to provide wind.

6. In paragraph 4, the user-specific data is Includes user preference control processes, The above user's preferred control processor is With the user's preferred fan control options, An artificial intelligence fan that includes front wind, side wind, wind provision time, wind provision interval, and wind speed.

7. In the first paragraph, the fan control model An artificial intelligence fan that predicts the user's movement path through user motion analysis and controls the direction of the fan according to the predicted path.

8. In paragraph 7, the fan control model An artificial intelligence fan that recognizes a space in which a fan is placed, recognizes a human object moving in the recognized space, extracts a user preference control processor of the recognized human object, and controls the fan according to the extracted preference control process.

9. In paragraph 8, the fan control model An artificial intelligence fan that tracks a human object for a certain period of time, determines the movement path of the human object in the recognized space, and limits the rotation range of the fan to be included in the movement path.

10. In the first paragraph, the fan control model An artificial intelligence fan that recognizes a human object, estimates the body temperature of the recognized human object using an infrared camera installed in the fan, and if the estimated body temperature is outside the normal range, recognizes it as an abnormal event and transmits the recognized event to a user terminal.

Images