WO2024135885A1 - Artificial intelligence-based biometric data quality evaluation method, device, and computer program for securing quality-verified biometric data - Google Patents

Abstract

An artificial intelligence-based biometric data quality evaluation method, device, and computer program for securing quality-verified biometric data are provided. An artificial intelligence-based biometric data quality evaluation method for securing quality-verified biometric data according to various embodiments of the present invention relates to an artificial intelligence-based biometric data quality evaluation method performed by a computing device, the method comprising the steps of: obtaining a subject's biometric data; and evaluating the quality of the obtained biometric data by using a pre-trained artificial intelligence model, wherein the pre-trained artificial intelligence model is a model that reduces a dimension of the obtained biometric data and evaluates the quality of the obtained biometric data according to whether noise is detected from the dimension-reduced biometric data.

Description

Artificial intelligence-based biometric data quality evaluation method, device, and computer program to secure quality-verified biometric data

Various embodiments of the present invention relate to an artificial intelligence-based biometric data quality evaluation method, device, and computer program for securing quality-verified biometric data.

In order to determine the patient's health status or diagnose whether the patient has a specific disease, it is necessary to measure the patient's biometric data and analyze the measured biometric data.

Conventionally, medical staff at medical institutions use biometric data measurement devices to periodically measure a patient's biometric data such as blood pressure, pulse, body temperature, electrocardiogram, oxygen saturation, and brain waves, and the medical staff directly checks the periodically measured biometric data. and reading to understand the patient's condition or diagnose a disease. However, with the recent development of information and communication technology, various technologies have been developed in the medical field to analyze the patient's biometric data, and these technologies are used to analyze the patient's biometric data. Through analysis, an environment has been created that can more easily and accurately determine whether a patient's health status or disease is present.

Meanwhile, in order to determine a patient's health status or diagnose a patient's disease using biometric data analysis technology implemented in the form of a computer program, a minimum amount of biometric data (e.g., 5 minutes of biometric data) is required. In the case of biometric data measurement, human error is likely to occur because it is still performed directly by humans, and as a result, biometric data may not be measured properly or unnecessary noise may be included in the measured biometric data.

If normal biometric data cannot be obtained as described above, it is difficult to accurately determine the patient's health status or accurately diagnose the patient's disease, so a method to secure the minimum normal data for diagnosing the patient's condition and disease is required. It is becoming.

The problem to be solved by the present invention is to solve the above-described conventional problems by evaluating the quality of biometric data collected from patients using a pre-learned artificial intelligence model to determine whether the biometric data contains noise. It provides an artificial intelligence-based biometric data quality evaluation method, device, and computer program to secure quality-verified biometric data that can determine and thereby secure normal biometric data without noise.

The problems to be solved by the present invention are not limited to the problems mentioned above, and other problems not mentioned can be clearly understood by those skilled in the art from the description below.

An artificial intelligence-based biometric data quality evaluation method according to an embodiment of the present invention to solve the above-described problem is an artificial intelligence-based biometric data quality evaluation method performed by a computing device, which includes obtaining the subject's biometric data. It may include evaluating the quality of the acquired biometric data using steps and a previously learned artificial intelligence model.

In various embodiments, the pre-trained artificial intelligence model is a model that reduces the dimension of the acquired biometric data and evaluates the quality of the acquired biometric data depending on whether noise is detected from the dimensionally reduced biometric data. It can be.

In various embodiments, the pre-trained artificial intelligence model includes an encoder including one or more dimension reduction layers and a decoder including one or more dimension restoration layers, and uses a plurality of biometric data without noise as learning data. and an auto-encoder learned based on unsupervised learning, wherein the step of evaluating the quality of the acquired biometric data includes preprocessing the acquired biometric data, and processing the preprocessed biometric data into the encoder. It may include generating low-dimensional biometric data as input and detecting noise by analyzing the generated low-dimensional biometric data.

In various embodiments, the step of preprocessing the acquired biometric data includes a first biometric data value at a first time point included in the acquired biometric data and second biometric data at a second time point before the first time point. If the difference between the values and the first biometric data value and the third biometric data value at a third time point after the first time point is greater than or equal to a threshold value, initializing the first biometric data value and the second biometric data value It may include interpolating the initialized first biometric data value using the data value and the third biometric data value.

In various embodiments, the acquired biometric data is connected to first unit biometric data and the first unit biometric data measured over a predetermined period from a first point in time, and after measurement of the first unit biometric data is completed. It includes second unit biometric data measured over a predetermined period from a second time point after a predetermined time, and preprocessing the obtained biometric data includes the last biometric data value of the first unit biometric data and the second unit biometric data. calculating a difference between the first biometric data values of the unit biometric data; and when the calculated difference is greater than or equal to the threshold, the first biometric data of the second unit biometric data based on the last biometric data value of the first unit biometric data. It may include correcting a value and interpolating the first biometric data of the corrected second unit biometric data and the first biometric data value included in the second biometric data.

In various embodiments, the acquired biometric data includes photoplethysmogram (PPG) data collected in real time from the subject, and preprocessing the acquired biometric data includes a plurality of photoplethysmograms included in the photoplethysmographic data. Identifying a peak, calculating a plurality of peak prominences corresponding to each of the identified plurality of peaks, and generating a peak prominence histogram using the calculated plurality of peak prominences. Analyzing the generated peak saliency histogram to identify a peak saliency section having a predetermined pattern, setting the peak saliency corresponding to the identified peak saliency section as a reference peak saliency, and calculating the peak saliency It may include selecting a peak saliency lower than the set reference peak saliency among a plurality of peak saliencies, and removing a peak corresponding to the selected peak saliency among the plurality of peaks.

In various embodiments, the step of detecting noise includes clustering a plurality of biometric data values included in the generated low-dimensional biometric data and based on a result of clustering the plurality of biometric data values, the obtained It may include a step of determining whether the biometric data contains noise.

In various embodiments, generating a plurality of biometric data sections by dividing the acquired biometric data - the acquired biometric data includes photoplethysmographic data collected in real time from the subject - into preset time units. , calculating a peak saliency corresponding to each of the plurality of generated biometric data sections using biometric data values included in each of the plurality of generated biometric data sections, and selecting one of the plurality of generated biometric data sections. If the peak saliency calculated for one biometric data section is less than a threshold value, the method may further include determining that noise is included in the one biometric data section.

In various embodiments, the step of determining that noise is included in any one biometric data section includes, when it is determined that noise is included in any one biometric data section, dividing the one biometric data section into a plurality of Dividing into unit sections and calculating a peak saliency for each of the plurality of divided unit sections, and dividing at least one unit section in which the calculated peak saliency is less than a threshold value among the plurality of divided unit sections into a noise section. It may include a step of specifying.

In various embodiments, a divergence histogram corresponding to the acquired biometric data is generated, the generated divergence histogram is analyzed, and any one divergence section having a predetermined pattern among a plurality of divergence sections included in the generated divergence histogram is determined. If present, the method may further include determining that noise is included in the acquired biometric data.

In various embodiments, the pre-trained artificial intelligence model includes an encoder including one or more dimension reduction layers and a decoder including one or more dimension restoration layers, and uses a plurality of biometric data without noise as learning data. It includes an auto-encoder and a one-dimensional convolutional neural network model learned based on unsupervised learning, and the step of evaluating the quality of the acquired biometric data includes converting the acquired biometric data into the one-dimensional convolutional neural network model. Extracting features related to time series patterns of the acquired biometric data as they are input to a neural network model, and generating low-dimensional biometric data as the extracted features and the acquired biometric data are input to the encoder, and generating the low-dimensional biometric data. It may include detecting noise by analyzing low-dimensional biometric data.

In various embodiments, the step of evaluating the quality of the acquired biometric data includes, when a specific noise is detected by analyzing the acquired biometric data, determining that the acquired biometric data includes the specific noise; If the period in which the specific noise is detected is less than a preset period, the method may include determining that the acquired biometric data does not contain the specific noise.

A computing device that performs an artificial intelligence-based biometric data quality evaluation method according to another embodiment of the present invention to solve the above-described problem includes a processor, a network interface, a memory, and a memory, and is loaded into the memory, and is operated by the processor. Includes an executable computer program, wherein the computer program may include instructions for acquiring biometric data of a subject and instructions for evaluating the quality of the acquired biometric data using a previously learned artificial intelligence model. .

A computer program according to another embodiment of the present invention for solving the above-described problem is combined with a computing device to obtain biometric data of a target and the quality of the acquired biometric data using a previously learned artificial intelligence model. It can be stored in a recording medium that can be read by a computing device in order to execute an artificial intelligence-based biometric data quality evaluation method that includes the step of evaluating.

Other specific details of the invention are included in the detailed description and drawings.

According to various embodiments of the present invention, by evaluating the quality of biometric data collected from a patient using a previously learned artificial intelligence model, it is possible to determine whether the biometric data contains noise, and accordingly, noise is included. There is an advantage in being able to secure normal biometric data that has not been released.

The effects of the present invention are not limited to the effects mentioned above, and other effects not mentioned will be clearly understood by those skilled in the art from the description below.

Figure 1 is a diagram illustrating an artificial intelligence-based biometric data quality evaluation system according to an embodiment of the present invention.

Figure 2 is a hardware configuration diagram of a computing device that performs an artificial intelligence-based biometric data quality evaluation method according to another embodiment of the present invention.

Figure 3 is a flow chart of an artificial intelligence-based biometric data quality evaluation method according to another embodiment of the present invention.

FIG. 4 is a flowchart illustrating a method of preprocessing biometric data using the amount of change in biometric data values, in various embodiments.

FIG. 5 is a flowchart illustrating a method of preprocessing biometric data in which a plurality of units of biometric data are combined in various embodiments.

FIG. 6 is a graph illustrating one biometric data in which different unit biometric data are combined in various embodiments.

FIG. 7 is a flowchart illustrating a method of preprocessing biometric data using peak saliency in various embodiments.

FIG. 8 is a graph illustrating peak saliency of biometric data in various embodiments.

FIG. 9 is a graph illustrating biometric data with identified peaks in various embodiments.

Figure 10 is a diagram illustrating a peak saliency histogram according to various embodiments.

11 and 12 are diagrams illustrating pre-trained artificial intelligence models in various embodiments.

FIG. 13 is a diagram illustrating a process for detecting noise from biometric data using a previously learned artificial intelligence model in various embodiments.

FIG. 14 is a diagram illustrating a process for evaluating the quality of biometric data based on peak saliency in various embodiments.

Figure 15 is a diagram to explain the process of evaluating the quality of biometric data based on the divergence histogram.

FIG. 16 is a diagram for explaining a process of filtering temporarily detected noise in various embodiments.

17 and 18 are diagrams illustrating a user interface (UI) that outputs quality evaluation results derived as an artificial intelligence-based biometric data quality evaluation method is performed, in various embodiments.

The advantages and features of the present invention and methods for achieving them will become clear by referring to the embodiments described in detail below along with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below and may be implemented in various different forms. The present embodiments are merely provided to ensure that the disclosure of the present invention is complete and to provide a general understanding of the technical field to which the present invention pertains. It is provided to fully inform the skilled person of the scope of the present invention, and the present invention is only defined by the scope of the claims.

The terminology used herein is for describing embodiments and is not intended to limit the invention. As used herein, singular forms also include plural forms, unless specifically stated otherwise in the context. As used in the specification, “comprises” and/or “comprising” does not exclude the presence or addition of one or more other elements in addition to the mentioned elements. Like reference numerals refer to like elements throughout the specification, and “and/or” includes each and every combination of one or more of the referenced elements. Although “first”, “second”, etc. are used to describe various components, these components are of course not limited by these terms. These terms are merely used to distinguish one component from another. Therefore, it goes without saying that the first component mentioned below may also be a second component within the technical spirit of the present invention.

Unless otherwise defined, all terms (including technical and scientific terms) used in this specification may be used with meanings commonly understood by those skilled in the art to which the present invention pertains. Additionally, terms defined in commonly used dictionaries are not to be interpreted ideally or excessively unless clearly specifically defined.

As used in the specification, the term “unit” or “module” refers to a hardware component such as software, FPGA, or ASIC, and the “unit” or “module” performs certain roles. However, “part” or “module” is not limited to software or hardware. A “unit” or “module” may be configured to reside on an addressable storage medium and may be configured to run on one or more processors. Thus, as an example, a “part” or “module” refers to components such as software components, object-oriented software components, class components, and task components, processes, functions, properties, Includes procedures, subroutines, segments of program code, drivers, firmware, microcode, circuits, data, databases, data structures, tables, arrays, and variables. The functionality provided within components and “parts” or “modules” can be combined into smaller components and “parts” or “modules” or into additional components and “parts” or “modules”. Could be further separated.

Spatially relative terms such as “below”, “beneath”, “lower”, “above”, “upper”, etc. are used as a single term as shown in the drawing. It can be used to easily describe the correlation between a component and other components. Spatially relative terms should be understood as terms that include different directions of components during use or operation in addition to the directions shown in the drawings. For example, if a component shown in a drawing is flipped over, a component described as "below" or "beneath" another component will be placed "above" the other component. You can. Accordingly, the illustrative term “down” may include both downward and upward directions. Components can also be oriented in other directions, so spatially relative terms can be interpreted according to orientation.

In this specification, a computer refers to all types of hardware devices including at least one processor, and depending on the embodiment, it may be understood as encompassing software configurations that operate on the hardware device. For example, a computer can be understood to include, but is not limited to, a smartphone, tablet PC, desktop, laptop, and user clients and applications running on each device.

Hereinafter, embodiments of the present invention will be described in detail with reference to the attached drawings.

Each step described in this specification is described as being performed by a computer, but the subject of each step is not limited thereto, and depending on the embodiment, at least part of each step may be performed in a different device.

Figure 1 is a diagram illustrating an artificial intelligence-based biometric data quality evaluation system according to an embodiment of the present invention.

Referring to FIG. 1, the artificial intelligence-based biometric data quality evaluation system according to an embodiment of the present invention may include a computing device 100, a user terminal 200, an external server 300, and a network 400. there is.

Here, the artificial intelligence-based biometric data quality evaluation system shown in FIG. 1 is according to one embodiment, and its components are not limited to the embodiment shown in FIG. 1, and can be added, changed, or deleted as necessary. You can.

In one embodiment, the computing device 100 may perform an artificial intelligence-based biometric data quality evaluation method. For example, the computing device 100 may evaluate the quality of a patient's biometric data to secure quality-verified biometric data for the purpose of performing an accurate diagnosis of the patient.

Here, evaluating the quality of biometric data is for the purpose of performing an accurate diagnosis of the patient. In order to secure quality-verified biometric data, the patient's biometric data is analyzed to determine whether noise is detected from the biometric data. It can mean judging.

In various embodiments, the computing device 100 analyzes biometric data collected in real time from a patient to detect noise in the biometric data collected in real time, but when noise is not detected from the biometric data for a predetermined period of time, noise is detected. Normal biometric data can be secured by extracting biometric data corresponding to the section that is not working. For example, when no noise is detected from the biometric data within a first time point to a second time point (e.g., a time point after a predetermined period from the first time point), the computing device 100 detects the biometric data from the first time point to the second time point. By extracting the section, normal biometric data without noise can be secured.

In various embodiments, the computing device 100 may be connected to the user terminal 200 through the network 400, and may display a user interface (UI) that outputs quality results for biometric data of a specific patient (e.g., 17 and 18) can be provided.

Here, the user terminal 200 may refer to any type of entity(s) in the system that has a mechanism for communication with the computing device 100. For example, such user terminal 200 includes a personal computer (PC), a notebook (note book), a mobile terminal, a smart phone, a tablet PC, and a wearable device. etc., and may include all types of terminals that can access wired/wireless networks. Additionally, the user terminal 200 may include an arbitrary server implemented by at least one of an agent, an application programming interface (API), and a plug-in. Additionally, the user terminal 200 may include an application source and/or client application.

Also, here, the network 400 may mean a connection structure that allows information exchange between nodes, such as a plurality of terminals and servers. For example, the network 400 includes a local area network (LAN), a wide area network (WAN), the World Wide Web (WWW), a wired and wireless data communication network, a telephone network, and a wired and wireless television communication network. can do.

In addition, here, the wireless data communication network includes 3G, 4G, 5G, 3GPP (3rd Generation Partnership Project), 5GPP (5th Generation Partnership Project), LTE (Long Term Evolution), WIMAX (World Interoperability for Microwave Access), and Wi-Fi (Wi-Fi). Fi), Internet, LAN (Local Area Network), Wireless LAN (Wireless Local Area Network), WAN (Wide Area Network), PAN (Personal Area Network), RF (Radio Frequency), Bluetooth network, It may include, but is not limited to, a Near-Field Communication (NFC) network, a satellite broadcasting network, an analog broadcasting network, and a Digital Multimedia Broadcasting (DMB) network.

In one embodiment, the external server 300 may be connected to the computing device 100 through the network 400, and may provide various information and information necessary for the computing device 100 to perform an artificial intelligence-based biometric data quality evaluation method. You can store and manage data, or collect, store, and manage various information and data generated by performing an artificial intelligence-based biometric data quality evaluation method. For example, the external server 300 may be a storage server separately provided outside the computing device 100, but is not limited thereto. Hereinafter, with reference to FIG. 2, the hardware configuration of the computing device 100 that performs an artificial intelligence-based biometric data quality evaluation method will be described.

Figure 2 is a hardware configuration diagram of a computing device that performs an artificial intelligence-based biometric data quality evaluation method according to another embodiment of the present invention.

Referring to FIG. 2, in various embodiments, the computing device 100 includes one or more processors 110, a memory 120 that loads a computer program 151 performed by the processor 110, and a bus ( 130), a communication interface 140, and a storage 150 for storing a computer program 151. Here, only components related to the embodiment of the present invention are shown in Figure 2. Accordingly, anyone skilled in the art to which the present invention pertains will know that other general-purpose components may be included in addition to the components shown in FIG. 2.

The processor 110 controls the overall operation of each component of the computing device 100. The processor 110 includes a Central Processing Unit (CPU), Micro Processor Unit (MPU), Micro Controller Unit (MCU), Graphic Processing Unit (GPU), or any other type of processor well known in the art of the present invention. It can be.

Additionally, the processor 110 may perform operations on at least one application or program for executing methods according to embodiments of the present invention, and the computing device 100 may include one or more processors.

In various embodiments, the processor 110 includes random access memory (RAM) (not shown) and read memory (ROM) that temporarily and/or permanently store signals (or data) processed within the processor 110. -Only Memory, not shown) may be further included. Additionally, the processor 110 may be implemented in the form of a system on chip (SoC) that includes at least one of a graphics processing unit, RAM, and ROM.

Memory 120 stores various data, commands and/or information. Memory 120 may load a computer program 151 from storage 150 to execute methods/operations according to various embodiments of the present invention. When the computer program 151 is loaded into the memory 120, the processor 110 can perform the method/operation by executing one or more instructions constituting the computer program 151. The memory 120 may be implemented as a volatile memory such as RAM, but the technical scope of the present disclosure is not limited thereto.

Bus 130 provides communication functionality between components of computing device 100. The bus 130 may be implemented as various types of buses, such as an address bus, a data bus, and a control bus.

The communication interface 140 supports wired and wireless Internet communication of the computing device 100. Additionally, the communication interface 140 may support various communication methods other than Internet communication. To this end, the communication interface 140 may be configured to include a communication module well known in the technical field of the present invention. In some embodiments, communication interface 140 may be omitted.

Storage 150 may store the computer program 151 non-temporarily. When performing an artificial intelligence-based biometric data quality evaluation process through the computing device 100, the storage 150 can store various information necessary to provide an artificial intelligence-based biometric data quality evaluation process.

The storage 150 is a non-volatile memory such as Read Only Memory (ROM), Erasable Programmable ROM (EPROM), Electrically Erasable Programmable ROM (EEPROM), flash memory, a hard disk, a removable disk, or a device well known in the art to which the present invention pertains. It may be configured to include any known type of computer-readable recording medium.

The computer program 151, when loaded into the memory 120, may include one or more instructions that cause the processor 110 to perform methods/operations according to various embodiments of the present invention. That is, the processor 110 can perform the method/operation according to various embodiments of the present invention by executing the one or more instructions.

In one embodiment, the computer program 151 evaluates the quality of biometric data based on artificial intelligence, including acquiring biometric data of a subject and evaluating the quality of the acquired biometric data using a previously learned artificial intelligence model. It may contain one or more instructions to perform a method.

The steps of the method or algorithm described in connection with embodiments of the present invention may be implemented directly in hardware, implemented as a software module executed by hardware, or a combination thereof. The software module may be RAM (Random Access Memory), ROM (Read Only Memory), EPROM (Erasable Programmable ROM), EEPROM (Electrically Erasable Programmable ROM), Flash Memory, hard disk, removable disk, CD-ROM, or It may reside on any type of computer-readable recording medium well known in the art to which the present invention pertains.

The components of the present invention may be implemented as a program (or application) and stored in a medium in order to be executed in conjunction with a hardware computer. Components of the invention may be implemented as software programming or software elements, and similarly, embodiments may include various algorithms implemented as combinations of data structures, processes, routines or other programming constructs, such as C, C++, , may be implemented in a programming or scripting language such as Java, assembler, etc. Functional aspects may be implemented as algorithms running on one or more processors. Hereinafter, with reference to FIGS. 3 to 17 , the artificial intelligence-based biometric data quality evaluation method performed by the computing device 100 will be described in more detail.

Figure 3 is a flow chart of an artificial intelligence-based biometric data quality evaluation method according to another embodiment of the present invention.

Referring to FIG. 3, in step S110, the computing device 100 may acquire biometric data of the target. Here, the subject's biological data may be sinusoidal photoplethysmography (PPG) data collected from the subject through a photoplethysmography device, but is not limited thereto.

In various embodiments, the computing device 100 may be connected to a photovoltaic pulse wave measuring device and may acquire the patient's PPG data measured in real time through the photovoltaic pulse wave measuring device, but is not limited to this, and the computing device ( 100) can upload at least one PPG data whose quality is to be evaluated among a plurality of PPG data previously measured and stored through a photoelectric pulse wave measuring device.

In step S120, the computing device 100 may preprocess the patient's biometric data obtained through step S110. Hereinafter, it will be described in more detail with reference to FIGS. 4 to 10.

FIG. 4 is a flowchart illustrating a method of preprocessing biometric data using the amount of change in biometric data values, in various embodiments.

In various embodiments, when the patient's biometric data includes a plurality of biometric data values at each of a plurality of consecutive time points, the computing device 100 is configured to By calculating the differential of biometric data values at each viewpoint, the amount of change in biometric data values for each of a plurality of viewpoints can be calculated.

For example, the computing device 100 can calculate the amount of change in biometric data for a specific point in time by calculating the difference between the biometric data value at a specific point in time and the biometric data value at a point in time before the specific point in time. By calculating the difference between the data value and the biometric data value at a point in time after a specific point in time, the amount of change in biometric data for a point in time after a specific point in time can be calculated.

In step S220, the computing device 100 may preprocess the patient's biometric data based on the amount of change in biometric data value calculated through step S210.

In various embodiments, the computing device 100 identifies a time point at which the difference in biometric data values between adjacent time points among the plurality of time points is greater than or equal to a threshold value, based on the amount of change in biometric data for each of the plurality of time points, and identifies the identified time point. Biometric data values at the point in time can be corrected.

More specifically, first, the computing device 100 identifies a time point where the difference in biometric data value between adjacent time points among a plurality of time points is greater than or equal to a threshold value, thereby preventing the biometric data value from rapidly rising due to noise or You can identify when the value drops sharply. For example, the computing device 100 may determine that the difference between the first biometric data value at the first time point and the second biometric data value at the second time point before the first time point is greater than or equal to a threshold value, and the first biometric data value at the first time point is If the difference between the data value and the third biometric data value at the third time point after the first time point is also greater than or equal to the threshold, the first time point can be identified as a time point when the data value suddenly rises or falls rapidly due to noise. However, it is not limited to this.

Thereafter, the computing device 100 may initialize the biometric data value at the point in time when it suddenly rises or falls suddenly due to noise. For example, as described above, when the first time point is identified as a time point when the data suddenly rises or falls rapidly due to noise, the computing device 100 may initialize the first biometric data value at the first time point.

Here, initializing the first biometric data value may mean changing the first biometric data value to a value representing a numerical result that cannot be expressed (eg, a NaN (Not a Number) value), but is not limited to this.

Thereafter, when the biometric data value at a specific point in time is initialized, the computing device 100 may correct the biometric data value at a specific point in time using biometric data values before and after the specific point in time. For example, as described above, when the first biometric data value at the first time point is initialized, the computing device 100 may configure the second biometric data value at a second time point before the first time point and the first biometric data value after the first time point. The first biometric data value can be corrected by interpolating the third biometric data value at three points in time (e.g., linear interpolation, polynomial interpolation, spline interpolation, etc.).

FIG. 5 is a flowchart illustrating a method of preprocessing biometric data in which a plurality of units of biometric data are combined in various embodiments.

Referring to FIG. 5 , in step S310, the computing device 100 may generate one biometric data by combining a plurality of units of biometric data.

Here, the plurality of unit biometric data may refer to biometric data collected from the subject for different periods of time.

Since the EEG is usually different when the eyes are open (eye-open) and the EEG when the eyes are closed (eye-close), the EEG measurement protocol is divided into a session to measure EEG in the eyes-open state and a session to measure the EEG in the eye-closed state. When measuring brain waves, each session is changed and the brain waves when the eyes are opened and the brain waves when the eyes are closed are measured separately.

At this time, when EEG and PPG data are measured together from the subject, the PPG data measurement is also stopped while changing the session for EEG measurement, and as a result, one PPG data is not measured and two or more unit PPG data are generated. do.

In order to analyze the PPG data for the target and diagnose the target, it is necessary to connect two or more unit PPG data to generate one PPG data, and the computing device 100 combines the two or more unit PPG data in chronological order. Thus, one PPG data can be generated.

For example, the plurality of unit biometric data includes first unit biometric data measured over a predetermined period from a first point in time and a predetermined period of time from a second point in time after the measurement of the second unit biometric data ends. It may include second unit biometric data, and the computing device 100 may connect the first biometric unit biometric data and the second unit biometric data to generate one biometric data.

In step S320, the computing device 100 may identify a point in time when correction is needed from one piece of biometric data generated through step S310.

When two or more unit PPG data are connected to generate one PPG data, since the two or more different unit PPG data are measured at different times, as shown in FIG. 6, the two or more different unit PPG data are connected. There is a problem that parts may show drastic differences, and if they are simply connected, PPG data containing noise may be generated.

In consideration of this, the computing device 100 may combine a plurality of unit biometric data to identify a point in time when correction is needed, that is, a point in time when unit biometric data with a sharp difference is connected from the connected biometric data.

In various embodiments, the computing device 100 may identify a point in time at which biometric data having a difference greater than a threshold value, among a plurality of points in time at which two or more units of biometric data are linked, as a point in time requiring correction. For example, the computing device 100 is connected to the first unit biometric data and the first unit biometric data measured for a predetermined period from the first point in time, and the second unit biometric data is connected to the first unit biometric data a predetermined time after the measurement of the first unit biometric data ends. For the second unit biometric data measured over a predetermined period from the point in time, the difference between the last biometric data value of the first unit biometric data and the first biometric data value of the second unit biometric data is calculated, and the calculated difference is the threshold value. In the above case, the time when the first unit biometric data and the second unit biometric data are connected can be identified as the time when correction is required.

In various embodiments, the computing device 100 calculates the difference between the biometric data at each of the plurality of time points and the biometric data value at the previous time point before each of the plurality of time points, with respect to one piece of biometric data in which a plurality of units of biometric data are connected. , the point in time when the calculated difference is above the threshold can be identified as the point in time when correction is necessary.

In step S330, the computing device 100 may preprocess biometric data values at the identified time point through step S320.

In various embodiments, the computing device 100 determines the first biometric data value of the second unit of biometric data when the difference between the last biometric data value of the first unit of biometric data and the first biometric data value of the second unit of biometric data is greater than or equal to a threshold value. may be corrected, and the first biometric data of the corrected second unit biometric data and the first biometric data value included in the second biometric data may be interpolated.

For example, as shown in FIG. 6, the computing device 100 corrects the first biometric data value of the second unit biometric data to the same value as the last biometric data value of the first unit biometric data, and uses an interpolation method (e.g., linear interpolation). , polynomial interpolation, spline interpolation, etc.), by interpolating the last biometric data value of the first unit biometric data and the first biometric data value of the corrected second unit biometric data, the first unit biometric data and the second unit biometric data are smoothed. It can be connected in a curved shape.

Here, it is explained that the computing device 200 corrects the first biometric data value of the second unit of biometric data to the same value as the last biometric data value of the first unit of biometric data, but the present invention is not limited to this, and in some cases, the first biometric data value is corrected to the same value as the last biometric data value of the first unit of biometric data. The last biometric data value of the unit biometric data may be corrected to the same value as the first biometric data value of the second unit biometric data.

FIG. 7 is a flowchart illustrating a method of preprocessing biometric data using peak saliency in various embodiments.

Referring to FIG. 7 , in step S410, the computing device 100 may generate a peak saliency histogram using the patient's biometric data. For example, when the biometric data collected from the patient is PPG data in the form of a sinusoidal wave, the computing device 100 may identify a plurality of peaks included in the PPG data by analyzing the PPG data in the form of a sinusoidal wave, and may identify a plurality of peaks included in the PPG data. Multiple peak saliencies can be calculated for each, and a peak saliency histogram can be generated using the calculated multiple peak saliencies.

Here, peak prominence may mean the difference in height between the high and low points of the peak. For example, referring to Figure 8, in the section including the first peak and the second peak, the peak prominence of the second peak is the height value of the second peak and the section including the first peak and the second peak. It may mean the difference between the height values of the low points.

In addition, here, as shown in FIG. 10, the peak saliency histogram sets a plurality of peak saliency sections according to a preset peak saliency size and counts the number of peak saliency sections for each of the multiple peak saliency sections. It may be generated accordingly, but is not limited to this.

In step S420, the computing device 100 may set a reference peak saliency using the peak saliency histogram generated through step S410.

In various embodiments, the computing device 100 may analyze the peak saliency histogram to identify a peak saliency section having a predetermined pattern, and set the peak saliency corresponding to the identified peak saliency section as the reference peak saliency. there is.

In order to diagnose the patient's condition based on sinusoidal PPG data, points corresponding to peaks in the PPG data are identified, and properties for the identified points (e.g., peak value, distance between peaks (RR peak), etc. ) is necessary. Typically, the point corresponding to the peak is identified based on the difference or amount of change from adjacent values, so as shown in FIG. 9, not only the point corresponding to the peak but also the valley Corresponding points can also be identified together.

In the case of points corresponding to valleys, since they have lower peak saliency compared to points corresponding to peaks, independent counting occurs in the low peak saliency section on the peak histogram, as shown in Figure 10, and low peak saliency It can be seen that after independent counting occurs in the degree section, there is a pattern in which the main distribution (peak saliency distribution of points corresponding to the peak) occurs in the peak saliency section spaced a certain distance apart.

Considering this, the computing device 100 may identify a section having a pattern in which points corresponding to the valley are visible as described above and set the reference peak saliency based on this. For example, the computing device 100 has a counting value for the first peak saliency section other than 0 for three adjacent peak saliency sections, and counts the second peak saliency section and the third peak saliency section. If the value is 0, the peak saliency corresponding to the third peak saliency section can be set as the reference peak saliency.

In step S430, the computing device 100 may preprocess the biometric data using the reference peak saliency set through step S420. For example, the computing device 100 may filter the points corresponding to the valley by identifying points corresponding to the peak from the patient's biometric data and deleting points whose peak saliency is less than the reference peak saliency among the identified points. there is.

Referring again to FIG. 3 , in step S130, the computing device 100 may evaluate the quality of the patient's biometric data by analyzing the biometric data preprocessed through step S120.

In various embodiments, the computing device 100 may detect noise in the pre-processed biometric data by analyzing the pre-processed biometric data using a previously learned artificial intelligence model, and determine whether noise is detected from the pre-processed biometric data. Accordingly, the quality of biometric data can be evaluated.

Here, the previously learned artificial intelligence model may be a model that reduces the dimension of biometric data and evaluates the quality of the biometric data depending on whether noise is detected from the reduced-dimensional biometric data.

For example, a previously learned artificial intelligence model may include an auto encoder 10, as shown in FIG. 11.

Here, the auto-encoder 10 may be an artificial neural network for outputting output data that is the same or similar to input data.

The auto-encoder 10 may include an encoder including one or more dimension reduction layers as a hidden layer and a decoder including one or more dimension restoration layers as a hidden layer.

At this time, a hidden layer including one or more dimension reduction layers and one or more dimension restoration layers is placed between the input layer and the output layer, but an odd number of hidden layers may be placed between the input layer and the output layer.

When the encoder includes multiple dimension reduction layers, the number of nodes included in each of the multiple dimension reduction layers gradually decreases from the input layer toward an intermediate layer called the bottleneck layer (coding layer) (e.g., moving away from the input layer). It can be implemented in a structure where the number of nodes decreases as the number increases.

In addition, when the decoder includes a plurality of dimensional restoration layers, the number of nodes included in each of the plurality of dimensional restoration layers gradually increases from the bottleneck layer to the output layer (e.g., the number of nodes increases as it gets closer to the output layer). structure) can be implemented.

That is, the number of nodes in the plurality of dimension reduction layers included in the encoder gradually decreases in the direction of the bottleneck layer, and the number of nodes in the plurality of dimension restoration layers included in the decoder gradually increases in the output layer direction, Accordingly, the plurality of dimension reduction layers included in the encoder and the plurality of dimension restoration layers included in the decoder can be implemented in a symmetrical form with respect to the bottleneck layer. However, it is not limited to this.

Here, for the purpose of generating two-dimensional biometric data by reducing the dimension of high-dimensional biometric data, the bottleneck layer of the autoencoder 10 shown in FIG. 11 may include two nodes, but this is only an example. , the number of nodes included in the bottleneck layer is not limited to this.

In various embodiments, the autoencoder 10 may be a model learned based on unsupervised learning using a plurality of biometric data without noise (i.e., normal biometric data) as learning data.

Here, the learning of the auto encoder 10 is to minimize output errors. In learning of the auto encoder (10), learning data is repeatedly input into the auto encoder (10), the output of the auto encoder (10) and the error of the target for the learning data are calculated, and the auto encoder (10) is used in a direction to reduce the error. ) is a process of backpropagating the error from the output layer of the autoencoder 10 to the input layer to update the weight of each node of the autoencoder 10.

Here, unsupervised learning may refer to a method of learning using learning data in which the correct answer is not labeled. For example, in the case of supervised learning, learning data labeled with the correct answer is input to an artificial intelligence model, and an error can be calculated by comparing the output of the artificial intelligence model and the label of the learning data. On the other hand, in the case of unsupervised learning, learning data that is not labeled with the correct answer is input to the artificial intelligence model, and the error can be calculated by comparing the output of the artificial intelligence model and the input learning data.

The calculated error is back-propagated in the artificial intelligence model in the reverse direction (i.e., from the output layer to the input layer), and the connection weight of each node in each layer of the artificial intelligence model can be updated according to back-propagation. The amount of change in the connection weight of each updated node may be determined according to the learning rate. The artificial intelligence model's calculation of input data and backpropagation of errors can constitute a learning cycle (epoch). The learning rate can be applied differently depending on the number of repetitions of the learning cycle of the artificial intelligence model. For example, in the early stages of training an artificial intelligence model, a high learning rate can be used to ensure that the artificial intelligence model quickly achieves a certain level of performance to increase efficiency, and in the later stages of training, a low learning rate can be used to increase accuracy.

As another example, the previously learned artificial intelligence model may include a one-dimensional convolutional neural network model 20 and an auto encoder 10, as shown in FIG. 12, and the one-dimensional convolutional neural network model 20 and the auto encoder The input layers of (10) may have a combined structure.

In other words, as the subject's biometric data is input into the one-dimensional convolutional neural network model (20), features corresponding to the time series pattern of the biometric data can be extracted, and the features and biometric data corresponding to the time series pattern are combined with the auto encoder (10). By inputting, low-dimensional biometric data can be generated by considering the time-series pattern of biometric data through the auto-encoder 10.

In various embodiments, the computing device 100 may detect noise included in the biometric data by analyzing the biometric data using a previously learned artificial intelligence model, and determine the size of the biometric data depending on whether noise is detected from the biometric data. Quality can be evaluated.

More specifically, referring to FIG. 13, first, the computing device 100 may generate low-dimensional biometric data by inputting the target's biometric data (preprocessed biometric data) into the encoder of the auto encoder 10. Here, the low-dimensional biometric data may be two-dimensional biometric data, but is not limited thereto.

Here, the computing device 100 may input the subject's biometric data into the auto encoder 10, but is not limited to this, and may window the subject's biometric data in predetermined time units (e.g., in units of 4 seconds). By performing , only biometric data corresponding to a predetermined time interval can be sequentially input into the encoder of the auto encoder 10 according to time series. That is, the computing device 100 can generate a plurality of biometric data corresponding to a predetermined time interval by dividing the biometric data into certain time sections according to the time series, and each of the generated plurality of biometric data is used as input data. It can be sequentially input to the encoder of the auto encoder (10).

At this time, the biometric data included in the input data sequentially input to the encoder of the auto encoder 10 may be implemented in a form where at least part of the biometric data is overlapped. For example, when the predetermined time unit is 4 seconds, the computing device 100 may input biometric data corresponding to a period of 0 to 4 seconds with respect to the biometric data into the encoder of the auto encoder 10, and then Biometric data with an overlapping 50% section, that is, biometric data corresponding to a 2 to 6 second section, can be input to the encoder of the auto encoder 10 as input data.

Thereafter, the computing device 100 may cluster a plurality of biometric data values included in the low-dimensional biometric data. Here, various techniques are known for methods of creating one or more clusters by clustering a plurality of data values based on a clustering algorithm, and these known techniques can be selectively applied. In this specification, a plurality of biometric data It does not describe specific methods for clustering values.

Thereafter, the computing device 100 may determine whether the biometric data contains noise based on the result of clustering the plurality of biometric data values.

For example, when one cluster is created by clustering a plurality of biometric data values, the computing device 100 may determine that the biometric data does not contain noise, and by clustering the plurality of biometric data values, the computing device 100 may determine that the biometric data does not contain noise. If the above clusters are generated, it may be determined that the biometric data contains noise.

As another example, when one cluster is created by clustering a plurality of biometric data values, the computing device 100 may determine that the biometric data does not contain noise, and by clustering the plurality of biometric data values, the computing device 100 may determine that one cluster is generated by clustering the plurality of biometric data values. If a large number of biometric data values that are not clustered with the clusters show a distributed pattern in various areas, it can be determined that the biometric data contains noise.

When training the autoencoder 10 using normal biometric data (e.g., biometric data without noise) as learning data, the output data (Latent vector) output from the bottleneck layer of the autoencoder is compressed for the normal biometric data. When biometric data containing noise is input to the auto encoder 10, another cluster (corresponding to noise) is distinguished from the cluster corresponding to normal biometric data, as shown in FIG. 13. Clusters) may be created, or normal biometric data values may be clustered into one cluster, but values corresponding to noise may not be clustered into one cluster but may be distributed in various areas.

That is, when one cluster is generated as low-dimensional biometric data is analyzed through the auto encoder 10, the computing device 100 determines that only the cluster corresponding to normal biometric data has been created, so that the biometric data is free of noise. It can be judged as not included.

Meanwhile, as the computing device 100 analyzes low-dimensional biometric data through the auto encoder 10, two or more clusters are created, or a plurality of biometric data values that are not clustered with one cluster are distributed in various areas. If a pattern is visible, it can be determined that the biometric data contains noise. Hereinafter, as another embodiment of evaluating the quality of biometric data, a quality evaluation method based on peak saliency and a quality evaluation method based on divergence histogram will be described.

The peak saliency-based quality evaluation method and the divergence histogram-based quality evaluation method, which will be described below, are performed independently from the previously learned artificial intelligence model-based quality evaluation method (e.g., Figure 3), so they are used in biometric data. It can be implemented in a form that derives individual quality evaluation results for each product, and in some cases, it can be implemented in a form that verifies or supplements the results derived according to a quality evaluation method based on a pre-learned artificial intelligence model.

For example, the computing device 100 compares the results derived from a quality evaluation method based on a previously learned artificial intelligence model with the results derived from a quality evaluation method based on peak saliency and a quality evaluation method based on divergence histogram, respectively. The results derived from the quality evaluation method based on the learned artificial intelligence model can be verified.

As another example, the computing device 100 may provide a first quality evaluation result derived from a quality evaluation method based on a previously learned artificial intelligence model (e.g., Type 1 in FIG. 17) and a second quality evaluation result derived from a quality evaluation method based on peak saliency. The quality evaluation results (e.g., Type 2 in FIG. 17) and the third quality evaluation results (e.g., Type 3 in FIG. 17) derived from the divergence histogram-based quality evaluation method can be collected to form one quality evaluation result for biometric data. You can.

In various embodiments, the computing device 100 may evaluate the quality of biometric data according to a peak saliency-based quality evaluation method.

More specifically, the computing device 100 may detect noise included in the biometric data based on peak saliency corresponding to each of a plurality of peaks included in the subject's biometric data.

For example, the computing device 100 may generate a plurality of biometric data sections by dividing the biometric data into preset time units, calculate peak saliency corresponding to each of the plurality of biometric data sections, and calculate the calculated peak saliency. By comparing a plurality of peak saliencies and a threshold value, it can be determined whether the biometric data contains noise.

Typically, in order to measure biometric data about a subject, the subject and the biometric data measuring device are physically connected. However, if the biometric data measuring device malfunctions or, in some cases, the subject and the biometric data measuring device are not properly connected, the biometric data may not be collected properly, and in cases where biometric data is not properly collected, there are cases where data other than biometric data is measured.

In the case of data other than biometric data, since it typically has the form of a signal without pulses, a very small peak saliency can be calculated as shown in FIG. 14.

That is, if there is a biometric data section whose peak saliency is less than the threshold among the plurality of biometric data sections divided from the biometric data, the computing device 100 determines that the corresponding biometric data section contains noise, that is, the corresponding biometric data section. It can be determined that data other than biometric data (signal without pulse) was collected in the section.

In various embodiments, when it is determined that noise is included in any one of the plurality of biometric data sections divided from biometric data, the computing device 100 specifies a noise section within any one of the biometric data sections. can do. For example, when it is determined that noise is included in one biometric data section, the computing device 100 divides one biometric data section into a plurality of unit sections and calculates the peak saliency for each of the plurality of unit sections. It can be calculated, and among a plurality of unit sections, at least one unit section whose peak saliency is less than the threshold can be specified as the noise section.

That is, the computing device 100 not only determines whether noise is included in a specific biometric data section, but also specifies the section containing noise within the specific biometric data section, thereby removing the section specified as the noise section later. It is possible to create an environment in which normal biometric data without noise can be obtained more quickly and accurately.

In various embodiments, the computing device 100 may evaluate the quality of biometric data according to a divergence histogram-based quality evaluation method.

More specifically, the computing device 100 may detect noise included in the biometric data using a divergence histogram corresponding to the biometric data.

For example, the computing device 100 generates a divergence histogram corresponding to biometric data, analyzes the divergence histogram, and determines whether there is a divergence section with a predetermined pattern among a plurality of divergence sections included in the divergence histogram, thereby It can be determined whether noise is included.

Typically, looking at the divergence histogram corresponding to the biometric data shown in FIG. 15, if the biometric data value at a specific time point included in the biometric data has a suddenly high value due to noise, the divergence histogram is far from the main distribution. Since counting occurs independently of the section, the divergence histogram can be analyzed to identify a divergence section that has a pattern in which the counting value is not 0 and the counting value corresponding to the adjacent divergence section is 0 among the plurality of divergence sections included in the divergence histogram. When a divergence section having this pattern is identified, it may be determined that the section of biometric data corresponding to the identified divergence section contains noise.

In various embodiments, the computing device 100 may evaluate the quality of biometric data depending on whether noise is detected from the biometric data, and thereby derive a quality evaluation result.

Here, the quality evaluation result (Signal Quality Index, SQI) is the result of determining whether noise is detected for a specific section (e.g., a windowed predetermined time section) of biometric data. For example, the quality evaluation result is a result of determining whether noise is detected. 1, if no noise is detected, it can be output as 0.

In this way, the quality evaluation results derived for the entire biometric data section are graphs in the form of pulse waves that reflect whether or not noise is included for each biometric data section, as shown in (C) of FIGS. 17 and 18. It can be visualized.

In various embodiments, the computing device 100 analyzes the biometric data and, if a specific noise is detected, determines that the biometric data contains a specific noise, but if the period during which the specific noise is detected is less than a preset period, the biometric data It can be determined that no specific noise is included. For example, as shown in FIG. 16 , the computing device 100 may determine that the biometric data does not contain the specific noise when a specific noise is detected at a specific point in time or in a very short section while analyzing the biometric data.

In other words, when the computing device 100 detects all noise included in the biometric data, it is possible to build an environment in which biometric data free of noise can be secured. However, in reality, it is difficult to secure biometric data free of noise. , Even if it is possible, it may take a very long time, so noise that does not cause a major problem when diagnosing a patient (e.g., noise that occurs for an extremely short period of time) is ignored and has the possibility of causing problems when diagnosing a patient. Only noise that exists (for example, noise that is included over a certain period of time) can be detected.

For example, the computing device 100 performs windowing on the subject's biometric data in predetermined time units (for example, in units of 4 seconds) and detects noise in the biometric data. However, when temporary noise is detected, the temporary noise is Detected parts can be removed using a preceding window.

The artificial intelligence-based biometric data quality evaluation method for securing biometric data with verified quality described above was explained with reference to the flow chart shown in the drawing. For simple explanation, the artificial intelligence-based biometric data quality evaluation method for securing quality-verified biometric data is illustrated as a series of blocks, but the present invention is not limited to the order of the blocks, and some blocks are included in the specification. It may be performed in a different order than that shown and performed in, or may be performed simultaneously. Additionally, new blocks not described in the specification and drawings may be added, or some blocks may be deleted or changed.

Above, embodiments of the present invention have been described with reference to the attached drawings, but those skilled in the art will understand that the present invention can be implemented in other specific forms without changing its technical idea or essential features. You will be able to understand it. Therefore, the embodiments described above should be understood in all respects as illustrative and not restrictive.

Claims (14)

1. In an artificial intelligence-based biometric data quality evaluation method performed by a computing device,

   Obtaining biometric data of a subject; and

   Including evaluating the quality of the acquired biometric data using a previously learned artificial intelligence model,

   Artificial intelligence-based biometric data quality evaluation method.
2. According to paragraph 1,

   The previously learned artificial intelligence model is,

   A model that reduces the dimension of the acquired biometric data and evaluates the quality of the acquired biometric data depending on whether noise is detected from the dimensionally reduced biometric data,

   Artificial intelligence-based biometric data quality evaluation method.
3. According to paragraph 1,

   The previously learned artificial intelligence model is,

   An auto-encoder comprising an encoder including one or more dimension reduction layers and a decoder including one or more dimension restoration layers, and learned based on unsupervised learning using a plurality of biometric data without noise as learning data. ), including

   The step of evaluating the quality of the acquired biometric data is,

   Preprocessing the obtained biometric data;

   generating low-dimensional biometric data by inputting the pre-processed biometric data into the encoder; and

   Comprising the step of detecting noise by analyzing the generated low-dimensional biometric data,

   Artificial intelligence-based biometric data quality evaluation method.
4. According to paragraph 3,

   The step of preprocessing the obtained biometric data is,

   The difference between the first biometric data value at a first time point included in the acquired biometric data and the second biometric data value at a second time point before the first time point and the first biometric data value and after the first time point If the difference between the third biometric data values at the third point in time is greater than or equal to a threshold value, initializing the first biometric data values; and

   Comprising the step of interpolating the initialized first biometric data value using the second biometric data value and the third biometric data value,

   Artificial intelligence-based biometric data quality evaluation method.
5. According to paragraph 3,

   The obtained biometric data is,

   First unit biometric data measured for a predetermined period from a first point in time and connected to the first unit biometric data, and a predetermined period starting from a second point in time after a predetermined time has elapsed after the measurement of the first unit biometric data ends. Contains second unit biometric data measured during the

   The step of preprocessing the obtained biometric data is,

   calculating a difference between a last biometric data value of the first unit of biometric data and a first biometric data value of the second unit of biometric data; and

   If the calculated difference is greater than or equal to the threshold value, the first biometric data value of the second unit biometric data is corrected based on the last biometric data value of the first unit biometric data, and the first biometric data value of the corrected second unit biometric data is corrected. Comprising the step of interpolating biometric data and a first biometric data value included in the second biometric data,

   Artificial intelligence-based biometric data quality evaluation method.
6. According to paragraph 3,

   The obtained biometric data is,

   It includes photoplethysmogram (PPG) data collected in real time from the subject,

   The step of preprocessing the obtained biometric data is,

   Identifying a plurality of peaks included in the photoplethysmographic pulse wave data;

   Calculating a plurality of peak prominences corresponding to each of the identified plurality of peaks, and generating a peak prominence histogram using the calculated plurality of peak prominences;

   Analyzing the generated peak saliency histogram to identify a peak saliency section having a predetermined pattern, and setting the peak saliency corresponding to the identified peak saliency section as a reference peak saliency; and

   Comprising the step of selecting a peak saliency lower than the set reference peak saliency among the calculated plurality of peak saliencies, and removing a peak corresponding to the selected peak saliency among the plurality of peaks.

   Artificial intelligence-based biometric data quality evaluation method.
7. According to paragraph 3,

   The step of detecting the noise is,

   Clustering a plurality of biometric data values included in the generated low-dimensional biometric data; and

   Comprising the step of determining whether the acquired biometric data contains noise based on a result of clustering the plurality of biometric data values,

   Artificial intelligence-based biometric data quality evaluation method.
8. According to paragraph 1,

   generating a plurality of biometric data sections by dividing the acquired biometric data - the acquired biometric data includes photoplethysmographic data collected in real time from the subject - into preset time units;

   calculating a peak saliency corresponding to each of the plurality of generated biometric data sections using biometric data values included in each of the plurality of generated biometric data sections; and

   If the peak saliency calculated corresponding to one of the plurality of biometric data sections generated is less than a threshold value, determining that noise is included in the one biometric data section, further comprising:

   Artificial intelligence-based biometric data quality evaluation method.
9. According to clause 8,

   The step of determining that noise is included in any one biometric data section is:

   When it is determined that noise is included in any one of the biometric data sections, dividing the one biometric data section into a plurality of unit sections and calculating peak saliency for each of the plurality of divided unit sections. ; and

   An artificial intelligence-based biometric data quality evaluation method comprising: specifying at least one unit section in which the calculated peak saliency is less than a threshold value among the plurality of divided unit sections as a noise section.
10. According to paragraph 1,

    Generate a divergence histogram corresponding to the acquired biometric data, analyze the generated divergence histogram, and if any one divergence section with a predetermined pattern exists among a plurality of divergence sections included in the generated divergence histogram, Further comprising determining that the acquired biometric data contains noise,

    Artificial intelligence-based biometric data quality evaluation method.
11. According to paragraph 1,

    The previously learned artificial intelligence model is,

    An auto-encoder comprising an encoder including one or more dimension reduction layers and a decoder including one or more dimension restoration layers, and learned based on unsupervised learning using a plurality of biometric data without noise as learning data. ) and a one-dimensional convolutional neural network model,

    The step of evaluating the quality of the acquired biometric data is,

    extracting features related to a time series pattern of the acquired biometric data by inputting the acquired biometric data into the one-dimensional convolutional neural network model;

    generating low-dimensional biometric data by inputting the extracted features and the acquired biometric data into the encoder; and

    Comprising the step of detecting noise by analyzing the generated low-dimensional biometric data,

    Artificial intelligence-based biometric data quality evaluation method.
12. According to paragraph 1,

    The step of evaluating the quality of the acquired biometric data is,

    When specific noise is detected by analyzing the acquired biometric data, it is determined that the acquired biometric data contains the specific noise, but if the period during which the specific noise was detected is less than a preset period, the acquired biometric data Including determining that the specific noise is not included in

    Artificial intelligence-based biometric data quality evaluation method.
13. processor;

    network interface;

    Memory; and

    Includes a computer program loaded into the memory and executed by the processor,

    The computer program is,

    Instructions for acquiring biometric data of a subject; and

    Containing instructions for evaluating the quality of the acquired biometric data using a previously learned artificial intelligence model,

    A computing device that performs an artificial intelligence-based biometric data quality evaluation method.
14. Combined with a computing device,

    Obtaining biometric data of a subject; and

    A computer program stored in a recording medium readable by a computing device to execute an artificial intelligence-based biometric data quality evaluation method including the step of evaluating the quality of the acquired biometric data using a previously learned artificial intelligence model.

Images