KR20230062109A - Biosignal processing device and method - Google Patents

Abstract

A biosignal processing apparatus according to an embodiment includes a communication unit that receives a user's biosignal from a device that collects biosignals, and first machine learning for a target biosignal among a plurality of partial biosignals divided from the received biosignal. and a processor for identifying a noise source that causes a noise component included in the target physiological signal based on output data output by applying the model.

Description

Biosignal processing device and method {Biosignal processing device and method}

Hereinafter, technologies related to devices and methods for processing received bio-signals are disclosed.

Wearable devices for measuring bio-signals are developed in various forms such as watches, rings, bands, and patches. Individuals can measure and monitor biosignals through these wearable devices. For example, AliveCor and the like provided users with an opportunity to diagnose arrhythmia without visiting a hospital through an electrocardiogram measurement function, and also sold devices certified as medical devices through FDA approval. In particular, patch-based electrocardiogram measurement technology overcomes the limits of accuracy of the PPG sensor and enables continuous monitoring for a period of two weeks. am. In Korea, five companies have obtained approval from the Ministry of Food and Drug Safety for Holter electrocardiographs (class 2 medical devices), and the products of three companies are in the process of being incorporated into the domestic medical system, such as being listed in insurance. In the case of an electrocardiogram, an electrocardiogram measuring device can be of great significance in that it can replace the heavy and inconvenient Holter equipment connected with a long wire.

Sensors designed to collect vital signals are based on electromagnetic properties. Because of this, the biological signal collected from the sensor may include noise components due to external magnetic force, electromagnetic waves, temperature change, and the like. The amplitude of the biosignal including the noise component may be greater than the amplitude of the biosignal not including the noise component, and information about the biosignal collected in a corresponding section has a problem in that it is difficult to utilize. In the bio-signal analysis process performed after the bio-signals are collected, a method of removing sections corresponding to bio-signals including noise components through frequency analysis is used, but bio-signals including all noise components are classified. It is a very difficult thing to do. Above all, it can be a big problem that important information can be missed when it is difficult to analyze biosignals, which are objects of collection, due to noise. The lifestyle of modern people, who frequently use mobile phones and have various electronic devices around them, may be a factor that makes it difficult to collect biosignals.

A biosignal processing apparatus according to an embodiment includes a communication unit that receives a user's biosignal from a device that collects biosignals, and first machine learning for a target biosignal among a plurality of partial biosignals divided from the received biosignal. and a processor for identifying a noise source that causes a noise component included in the target physiological signal based on output data output by applying the model.

The processor may determine the noise source as one or a combination of two or more of electric field, magnetic field, motion, and temperature.

The biosignal processing apparatus according to an embodiment may further include a display displaying at least one or a combination of two or more of a text guidance message mapped with the noise source identified by the processor, a behavior correction guide, and a graphic object. .

The biosignal processing apparatus according to an embodiment may further include a speaker that reproduces one or a combination of two or more of a voice guide message mapped with the noise source identified by the processor, a behavior correction guide, and an alarm.

The biosignal processing device according to an embodiment may further include a lighting unit displaying a color mapped with the noise source identified by the processor.

The communication unit may further receive acceleration data of the user from an accelerometer together with the biosignal, and the processor may generate noise for a partial biosignal corresponding to the acceleration data based on the acceleration data together with the output data. The source can be determined by exercise.

The processor may select the target biosignal from among the plurality of partial biosignals based on probability scores output by applying a second machine learning model to each of the plurality of partial biosignals.

The processor is a provisional likelihood score generated by applying the second machine learning model to a training biosignal including a valid signal measured in a controlled clinical environment and a noise signal including a noise component obtained from a biosignal collection device. An objective function value between the ground truth and the ground truth may be calculated, and updating of parameters of the second machine learning model may be repeated so that the calculated objective function value converges.

The processor may calculate the objective function value further based on a composite signal generated by adding a noise component to the valid signal together with the valid signal and the noise signal.

The processor may train a new second machine learning model using a training partial bio-signal selected based on the reliability of the probability score output by the second machine learning model.

A biosignal processing method according to an embodiment includes the steps of receiving a biosignal of a user from a device that collects biosignals, and performing first machine learning on a target biosignal among a plurality of partial biosignals divided from the received biosignal. The method may further include identifying a noise source generated in the target physiological signal based on output data output by applying the model.

1 illustrates a biosignal processing device according to an exemplary embodiment.
2 illustrates identifying a noise source that causes a noise component included in a biosignal by applying a first machine learning model to the biosignal.
3 illustrates selecting a target bio-signal by applying a second machine learning model.
4 illustrates training of a second machine learning model according to one embodiment.
5 illustrates training a new second machine learning model based on selected training partial biosignals according to an embodiment.
6 shows a biosignal processing device that presents information about a noise source to a user.
7A and 7B show a biosignal processing device that presents information about a noise source and noise intensity to a user.
8 illustrates identifying a noise source using an accelerometer together with a biosignal collection device.

Specific structural or functional descriptions of the embodiments are disclosed for illustrative purposes only, and may be changed and implemented in various forms. Therefore, the form actually implemented is not limited only to the specific embodiments disclosed, and the scope of the present specification includes changes, equivalents, or substitutes included in the technical idea described in the embodiments.

Although terms such as first or second may be used to describe various components, such terms should only be construed for the purpose of distinguishing one component from another. For example, a first element may be termed a second element, and similarly, a second element may be termed a first element.

It should be understood that when an element is referred to as being “connected” to another element, it may be directly connected or connected to the other element, but other elements may exist in the middle.

Singular expressions include plural expressions unless the context clearly dictates otherwise. In this specification, terms such as "comprise" or "have" are intended to designate that the described feature, number, step, operation, component, part, or combination thereof exists, but one or more other features or numbers, It should be understood that the presence or addition of steps, operations, components, parts, or combinations thereof is not precluded.

Unless defined otherwise, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art. Terms such as those defined in commonly used dictionaries should be interpreted as having a meaning consistent with the meaning in the context of the related art, and unless explicitly defined in this specification, it should not be interpreted in an ideal or excessively formal meaning. don't

Hereinafter, embodiments will be described in detail with reference to the accompanying drawings. In the description with reference to the accompanying drawings, the same reference numerals are given to the same components regardless of reference numerals, and overlapping descriptions thereof will be omitted.

1 illustrates a biosignal processing device according to an exemplary embodiment.

The biosignal processing device 100 may include a processor 110, a communication unit 120, and an output unit 105.

The communication unit 120 may establish wired communication and/or wireless communication with an external device and receive data from the external device. For example, the communication unit 120 may receive one or more of a user's bio-signal and data corresponding to the bio-signal from a device that collects the bio-signal. A device that collects bio-signals may be a wearable device for collecting user's bio-signals. A wearable device may be, for example, a device worn or attached to a user's body in the form of a watch, ring, band, and/or patch.

Biological signals are biological signals generated from activities of living cells, and may include, for example, body temperature, pulse, blood pressure, respiration, blood sugar, brain waves, electrocardiogram, and oxygen saturation. The bio-signal collected by the bio-signal collection device includes not only a bio-signal component corresponding to the user's actual bio-signal, but also the environment around the bio-signal collection device (eg, electric field, magnetic field, temperature change, etc.) and the user's activity ( For example, noise components caused by exercise, etc.) may be included. An electrocardiogram signal received from a biosignal collection device affected by an electric field may be a signal in which an electrocardiogram component based on potential values of the heart and a noise component due to the electric field are combined. The noise component caused by the electric field may represent a difference between the ECG signal received from the biosignal collection device and the ECG signal component (eg, heart potential values of the user).

For reference, in the present specification, an example in which a biosignal is an electrocardiogram signal is mainly described. For example, an electrocardiogram (ECG) signal is data obtained by sampling potential values of the heart appearing in relation to a user's heartbeat at a sampling rate. can indicate For example, the sampling rate may be 300 times per minute.

The processor 110 may process at least one of biosignals and acceleration data received by the communication unit 120 . The processor 110 may identify a noise source that causes a noise component included in the biosignal by processing at least one of the received biosignal and acceleration data. The noise source is a surrounding environment or a user's activity that affects the bio-signal collection device, and may be, for example, an electric field, a magnetic field, a change in temperature, or motion. An operation of identifying a noise source by the processor will be described later with reference to FIGS. 2 to 5 .

When a noise source is identified by the processor 110, the output unit 105 may guide a user to a content element mapped with a corresponding noise source for each of a plurality of noise sources. The output unit 105 may include, for example, a display 130 , a lighting unit 140 , and a speaker 150 . Operations by the display 130, the lighting unit 140, and the speaker 150 presenting the content elements mapped with the noise source to the user will be described later with reference to FIGS. 6, 7A, and 7B.

2 illustrates identifying a noise source that causes a noise component included in a biosignal by applying a first machine learning model to the biosignal.

In step 210, the communication unit may receive the user's biosignal from the biosignal collection device. For example, when a bio-signal collection device is affected by a noise source while measuring a bio-signal, the bio-signal measured by the bio-signal collection device may include a noise component. The communication unit may receive a bio signal including a noise component.

In operation 220, the processor may apply the first machine learning model 221 to the target biosignal and identify a noise source of the biosignal based on the output data. The first machine learning model 221 is a model designed and trained to output output data from target biosignals, and may include, for example, a neural network. A schematic structure of a neural network will be described as follows. According to an embodiment, a neural network may include a plurality of layers composed of a plurality of nodes. Also, the neural network may include connection weights for connecting a plurality of nodes included in each of a plurality of layers to a node included in another layer. For example, a neural network may represent a recognition model that mimics the computational capability of a biological system by using a large number of nodes connected by edges. A neural network may include a plurality of layers. For example, a neural network may include an input layer, a hidden layer, and an output layer.

The biosignal processing apparatus according to an embodiment may output the aforementioned output data by applying a first machine learning model including a neural network to data corresponding to a target biosignal. The output data may indicate a possibility that the target bio-signal is a valid signal and a possibility that the corresponding noise source has caused a noise component included in the target bio-signal for each of a plurality of noise sources.

For example, the output data may include a probability score representing the possibility that the target bio-signal is a valid signal. A valid signal is a signal that includes a biosignal component corresponding to a user's actual biosignal and does not include a noise component, and may represent a signal capable of deriving a meaningful result on the user's health status by being collected and then analyzed. . For reference, when the collected bio-signal includes a noise component along with the bio-signal component, the bio-signal cannot be analyzed or a meaningful result may not be derived even if analyzed. A valid signal may represent a signal including a biosignal component and a noise component below a threshold value. A noise component below the threshold value may be a value small enough to be ignored. For example, valid signals may include signals collected in a controlled clinical environment. In addition, the output data may further include a probability score indicating a possibility that a noise source causing a noise component included in the target bio-signal is an electric field.

Exemplarily, the biosignal processing apparatus may input data corresponding to a target biosignal to an input layer of a neural network. The biosignal processing apparatus may propagate data corresponding to a target biosignal through one or more layers from an input layer to an output layer. Data corresponding to the target bio-signal may be extracted as abstracted feature data (eg, feature vector) while being propagated, and the bio-signal processing device may generate output data from the feature data. However, this is a pure example, and the structure of the first machine learning model is not limited to the aforementioned neural network.

The processor may identify a noise source based on output data output by the first machine learning model. Depending on the likelihood scores included in the output data, the processor can identify noise sources. For example, when the probability score that the target biosignal is a valid signal among the plurality of likelihood scores is the maximum likelihood score, the processor may determine the target biosignal as a valid signal. When a likelihood score corresponding to the electric field among the plurality of likelihood scores is the maximum likelihood score, the processor may determine the noise source as the electric field. For another example, the processor may determine a noise source corresponding to a probability score equal to or greater than the threshold value as a noise source for the target biosignal by comparing each probability score with a threshold value.

The processor may obtain a plurality of partial bio-signals by pre-processing the received bio-signal. Each partial bio-signal may represent a signal generated by dividing the bio-signal into an input format of the first machine learning model. The processor extracts one or more signal values (eg, 300 signal values when the sampling rate is 300 times/min) corresponding to a predefined time length (eg, 1 minute) from the received biosignal. Bio-signals of each part can be obtained. The communication unit may receive biosignals (eg, biosignals including 1500 heart potential values) collected for 5 minutes by the biosignal collection device. The processor may acquire 5 partial biosignals each including 300 heart potential values by dividing the biosignal at intervals of 1 minute.

For reference, in this specification, it is mainly described that pre-processing of a biosignal is performed by a processor of a biosignal processing device, but is not limited thereto. For example, the biosignal processing device may receive a biosignal preprocessed and divided by a biosignal collecting device from a biosignal collecting device.

The processor may obtain a target bio-signal by selecting at least some of a plurality of partial bio-signals. The target bio-signal is a signal determined to potentially include a noise component among a plurality of partial bio-signals, and may be an application target of one machine learning model. The processor may apply the first machine learning model to the target physiological signal in order to identify a noise source that causes the noise component. An operation of selecting a target bio-signal from among a plurality of partial bio-signals will be described later with reference to FIG. 3 .

3 illustrates selecting a target bio-signal by applying a second machine learning model. The biosignal processing apparatus may identify a noise source for all of the plurality of partial biosignals, but may also identify a noise source for only some of the plurality of partial biosignals. The processor may select, as the target bio-signal 313 , a part determined to include a noise component among the plurality of partial bio-signals 311 .

In step 310, the processor may select a target biosignal to identify a noise source from among a plurality of partial biosignals. The target bio-signal may represent at least some of a plurality of partial bio-signals determined to include a noise component. The processor may select a target biosignal based on the second machine learning model 312 .

The second machine learning model is a model designed and trained to output a probability score from partial bio-signals, and may include, for example, a neural network. The biosignal processing apparatus according to an embodiment may output a probability score by applying a second machine learning model including a neural network to data corresponding to partial biosignals. Exemplarily, the biosignal processing apparatus may input input data corresponding to a partial biosignal to an input layer of a neural network. The biosignal processing apparatus may propagate data corresponding to a partial biosignal from input data to an output layer through one or more layers. Data corresponding to the partial biosignal may be extracted as abstracted feature data (eg, feature vector) while propagating, and the biosignal processing device may generate a probability score from the feature data. However, this is a pure example, and the structure of the second machine learning model is not limited to the aforementioned neural network.

The processor may output a probability score by applying the second machine learning model to each partial bio-signal. The likelihood score may indicate a possibility that the partial bio-signal includes a noise component. The processor may determine the partial bio-signal as the target bio-signal based on the likelihood score. For example, the processor may determine whether the partial biosignal includes a noise component by comparing the likelihood score with a threshold score.

The processor may identify a noise source for the target physiological signal based on the first machine learning model 314 . As described above in step 220 of FIG. 2 , the processor may identify a noise source of the target biosignal based on the output data of the first machine learning model 314 .

4 illustrates training of a second machine learning model according to one embodiment.

Training data of the second machine learning model may include one or a combination of two or more of a valid signal, a noise signal, and a synthesized signal.

The effective signal may represent a signal composed of biosignal components corresponding to the actual biosignal of the user. A valid signal can be measured in a controlled clinical environment. For example, the valid signal is an electrocardiogram signal composed of biosignal components, and an electrocardiogram signal including a noise component below a threshold (eg, the difference between the user's heart potential value and the received biosignal is a critical value) along with the biosignal components. signals below). A valid signal may be labeled a valid signal when collected.

The noise signal is a signal including a noise component equal to or greater than a threshold value together with the user's actual biosignal, and may represent a biosignal including a biosignal component and a noise component. For example, the noise signal may represent a biosignal obtained from a biosignal collection device affected by a noise source. For reference, the biosignal received from the biosignal collection device may or may not include a noise component along with the biosignal component, depending on whether the biosignal collection device is affected by a noise source. A noise signal may be labeled as a combination of a noise signal and a noise source. For example, a biosignal collected from a biosignal collecting patch affected by an electric field may be labeled as a combination of a noise signal and an electric field.

The synthesized signal may represent a signal generated by adding a noise component to a biosignal component. A noise source may be labeled as a noise component when the noise component is collected. A composite signal may be labeled with a combination of the composite signal and a noise source labeled noise components. For example, a signal generated by adding a noise component labeled as a magnetic field to a valid signal may be labeled as a combination of the composite signal and the magnetic field.

The processor may train the temporary second machine learning model using the valid signal and the noise signal. The processor may further use the synthesized signal along with the valid signal and the noise signal to train the temporary second machine learning model. The temporary second machine learning model is a machine learning model in training, and may be determined as the second machine learning model when training is completed.

Details of training of the second machine learning model will be described as follows. The second machine learning model may include a neural network. The training device may obtain the neural network from an internal database stored in a memory or receive the neural network from an external server through a communication unit. The training device may be implemented independently of the biosignal processing device, but is not limited thereto, and may be integrated into the biosignal processing device.

A training apparatus according to an embodiment may train at least a portion of a neural network through supervised learning. The training device may be implemented as a software module, a hardware module, or a combination thereof. Supervised learning is a technique of inputting training inputs of training data and corresponding training outputs to a neural network, and updating connection weights of connection lines so that output data corresponding to the training outputs of the training data are output. The training data may represent a data set composed of a plurality of training pairs. For example, a training pair may include a training input and a training output, and the training output may indicate a value (eg, ground truth) that should be output from the paired training input. Thus, the training data may include a plurality of training inputs, and may include a training output mapped to each of the plurality of training inputs.

However, training is not limited to supervised learning, and the training apparatus may train at least a portion of the neural network through unsupervised learning. Unsupervised learning may calculate a loss based on an output obtained by forward propagating a training input of training data, and may represent a technique of updating connection weights of connection lines so that the loss is reduced.

The training device may continuously change connection weights based on a result of an objective function defined to measure how close to optimum the currently set connection weights are, and repeatedly perform training. For example, the objective function may be a loss function for calculating a loss between an output value actually output by a neural network based on training input of training data and an expected value desired to be output. The training device may update connection weights in a direction of reducing the value of the loss function.

5 illustrates training a new second machine learning model based on selected training partial biosignals according to an embodiment.

The new second machine learning model may be a second machine learning model trained using biosignals collected from a biosignal collection device of the same type as the user's biosignal collection device. Types of biosignal collection devices may include watches, rings, bands, and patches. However, it is not limited thereto, and the bio-signal collection device of the same type as the bio-signal collection device may be determined based on an attachment position of the user's body, a device model, a bio-signal collection method, and the like.

For example, a new second machine learning model may be additionally trained for one type of biosignal collection device. The biosignal processing device may use a new, additionally trained second machine learning model in order to process data collected through the aforementioned one type of biosignal collection device. The biosignal processing device may exclude data collected from other types of biosignal collection devices from training and/or inference of the second machine learning model for the corresponding type of biosignal collection devices. For example, biosignals collected by the smart watch may be excluded from application of the second machine learning model for the biosignal collection patch.

Based on the second machine learning model (eg, the second machine learning model 312 described above in FIG. 3 ), the processor converts at least some of the plurality of training part biosignals into training data of a new second machine learning model. can be selected The training part biosignals may be candidates for training data of a new second machine learning model for a type of biosignal collecting device. The training part biosignals may be composed of signals collected from a corresponding type of biosignal collecting device. The training part biosignals may be selected based on reliability from the second machine learning model. Confidence may be determined based on the likelihood score output from the second machine learning model.

The processor may train a new temporary second machine learning model using signals selected from among the physiological signals of the training portion. Training of the new temporary second machine learning model may be performed similarly to that described above in FIG. 5 . A new temporary second machine learning model that has been trained may be acquired as a new second machine learning model.

6 shows a biosignal processing device that presents information about a noise source to a user.

The biosignal processing apparatus may induce the user to configure an environment in which an effective biosignal can be collected from the biosignal collecting device by presenting information corresponding to each noise source to the user. A valid bio-signal is a signal composed of components of the user's bio-signal, and may represent a signal that can be an analysis target based on the bio-signal. The bio-signal collection device presents information mapped to the noise source (eg, information on the noise source, behavior correction, etc.) along with information that the bio-signal contains noise components, so that the bio-signal collection device can be obtained without expert intervention. It is possible to present information necessary for removing the noise component to the user.

The biosignal processing apparatus may further include an output unit (eg, the output unit 105 of FIG. 1 ) presenting information about the noise source to the user. The output unit may include one or a combination of two or more of a display, a lighting unit, and a speaker.

A biosignal processing device according to an embodiment may further include a display. The display may visually present information to the user regarding the identified noise source. Each noise source may be mapped to one or a combination of two or more of a text guide message, a behavior correction guide, and a graphical object. The text guide message may be displayed on the display to indicate a message including information about the noise source, such as the noise source, the duration of the noise source, and the strength of the noise source, to the user. The behavior correction guide may indicate a guide including a user's behavior for removing the influence of the noise source identified in the biosignal collecting device. The graphical object may represent a graphical object that visually describes the noise source to the user. For example, when the noise source is identified as an electric field, the display may display a text guide message 611 and a behavior correction guide 612 mapped to the electric field.

The biosignal processing device according to an embodiment may further include a speaker. The speaker may aurally present information about the identified noise source to the user. Each noise source may be mapped to one or a combination of two or more of a voice guidance message, a behavioral correction guide, and an alarm. The voice guidance message is reproduced by the speaker, so that a message including information about the noise source can be presented to the user. The behavior correction guide includes the user's behavior for removing the influence of the noise source identified by the bio-signal collecting device, and can be reproduced by the speaker. The alarm may be reproduced to directly or indirectly present information including noise components in biosignals and information on noise sources to the user.

A biosignal processing device according to an embodiment may further include a lighting unit. The lighting unit may display a color mapped to the identified noise source as brightness mapped to the noise source. A noise source can be mapped to one color or a combination of two or more colors. For example, the lighting unit may continuously display one mapped color. For another example, the lighting unit may blink and display one mapped color. As another example, when a combination of two or more colors is mapped to the identified noise source, the lighting unit may display the combination of colors by alternately flickering each color. However, it is not limited thereto, and as will be described later, the lighting unit may display a color mapped to noise intensity.

7A and 7B show a biosignal processing device that presents information about a noise source and noise intensity to a user.

The biosignal processing apparatus may present information mapped to a combination of a noise source and noise intensity to a user. The noise intensity may indicate a weight of a noise component among collected bio-signals. Noise intensity can be classified into a plurality of levels. Even if the noise source is the same, the biosignal processing apparatus presents information differently according to noise intensity, thereby providing more specific information to the user. The biosignal processing device may present information to the user differently according to the level of the electric field even when the noise source is the same as the electric field.

A biosignal processing device according to an embodiment may include an output unit. An output unit, for example, a display, speaker, and lighting unit may present information about a noise source and noise intensity to a user similar to that described above in FIG. 6 .

The case in which the noise component due to the second level of the electric field is included in the biosignal (720b) has a higher proportion of the noise component in the biosignal than the case in which the noise component due to the first level of the electric field is included in the biosignal (710a). can In other words, the second level of the electric field may represent a stronger noise level than the first level of the electric field.

The biosignal processing device may deliver different information to the user when the electric field is at the second level and when the electric field is at the first level. The biosignal processing device may guide the user's action so that a valid biosignal can be collected by presenting more specific and immediate information to the user at the second level of the electric field than at the first level of the electric field.

For example, the text guide message 711 may be mapped to a combination of an electric field and a first level. The combination of the electric field and the second level may be mapped to a combination of a text guide message 721 , a behavior correction guide 722 , and a graphic object 723 . Unlike the case of the combination of the electric field and the first level, in the case of the combination of the electric field and the second level, the behavior correction guide 722 and the graphic object 723 are presented together with a text guide message, so that the user's attention can be focused. there is. Even with the same noise source (e.g., the electric field in FIGS. 7A and 7B), the text guide message 721 mapped to the second level has a higher strength and stronger intensity than the text guide message 711 mapped to the first level. It may include information about having an expected damage due to noise.

The speaker may play one or a combination of two or more of a voice guidance message mapped to a noise source and noise intensity, a behavioral correction guide, and an alarm. For example, a speaker may play an alarm with a volume or frequency mapped to noise intensity. The alarm 724 mapped to the combination of the electric field and the second level may be played at a higher volume than the alarm 714 mapped to the combination of the electric field and the first level.

The lighting unit may display a combination of color and brightness mapped to a combination of a noise source and an intensity of the noise source. A noise source may be mapped to a light color, and the intensity of a noise source may be mapped to a light brightness. However, it is not limited thereto, and the intensity of the noise source may be mapped to the lighting color regardless of the noise source.

8 illustrates identifying a noise source using an accelerometer together with a biosignal collection device.

In operation 810, the communication unit may further receive acceleration data of the user from the accelerometer together with the biosignal received from the biosignal collection device. An accelerometer may represent a sensor that measures user's movement acceleration. An accelerometer may collect acceleration data. The acceleration data may include data representing a user's motion state. The processor may determine the user's movement state based on the acceleration data. For example, the processor may determine that the user is exercising when the acceleration calculated from the acceleration data is greater than or equal to the critical acceleration.

In operation 820, the biosignal processing apparatus may identify a noise source for the target biosignal based on the acceleration data received from the accelerometer together with the output data of the first machine learning model.

The processor may identify a noise source based on at least one of the acceleration data and output data of the first machine learning model.

The processor according to an embodiment may identify a noise source independently based on acceleration data and output data of the first machine learning model. The processor may determine the noise source as motion based on the acceleration data. The processor may identify a noise source based on output data output by applying the first machine learning model to the target bio-signal. Both the noise source determined based on the acceleration data and the noise source determined based on the output data of the first machine learning model may be determined as noise sources for the target biosignal.

The processor according to another embodiment may identify a noise source based on one of the acceleration data and the output data of the first machine learning model. For example, in response to a case where motion is determined as the noise source of the target biological signal based on acceleration data, identification of the noise source based on the first machine learning model may be omitted. For another example, in response to determining that the noise source of the target physiological signal is motion based on the first machine learning model, identification of the noise source based on acceleration data may be omitted.

However, in the present disclosure, the identification of the noise source based on the first machine learning model is mainly described as being separated from the identification of the noise source based on the acceleration data, but is not limited thereto. For example, the processor may output output data by further applying the target biosignal and acceleration data to the first machine learning model.

The embodiments described above may be implemented as hardware components, software components, and/or a combination of hardware components and software components. For example, the devices, methods and components described in the embodiments may include, for example, a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a field programmable gate (FPGA). array), programmable logic units (PLUs), microprocessors, or any other device capable of executing and responding to instructions. The processing device may execute an operating system (OS) and software applications running on the operating system. A processing device may also access, store, manipulate, process, and generate data in response to execution of software. For convenience of understanding, there are cases in which one processing device is used, but those skilled in the art will understand that the processing device includes a plurality of processing elements and/or a plurality of types of processing elements. It can be seen that it can include. For example, a processing device may include a plurality of processors or a processor and a controller. Other processing configurations are also possible, such as parallel processors.

Software may include a computer program, code, instructions, or a combination of one or more of the foregoing, which configures a processing device to operate as desired or processes independently or collectively. You can command the device. Software and/or data may be any tangible machine, component, physical device, virtual equipment, computer storage medium or device, intended to be interpreted by or provide instructions or data to a processing device. , or may be permanently or temporarily embodied in a transmitted signal wave. Software may be distributed on networked computer systems and stored or executed in a distributed manner. Software and data may be stored on computer readable media.

The method according to the embodiment may be implemented in the form of program instructions that can be executed through various computer means and recorded on a computer readable medium. The computer readable medium may include program instructions, data files, data structures, etc. alone or in combination, and the program instructions recorded on the medium may be specially designed and configured for the embodiment or may be known and usable to those skilled in the art of computer software. may be Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks and magnetic tapes, optical media such as CD-ROMs and DVDs, and magnetic media such as floptical disks. - includes hardware devices specially configured to store and execute program instructions, such as magneto-optical media, and ROM, RAM, flash memory, and the like. Examples of program instructions include high-level language codes that can be executed by a computer using an interpreter, as well as machine language codes such as those produced by a compiler.

The hardware device described above may be configured to operate as one or a plurality of software modules to perform the operations of the embodiments, and vice versa.

As described above, although the embodiments have been described with limited drawings, those skilled in the art can apply various technical modifications and variations based on this. For example, the described techniques may be performed in an order different from the method described, and/or components of the described system, structure, device, circuit, etc. may be combined or combined in a different form than the method described, or other components may be used. Or even if it is replaced or substituted by equivalents, appropriate results can be achieved.

Therefore, other implementations, other embodiments, and equivalents of the claims are within the scope of the following claims.

Claims (11)

a communication unit that receives a user's biosignal from a device that collects biosignals; and
A noise source that causes a noise component included in the target bio-signal based on output data output by applying a first machine learning model to a target bio-signal among a plurality of partial bio-signals divided from the received bio-signal ( source)
Biometric signal processing device comprising a.
According to claim 1,
the processor,
Determining the noise source as one or a combination of two or more of electric field, magnetic field, motion, and temperature,
Bio-signal processing device.
According to claim 1,
The biosignal processing device,
A display displaying at least one or a combination of two or more of a text guide message, a behavior correction guide, and a graphic object mapped to the noise source identified by the processor.
Biometric signal processing device further comprising a.
According to claim 1,
The biosignal processing device,
A speaker that reproduces one or a combination of two or more of a voice guidance message mapped with the noise source identified by the processor, a behavior correction guide, and an alarm.
Biometric signal processing device further comprising a.
According to claim 1,
The biosignal processing device,
A lighting unit displaying one color or a combination of two or more colors mapped with the noise source identified by the processor
Biometric signal processing device further comprising a.
According to claim 1,
The communication department,
Further receiving acceleration data of the user from an accelerometer together with the biosignal;
the processor,
Further based on the acceleration data together with the output data, determining the noise source as motion for a partial biosignal corresponding to the acceleration data.
Bio-signal processing device.
According to claim 1,
the processor,
Selecting the target bio-signal from among the plurality of partial bio-signals based on probability scores output by applying a second machine learning model to each of the plurality of partial bio-signals;
Bio-signal processing device.
According to claim 7,
the processor,
Provisional likelihood score and true value generated according to the application of the second machine learning model to a training biosignal including a valid signal measured in a controlled clinical environment and a noise signal containing a noise component obtained from a biosignal collection device ( calculate the objective function value between the ground truth),
Repeating updating parameters of the second machine learning model so that the calculated objective function value converges,
Bio-signal processing device.
According to claim 8,
the processor,
calculating the objective function value further based on a synthesized signal generated by adding a noise component to a biosignal component together with the valid signal and the noise signal;
Bio-signal processing device.
According to claim 7,
the processor,
Training a new second machine learning model using a training part biosignal selected based on the reliability of the likelihood score output by the second machine learning model,
Bio-signal processing device.
In the biological signal processing method,
Receiving a user's biosignal from a device that collects biosignals; and
identifying a noise source generated in the target bio-signal based on output data output by applying a first machine learning model to a target bio-signal among a plurality of partial bio-signals divided from the received bio-signal;
Biometric signal processing method comprising a.

Images