Method and apparatus for detecting abnormal bio-signal

Abstract

A method and an apparatus for detecting an abnormal bio-signal can be provided, the method comprising the steps of: generating a plurality of pieces of sample data by randomly extracting, from raw data including bio-signal data, a frame with a first data length; generating a plurality of pieces of unit data by sequentially extracting, from the sample data, a frame with a second data length; generating a training data set on the basis of the plurality of pieces of unit data; and generating an abnormal bio-signal detection model by training an artificial intelligence model with the training data set.

Description

Method and device for detecting abnormal biosignals

[0001] The present invention relates to a method and device for detecting an abnormal biosignal, and more particularly, to a method and device for generating a model for detecting an abnormal biosignal, and further, to a method and device for detecting an abnormal biosignal through biosignal monitoring using the generated model.

[0002] Artificial neural networks (ANNs) mimic biological neural networks. They can learn from a large amount of input data and are used to estimate or approximate results that are difficult to derive using conventional techniques. An ANN consists of layers of interconnected neurons that exchange signals, with synapses having weights that can be tuned based on learning or experience. Training an ANN requires high-quality training data, and its performance can vary depending on the training data.

[0003] Meanwhile, heart disease accounts for a significant proportion of the leading causes of death both domestically and internationally. In particular, sudden, unintentional deaths resulting from suffocation during sleep are most common in infants aged two to four months. Furthermore, as the rate of unaccompanied deaths continues to rise, the need to reduce social costs through the prevention and management of lonely deaths is growing.

[0004] The background technology described above is technical information that the inventor possessed for the purpose of deriving the present invention or acquired in the process of deriving the present invention, and cannot necessarily be considered as publicly known technology disclosed to the general public prior to the application for the present invention.

[0005] The purpose of the present disclosure is to provide a method and device for detecting abnormal biosignals. The problems addressed by the present disclosure are not limited to the technical problems mentioned above. Other technical problems not mentioned above will be clearly understood by those skilled in the art from the description of the present disclosure, and will be further understood by the embodiments of the present disclosure. Furthermore, it will be appreciated that the problems and advantages addressed by the present disclosure can be realized by the means and combinations thereof set forth in the claims.

[0006] As a means for solving the above-described technical problem, a first aspect of the present disclosure may provide a method for detecting an abnormal biosignal, including: a step of generating a plurality of sample data by randomly extracting a frame having a first data length from raw data including biosignal data; a step of generating a plurality of unit data by extracting a frame having a second data length from the sample data in chronological order; a step of generating a learning data set based on the plurality of unit data; and a step of generating an abnormal biosignal detection model by training an artificial intelligence model with the learning data set.

[0007] A second aspect of the present disclosure provides a device for detecting an abnormal biosignal, comprising: a memory having at least one program stored therein; a processor for performing an operation by executing the at least one program; wherein the processor randomly extracts frames having a first data length from raw data including biosignal data to generate a plurality of sample data, extracts frames having a second data length from the sample data in chronological order to generate a plurality of unit data, generates a learning data set based on the plurality of unit data, and trains an artificial intelligence model with the learning data set to generate an abnormal biosignal detection model.

[0008] A third aspect of the present disclosure can provide a computer-readable recording medium having recorded thereon a program for executing the method of the first aspect on a computer.

[0009] A fourth aspect of the present disclosure provides a neuromorphic device that implements a neural network for detecting an abnormal biosignal, the device comprising: at least one processor for driving the neural network; and an on-chip memory including a crossbar array circuit for receiving a command from the at least one processor and performing an operation in the memory; wherein the at least one processor receives input data collected by a sensor, and inputs the input data to the neural network trained based on predetermined learning data to detect the abnormal biosignal, the input data including a biosignal and a noise signal, and the neural network includes an input layer for receiving the input data, one or more hidden layers for extracting features of the input data, and an output layer for outputting an output signal.

[0010] A fifth aspect of the present disclosure provides a method for detecting an abnormal biosignal using a neural network, comprising: receiving input data collected by a sensor; and inputting the input data to the neural network trained based on predetermined learning data to detect the abnormal biosignal; wherein the input data includes a biosignal and a noise signal, and the neural network includes an input layer for receiving the input data, one or more hidden layers for extracting features of the input data, and an output layer for outputting the abnormal biosignal.

[0011] A sixth aspect of the present disclosure can provide a computer-readable recording medium having recorded thereon a program for executing the method of the fifth aspect on a computer.

[0012] In addition, other methods for implementing the present invention, other devices, and computer-readable recording media recording a program for executing the method may be further provided.

[0013] Other aspects, features and advantages other than those described above will become apparent from the following drawings, patent claims and detailed description of the invention.

[0014] According to the problem solving means of the present disclosure described above, unintentional sudden death or lonely death due to suffocation during sleep of infants and young children can be prevented and managed, thereby reducing social costs.

[0015] In addition, according to the problem solving means of the present disclosure, it is possible to construct a machine learning model with high accuracy and minimize abnormal biosignal misrecognition that significantly increases social costs.

[0016] The effects of the embodiments 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 of the present invention.

[0017] FIG. 1 is a diagram for explaining the architecture of a fully-connected neural network (FCNN) according to one embodiment of the present invention.

[0018] FIG. 2 and FIG. 3 are exemplary diagrams for comparing a Von Neumann structure and a CIM (Computing-In-Memory) structure according to one embodiment of the present invention.

[0019] FIG. 4 is a diagram illustrating a neural network according to one embodiment of the present invention.

[0020] FIG. 5 and FIG. 6 are exemplary diagrams of an environment for detecting abnormal biosignals according to one embodiment of the present invention.

[0021] FIG. 7 is a diagram for explaining raw data according to one embodiment of the present invention.

[0022] FIG. 8 is a diagram for explaining a method for generating a plurality of sample data according to one embodiment of the present invention.

[0023] FIG. 9 and FIG. 10 are drawings for explaining a method for generating a plurality of unit data according to one embodiment of the present invention.

[0024] FIGS. 11 to 13 are diagrams for explaining a method for generating a learning data set according to one embodiment of the present invention.

[0025] Figure 14 is a graph showing the results of training an artificial intelligence model with a learning data set according to one embodiment of the present invention.

[0026] FIG. 15 is a block diagram of a device for detecting an abnormal biosignal according to one embodiment of the present invention.

[0027] A method and device for detecting an abnormal biosignal may be provided, comprising: a step of generating a plurality of sample data by randomly extracting a frame having a first data length from raw data including biosignal data; a step of generating a plurality of unit data by extracting a frame having a second data length from the sample data in chronological order; a step of generating a learning data set based on the plurality of unit data; and a step of generating an abnormal biosignal detection model by training an artificial intelligence model with the learning data set.

[0028] In describing the present invention, if it is determined that a detailed description of a related known technology may obscure the gist of the present invention, the detailed description may be omitted, and unless defined otherwise, all terms used in this specification have the same meaning as generally understood by a person of ordinary skill in the technical field to which the present invention belongs.

[0029] The appearances of phrases such as “according to one embodiment,” “relating to one embodiment,” or “according to an implementation of one embodiment” in this specification do not necessarily all refer to the same embodiment.

[0030] The embodiments may be modified in various ways and may take various forms. Some embodiments are illustrated in the drawings and described in detail. However, this is not intended to limit the embodiments to a specific disclosed form, but rather to encompass all modifications, equivalents, and alternatives falling within the spirit and technical scope of the embodiments. The terminology used in the specification is solely for the purpose of describing the embodiments and is not intended to limit the embodiments.

[0031] The terms used in the examples are selected from widely used, common terms, taking into account the functions of the examples. However, these terms may vary depending on the intentions of engineers working in the technical fields to which the examples pertain, precedents, the emergence of new technologies, etc. Furthermore, in certain cases, terms may be arbitrarily selected by the applicant, in which case their meanings will be described in detail in the relevant section. Therefore, the terms used in the examples should be defined based on the meaning of the terms and the overall content of the examples, rather than simply their names.

[0032] Some embodiments of the present disclosure may be represented by functional block configurations and various processing steps. Some or all of these functional blocks may be implemented by various hardware and / or software components that perform specific functions. For example, the functional blocks of the present disclosure may be implemented by one or more microprocessors or by circuit configurations for specific functions.

[0033] Additionally, for example, the functional blocks of the present disclosure may be implemented in various programming or scripting languages. The functional blocks may be implemented as algorithms that run on one or more processors. Furthermore, the present disclosure may employ conventional techniques for electronic configuration, signal processing, and / or data processing.

[0034] Terms such as "database," "element," "means," and "configuration" are broadly used and are not limited to mechanical or physical components. Furthermore, terms such as "-unit" and "-module" described in the specification mean a unit that processes at least one function or operation, which may be implemented in hardware, software, or a combination of hardware and software.

[0035] Additionally, the connecting lines or connecting members between components depicted in the drawings are merely exemplary representations of functional connections and / or physical or circuit connections. In an actual device, connections between components may be represented by various functional connections, physical connections, or circuit connections that may be replaced or added.

[0036] Additionally, terms including ordinal numbers, such as "first" or "second," used herein may be used to describe various components, but the components should not be limited by the terms. The terms are used only to distinguish one component from another.

[0037] Additionally, some components in the drawings may be depicted with somewhat exaggerated sizes or proportions. Additionally, components depicted in one drawing may not be depicted in another drawing.

[0038] Throughout this specification, the term "embodiment" is an arbitrary distinction used to facilitate the description of the invention in this disclosure, and each embodiment is not necessarily mutually exclusive. For example, the configurations disclosed in one embodiment may be applied and / or implemented in other embodiments, and may be modified and applied and / or implemented without departing from the scope of the present disclosure.

[0039] Additionally, the terminology used in this disclosure is for the purpose of describing embodiments and is not intended to limit the embodiments. In this disclosure, singular forms also include plural forms unless specifically stated otherwise.

[0040] Below, embodiments of the present disclosure are described in detail with reference to the attached drawings so that those skilled in the art can easily implement the present disclosure. However, the embodiments of the present disclosure may be implemented in various different forms and are not limited to the embodiments described in the present disclosure.

[0041] Hereinafter, the present invention will be described in detail with reference to the drawings based on this.

[0042] Neuromorphic chips are hardware that mimic the human brain's functions by creating circuits that mimic the shape of neurons. In other words, neuromorphic chips are computer chips that mimic the structure of the nervous system.

[0043] Neuromorphic chips consist only of the circuitry necessary for neural network computation, offering hundreds of times greater power, area, and speed advantages. Unlike conventional computers, the human brain consumes minimal power even when processing vast amounts of data. Similarly, neuromorphic chips mimic the brain's workings. The structures connecting neurons and synapses are arranged in parallel, connecting and disconnecting when data processing is not in progress, saving energy.

[0044] For example, in the case of conventional computers, the von Neumann architecture processes data sequentially when it is input, so it is excellent at executing precisely written programs, but it has problems such as low efficiency in pattern recognition and real-time recognition, as well as power consumption limitations.

[0045] In contrast, neuromorphic chips use analog operations, where data gradually changes across various states, rather than digital ones like 0 and 1. In other words, artificial neurons, configured in parallel, operate in an event-driven manner, without a clock. This allows them to efficiently process atypical text, audio, and video, which are difficult for conventional computers to intuitively recognize. Specifically, neurons (nerve cells) and synapses (connecting wires) are distributed and implemented using silicon transistor circuits and memory devices, enabling parallel data processing.

[0046] In one embodiment, when input data such as an image, voice, or electromagnetic wave is input into a neuromorphic chip, the input data may be converted into predetermined output data through an operation within the neuromorphic chip. The predetermined output data may include a voice / image / electromagnetic wave recognition result through feature classification of the input data. For example, when an electromagnetic wave signal is input as input data, the predetermined output data may include an electromagnetic wave recognition result that classifies whether the input data contains a previously learned abnormal signal.

[0047] Meanwhile, data input into a neuromorphic chip is not limited to the aforementioned images, voices, or electromagnetic waves, and may include various forms of data such as video and text.

[0048] According to one embodiment, a neuromorphic device can be implemented as an Edge AI Chip. Edge AI refers to a technology that executes AI algorithms on hardware devices using edge computing based on system-generated data. AI processing is typically performed in cloud-based data centers that require massive computing capacity, resulting in a high dependence on servers. However, Edge AI allows AI algorithm computations to be performed locally, reducing dependence on the cloud (servers), reducing communication costs, and protecting privacy by eliminating the transmission of sensitive personal information to the cloud. Therefore, configuring a neuromorphic device as an Edge AI Chip not only reduces costs and enhances security, but also enables the implementation of a highly responsive system, as computations are processed instantly within the same hardware.

[0049] In the following description, the device for detecting abnormal biosignals according to one embodiment of the present invention may be the neuromorphic device described above. In other words, the neuromorphic device described above may function as the device for detecting abnormal biosignals according to one embodiment of the present invention.

[0050] FIG. 1 is a diagram for explaining the architecture of a fully-connected neural network (FCNN) according to one embodiment of the present invention.

[0051] Referring to Figure 1, FCNN, a fully connected layer, is included in a convolutional neural network in which all neurons in one layer are connected to all neurons in the next layer, and is a layer used to classify data through a flattened matrix in the form of a one-dimensional array.

[0052] In FCNN, each node has a value, and each neuron has a weight and bias value. When moving from one layer to another, the value obtained by multiplying the weight of each node and adding the bias becomes the node value of the next layer. At this time, if the output value satisfies a specific condition, the output value that has gone through the activation function is input as the node value of the next layer and activated. If the output value does not satisfy a specific condition, an activation function is involved that goes through the activation function and deactivates the node.

[0053] Therefore, since the output value varies depending on the type of activation function, it is important to use an appropriate activation function as needed. Representative examples include the ReLU function and the Softmax function.

[0054] Since FCNN can only input data in a one-dimensional array, if the input data is an image in a three-dimensional array consisting of vertical, horizontal, and channel (color), it must be flattened into one-dimensional data and then input to FCNN, which has the disadvantage of being a drawback. In other words, since the spatial information of the image is ignored, there is a disadvantage in that features contained in the shape cannot be extracted. Therefore, FCNN can be highly utilized when inputting data that can be implemented as a one-dimensional array of data, such as electromagnetic wave data. According to one embodiment, the neural network described below can be an FCNN.

[0055] FIG. 2 and FIG. 3 are exemplary diagrams for comparing a Von Neumann structure and a CIM (Computing-In-Memory) structure according to one embodiment of the present invention.

[0056] Referring to Figure 2, the von Neumann architecture is a computer architecture proposed by John von Neumann, and is a program-stored computer architecture consisting of a typical three-stage structure of a main memory device, a central processing unit, and an input / output device.

[0057] The von Neumann architecture offers the advantage of greatly enhancing versatility, as changing tasks on a computing device only requires changing the software (program) without having to rearrange hardware (wires, etc.). However, because it sequentially executes a series of listed instructions, and those instructions consist of operations that modify values ​​in specific memory locations, it poses a serious problem in the design of high-speed computers. This phenomenon is known as the von Neumann bottleneck.

[0058] To solve the von Neumann bottleneck, alternatives are being proposed, such as the Harvard architecture, which divides memory into where instructions are stored and where data is stored, the CIM architecture, which performs not only data storage but also data operations in memory, and neuromorphic computing, which is an integrated circuit in the form of an artificial neural network that imitates the brain structure of higher animals, and connects countless units that integrate computation and memory functions in parallel like a network, and then operates each unit in an event-driven manner.

[0059] Referring to FIG. 3, it can be seen that the CIM structure is composed of a processor and a memory having computing capabilities.

[0060] Unlike the conventional von Neumann architecture, where all data within memory is transferred to the processor for computation, the CIM architecture performs computations within memory when a processor instruction is received, and only the resulting data is transferred to the processor. This eliminates the movement of large amounts of data, effectively resolving the aforementioned von Neumann bottleneck. Furthermore, it offers the advantage of significantly reduced power consumption.

[0061] A neural network according to one embodiment of the present invention can perform computations using only on-chip memory, without using external memory. For example, the neural network performs computations for each layer based on CIM using only on-chip memory, without using external memory (e.g., off-chip memory), thereby enabling computations to be performed without memory updates while processing input signals. Specifically, the neural network can perform CIM-based computations in which each memory cell is directly connected to the processor.

[0062] However, compared to general memory such as RAM (Random Access Memory), CIM may have limitations in memory capacity and power consumption of computing devices. Therefore, a new hardware structure may be required to produce a high-performance CIM structure chip, which may increase manufacturing costs, etc. Accordingly, the CIM structure may be advantageous for performing relatively small amounts or simple calculations. On the other hand, a neural network according to an embodiment of the present invention may be configured with a memory capable of multi-bit implementation to overcome the shortcomings of the CIM structure. For example, the neural network may be configured with a memory capable of implementing 7 bits 128 analog memory states. By configuring the neural network with a large capacity, unlike general CIM chips that exhibit problems such as heat generation or performance degradation, it can perform massive data processing with low power and high performance even when used for a long time.

[0063] Meanwhile, the on-chip memory can be implemented by a crossbar array circuit. That is, the crossbar array can receive commands from the processor and perform operations, and the on-chip memory can achieve CIM operations by integrating memory elements with a crossbar array structure. As an example, the processor can receive input data collected by a sensor, drive a neural network circuit trained based on predetermined training data, and receive an abnormal biosignal output by the neural network circuit, thereby detecting an abnormal biosignal. Specifically, the processor can input input data to the neural network to detect an abnormal biosignal. Hereinafter, the configuration of the neural network and the training data for generating an abnormal biosignal detection model implemented by the neural network will be described in detail.

[0064] FIG. 4 is a diagram illustrating a neural network according to one embodiment of the present invention.

[0065] Referring to FIG. 4, in one embodiment, the neural network may include an input layer (410) that receives input data (400), one or more hidden layers (420, 430, 440) that extract features of the input data (400), and an output layer (450) that outputs an output signal (460). In this case, the input layer (410), one or more hidden layers (420, 430, 440), and the output layer (450) may refer to each layer of the fully connected layer (FCNN) described above.

[0066] In one embodiment, the neural network can classify input data (400) into one of a normal biosignal, an abnormal biosignal, and noise. Specifically, the neural network can include an input layer (410), one or more hidden layers (420, 430, 440), and an output layer (450) for classifying the input data (400).

[0067] In one embodiment, input data (400) including biosignals and noise signals may be input into the input layer (410) of the neural network. Biosignals are signals generated by the movement of the human body, and noise signals may include simple movements of the monitoring target (e.g., tossing and turning) rather than the biosignals to be monitored (e.g., breathing or heartbeat), or movements of objects other than the monitoring target, such as bystanders or pets.

[0068] In one embodiment, the hidden layers (420, 430, 440) of the neural network can extract or classify features of the input data (400) by reflecting weights on the input data (400). Therefore, the greater the number of hidden layers (420, 430, 440), the more precisely the input data (400) can be processed, thereby improving the algorithm's functionality.

[0069] In one embodiment, the output layer (450) of the neural network can output whether the input data (400) corresponds to a normal biosignal, an abnormal biosignal, or noise, based on features extracted by at least one hidden layer (420, 430, 440). That is, the output layer (450) can output, as an output signal, whether the neural network has classified the input data (400) as a normal biosignal, an abnormal biosignal, or noise.

[0070] FIG. 5 and FIG. 6 are exemplary diagrams of an environment for detecting abnormal biosignals according to one embodiment of the present invention.

[0071] Referring to FIG. 5, a device (hereinafter referred to as “device”) (1) for detecting an abnormal biosignal can communicate with a sensor (2) and one or more external control devices (31, 32, 33, 34).

[0072] As an example, the device (1) can receive input data collected by the sensor (2).

[0073] In addition, when an abnormal bio-signal is detected, the device (1) can transmit an emergency signal to an external control device including at least one of a control center (31), a government office (32), an external terminal (33), and an external server (34).

[0074] In one embodiment, biosignals and noise signals included in the input data can be collected by the sensor (2).

[0075] The change in the thoracic cavity due to breathing of the human body is 4 to 12 mm, and the change in the thoracic cavity due to heartbeat is 0.6 to 2 mm. Therefore, in one embodiment, the sensor (2) may be a millimeter wave radar (mmWave Radar) whose radio wave wavelength band corresponds to the millimeter (mm) band, or whose radio wave frequency band corresponds to 24 GHz to 77 GHz. However, the radio wave wavelength band and frequency band of the sensor (2) are not limited to the above-described embodiment, and the type of radar is not limited to the millimeter wave radar.

[0076] At this time, the sensor (2) can collect characteristics related to the shape or movement of the target human body by emitting radio waves to the target human body and receiving the radio waves reflected from the target's surface. Accordingly, the sensor (2) can collect changes in the chest of the target human body due to breathing or heartbeat.

[0077] Referring to FIG. 6, in one embodiment, the device (1) may include a communication unit (not shown) that uses at least one communication protocol among a Serial Peripheral Interface (SPI) (12), a Digital Video Port (DVP), a Parallel Serial Peripheral Interface (PSPI), an Inter-Integrated Circuit (I2C), a General Purpose Input / Output (GPIO), a Pulse Width Modulation (PWM), an Inter-IC Sound (I2S), and a Universal Asynchronous Receiver / Transmitter (UART) (13).

[0078] The serial peripheral interface (12) is a synchronous serial communication protocol for communication between multiple peripheral devices and a microcontroller or microprocessor. The serial peripheral interface (12) according to one embodiment can be used to exchange data between a single device (1) and multiple slave devices, such as a sensor (2) and an external memory (21). The universal asynchronous transceiver (13) is used to asynchronously transmit data as a form of a serial communication protocol. The universal asynchronous transceiver (13) according to one embodiment can be used for serial communication between the device (1) and a peripheral device, such as an external terminal (33).

[0079] As an example, the device (1) can receive biosignals and noise signals regarding the movement of the thorax of a human body (600) collected by the sensor (2) by communicating with the sensor (2) using a serial peripheral interface (12).

[0080] In addition, the device (1) can transmit and receive data to and from the external memory (21) by communicating with the external memory (21) using the serial peripheral interface (12). At this time, at least one signal among the normal bio-signals, abnormal bio-signals, and noise of the human body (600) detected by the device (1) can be stored in the external memory (21).

[0081] However, when the device (1) communicates with the sensor (2) and / or external memory (21), in addition to the serial peripheral interface (12), at least one communication protocol or communication device among DVP (Digital Video Port), PSPI (Parallel Serial Peripheral Interface), Universal Asynchronous Transceiver (UART), I2C (Inter-Integrated Circuit), GPIO (General Purpose Input / Output), and PWM (Pulse Width Modulation) may be used. That is, the device (1) may include a communication unit (not shown) that uses at least one communication protocol among DVP (Digital Video Port), PSPI (Parallel Serial Peripheral Interface), Universal Asynchronous Transceiver (UART), I2C (Inter-Integrated Circuit), GPIO (General Purpose Input / Output), and PWM (Pulse Width Modulation).

[0082] In one embodiment, the device (1) can transmit an emergency signal to an external terminal (33) or the like using a universal asynchronous transceiver (13). In this case, the external terminal (33) is merely an example, and it is obvious that the emergency signal can be transmitted to the aforementioned control center (31), government office (32), and external server (34). The external terminal (33) may be a terminal of a preset guardian, etc. The emergency signal can be transmitted to the external terminal (33) in the form of text, image, voice, notification, etc.

[0083] However, when the device (1) communicates with an external terminal (33), in addition to the universal asynchronous transceiver (13), at least one peripheral interface among GPIO (General Purpose Input / Output), I2C (Inter-Integrated Circuit), I2S (Inter-IC Sound), and PWM (Pulse Width Modulation) may be used. That is, the device (1) may include a communication unit (not shown) that uses at least one communication protocol among GPIO, I2C, I2S, and PWM.

[0084] However, the above-described communication protocol and peripheral device interface are only an example, and the method by which the device (1) communicates with the sensor (2), external memory (21), external terminal (33), etc. is not limited thereto.

[0085] In one embodiment, the device (1) is installed in a certain space to monitor the bio-signals of the human body (600), and when an abnormal bio-signal is detected from the human body (600), it transmits an emergency signal to a preset external terminal (33), thereby enabling an immediate response to an emergency situation. Accordingly, unintentional suffocation accidents in infants and toddlers, lonely deaths in the elderly, etc. can be prevented.

[0086] FIG. 7 is a diagram for explaining raw data according to one embodiment of the present invention.

[0087] A device according to one embodiment of the present invention is a neuromorphic device including a neural network, and may be an artificial intelligence model trained based on predetermined learning data. In other words, an abnormal biosignal detection model may be implemented using a neuromorphic device.

[0088] At this time, the predetermined learning data may be generated based on raw data. The raw data may be processed as described below to generate the predetermined learning data.

[0089] In one embodiment, the raw data may include normal biosignal data (710), abnormal biosignal data (720), and noise data (730).

[0090] In the following, the biosignal is described assuming a respiratory signal, but this is only for convenience of explanation, and the biosignal may include various signals of the human body, including a heartbeat signal.

[0091] Normal biosignal data (710) refers to data that measures only normal respiration signals. For example, this may be a predetermined amount of data collected by a sensor for a monitored subject who is breathing normally. Similarly, abnormal biosignal data (720) may refer to data that measures unstable respiration, such as when a respiratory disorder occurs. For example, this may be a predetermined amount of data collected by a sensor for a monitored subject who is breathing abnormally. In this case, the predetermined amount may vary depending on various embodiments, such as 3 hours and 20 minutes or 10 hours.

[0092] Noise data (730) may be 3 hours and 20 minutes of measured signals excluding biosignals of the monitored subject, such as the movement of the monitored subject, the observer, or a pet. By generating learning data for noise as well as normal and abnormal biosignals, the product can achieve higher performance when actually applied.

[0093] For example, the amounts of normal biosignal data (710), abnormal biosignal data (720), and noise data (730) may be identical. Accordingly, the model can be appropriately trained on normal / abnormal biosignals and noise, thereby achieving high performance. However, the amounts of each data may vary, and are not limited to the above-described embodiments.

[0094] In one embodiment, each raw data may include in-phase raw data ('I-data' in FIG. 7) and phase-shifted raw data ('Q-data' in FIG. 7). As an example, the in-phase raw data may be quadrature data. The in-phase raw data may refer to the real part of the raw data expressed as a complex number, and the phase-shifted raw data may refer to the imaginary part of the raw data expressed as a complex number. The in-phase raw data and the phase-shifted raw data may be used together to represent different dimensions of a complex signal.

[0095] FIG. 8 is a diagram for explaining a method for generating a plurality of sample data according to one embodiment of the present invention.

[0096] The devices described in FIGS. 1 to 7 refer to neuromorphic devices that detect abnormal biosignals, but a device that generates an abnormal biosignal detection model implemented as a neuromorphic device (hereinafter, “model generation device”) may refer to a device that generates or processes learning data for training an abnormal biosignal detection model.

[0097] In one embodiment, the model generation device may extract a plurality of frames from raw data to generate a plurality of sample data (800). At this time, the model generation device may extract a plurality of frames having a first data length (801) from the raw data.

[0098] As an example, the first data length (801) may be 30 seconds. Assuming that the length of raw data (802) is 3 hours and 20 minutes and that the raw data is sampled at 20 Hz, the number of data points corresponding to the length of raw data (802) may be 240,000, and the number of data points of the first data length (801) corresponding to 30 seconds may be 600.

[0099] In one embodiment, the model generation device can randomly extract multiple frames from raw data. That is, rather than sequentially extracting frames according to data points, the model generation device can randomly extract multiple frames having a first data length (801). In addition, the model generation device can extract multiple frames in a preset number (e.g., 2000) to generate multiple sample data (800) in the preset number.

[0100] Meanwhile, when the raw data includes in-phase raw data and moving-phase raw data, the model generation device can randomly extract a frame having a first data length (801) from the in-phase raw data to generate a plurality of in-phase sample data (800). In addition, the model generation device can extract a frame having a first data length (801) from the moving-phase raw data to correspond to a signal of each of the plurality of in-phase sample data (800) to generate a plurality of moving-phase sample data (810). That is, the in-phase sample data (800) can be randomly extracted, but the moving-phase sample data (810) can extract a signal at a time corresponding to the in-phase sample data (800). As an example, the model generation device can extract a frame shifted by a time point equal to the phase difference between the moving phase raw data and the moving phase raw data from the in-phase sample data (800) and generate the moving phase sample data (810) to correspond to the signal of the in-phase sample data (800).

[0101] FIG. 9 and FIG. 10 are drawings for explaining a method for generating a plurality of unit data according to one embodiment of the present invention.

[0102] Referring to FIG. 9, a plurality of unit data (900 to 990) generated from sample data having a first data length (801) are illustrated.

[0103] In one embodiment, the model generation device may extract frames having a second data length (901) from the sample data in chronological order to generate a plurality of unit data (900 to 990). As an example, the second data length (901) may be 7.5 seconds. If the first data length (801) of the sample data is 30 seconds and there are 600 data points of the first data length (801) according to a data sampling cycle of 20 Hz, the number of data points of the second data length (901) may be 150.

[0104] In one embodiment, the model generation device can sequentially extract frames having a second data length (901) such that the plurality of unit data (900 to 990) overlap by a third data length (902). Specifically, the model generation device can extract a first frame (900) including a first data point, and then extract a second frame (910) such that the second frame (900) overlaps by a third data length (902). Similarly, the model generation device can extract a third frame (not shown) such that the second frame (910) overlaps by a third data length (902).

[0105] For example, the third data length (902) may be 5 seconds. If the second data length (901) is 7.5 seconds, the model generation device may move by 2.5 seconds when extracting each frame to extract the next frame. At this time, if the number of data points of the second data length (901) is 150, the third data length (902) may be 100, and since a total of 10 frames are extracted, 10 pieces of the plurality of unit data (900 to 990) may be generated. In addition, since the plurality of unit data (900 to 990) are generated for each sample data, the unit data corresponding to one row data may be generated in the number obtained by multiplying the number of the plurality of sample data (e.g., 2000) by the number of unit data (e.g., 10) generated from each sample data. In FIG. 9, 10 unit data (900 to 990) are shown to be generated, but the number of multiple unit data (900 to 990) may vary depending on the third data length (902).

[0106] In one embodiment, the plurality of sample data may include a plurality of in-phase sample data and a plurality of moving-phase sample data. Accordingly, the model generation device may generate a plurality of in-phase unit data corresponding to the plurality of in-phase sample data and a plurality of moving-phase unit data corresponding to the plurality of moving-phase sample data. That is, the model generation device may generate a plurality of in-phase unit data for each of the in-phase sample data, and a plurality of moving-phase unit data for each of the moving-phase sample data.

[0107] In one embodiment, the model generation device can convert the generated plurality of unit data into frequency components. For example, the model generation device can perform a fast Fourier transform (FFT) on the generated plurality of unit data.

[0108] Referring to FIG. 10, a plurality of unit data (1010 to 1100) converted into frequency components are illustrated.

[0109] In one embodiment, the model generation device can convert a plurality of in-phase unit data into frequency components and can convert a plurality of moving phase unit data into frequency components.

[0110] FIGS. 11 to 13 are diagrams for explaining a method for generating a learning data set according to one embodiment of the present invention.

[0111] Referring to FIG. 11, multiple unit data (1100 to 1190) are illustrated.

[0112] In one embodiment, the model generation device can generate a learning data set based on a plurality of unit data (1100 to 1190). Specifically, the model generation device can synthesize a plurality of unit data (1100 to 1190) converted into frequency components in the order in which they were extracted.

[0113] Accordingly, a single training data corresponding to a single sample data can be generated. The reason for synthesizing multiple unit data (1100 to 1190) that have already been created by overlapping each other is that when a signal such as a delta function is included in the signal, whether it is noise can be confirmed based on whether there is continuity with adjacent unit data, and this can be reflected in the training data.

[0114] Referring to FIG. 12, a plurality of in-phase unit data (1100 to 1190) and a plurality of moving phase unit data (1101 to 1191) are illustrated.

[0115] In one embodiment, the model generation device can alternately synthesize a plurality of in-phase unit data (1100 to 1190) and a plurality of moving phase unit data (1101 to 1191). The alternately synthesized data may mean that the first moving phase unit data (1101) is sequentially synthesized so that the first in-phase unit data (1100) is positioned behind the first moving phase unit data (1101), the second in-phase unit data (1110) is positioned behind the first moving phase unit data (1101), and the second moving phase unit data (1111) is positioned behind the second in-phase unit data (1110). Similarly, the third in-phase unit data (1120) and the third moving phase unit data (1121) may be positioned behind the second moving phase unit data (1111), and depending on the number of unit data, the last in-phase unit data (1190) and the last moving phase unit data (1191) may be positioned last.

[0116] Referring to FIG. 13, since the raw data includes normal biosignal data, abnormal biosignal data, and noise data, the model generation device can generate normal biosignal learning data (1310) from normal biosignal data (710), abnormal biosignal learning data (1320) from abnormal biosignal data (720), and noise signal learning data (1330) generated from noise data (730).

[0117] A plurality of sample data and a plurality of unit data can be generated for each of normal biosignal data, abnormal biosignal data, and noise data. For example, the model generation device can generate N (e.g., 2000) in-phase sample data (800) and N (e.g., 2000) moving-phase sample data (810) from normal biosignal data (710).

[0118] In addition, since multiple unit data (not shown) generated from one sample data (800) are synthesized again into one learning data, N (e.g., 2000) normal biosignal learning data (1310), abnormal biosignal learning data (1320), and noise learning data (1330) can also be generated.

[0119] In one embodiment, the learning data set generated by the model generation device may include normal biosignal learning data (1310), abnormal biosignal learning data (1320), and noise learning data (1330).

[0120] Meanwhile, in one embodiment, the model generation device can generate an abnormal biosignal detection model by training an artificial intelligence model with a learning data set.

[0121] Figure 14 is a graph showing the results of training an artificial intelligence model with a learning data set according to one embodiment of the present invention.

[0122] In one embodiment, the model generation device can adjust the parameters of an artificial intelligence model using a training data set. The artificial intelligence model may be a machine learning model.

[0123] Referring to Figure 14, as the number of times (the number of epochs) of learning with the training data set increases, the training loss function and validation loss function of the machine learning model change as shown in Figure 14. The training loss function (train_loss) is a value representing the loss for the training data set, and the more accurately the machine learning model predicts the training data set, the lower the training loss function becomes. The validation loss function (val_loss) is a value representing the loss for new data (i.e., validation data) that the machine learning model has never seen before, and the more the machine learning model generalizes, the lower the validation loss function becomes.

[0124] In one embodiment, the model generation device may select a machine learning model having the lowest loss function and the highest accuracy among these and determine it as an abnormal biosignal detection model.

[0125] The abnormal biosignal detection model generated in this way is implemented as a neural network, and the processor of the neuromorphic device inputs the input data collected by the sensor into the abnormal biosignal detection model, and can detect abnormal biosignals based on the output of the abnormal biosignal detection model.

[0126] FIG. 15 is a block diagram of a device for detecting an abnormal biosignal according to one embodiment of the present invention.

[0127] Referring to FIG. 15, the device (1500) may include a communication unit (1510), a processor (1520), and a database (1530). Only components related to the embodiment are illustrated in the device (1500) of FIG. 15. Therefore, those skilled in the art will understand that other general components may be included in addition to the components illustrated in FIG. 15.

[0128] The communication unit (1510) may include one or more components that enable wired / wireless communication with an external server or external device. For example, the communication unit (1510) may include at least one of a short-range communication unit (not shown), a mobile communication unit (not shown), and a broadcast reception unit (not shown). In one embodiment, the communication unit (1510) may use at least one communication protocol among a serial peripheral interface (SPI) and a universal asynchronous receiver / transmitter (UART). Furthermore, in one embodiment, the communication unit (1510) may communicate with sensors, external memory, and external control devices.

[0129] DB (1530) is hardware that stores various data processed within the device (1500), and can store a program for processing and controlling the processor (1520).

[0130] DB (1530) may include random access memory (RAM) such as dynamic random access memory (DRAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), CD-ROM, Blu-ray or other optical disk storage, hard disk drive (HDD), solid state drive (SSD), or flash memory.

[0131] The processor (1520) controls the overall operation of the device (1500). For example, the processor (1520) can control the input unit (not shown), the display (not shown), the communication unit (1510), the DB (1530), etc., by executing programs stored in the DB (1530). The processor (1520) can control the operation of the device (1500) by executing programs stored in the DB (1530).

[0132] The processor (1520) can control at least some of the operations of the device for detecting abnormal biosignals or the model generation device described above in FIGS. 1 to 14.

[0133] The processor (1520) may be implemented using at least one of application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, micro-controllers, microprocessors, and other electrical units for performing functions.

[0134] In one embodiment, the device (1500) may be a server. The server may be implemented as a computer device or multiple computer devices that communicate over a network to provide commands, code, files, content, services, etc. The server may receive data necessary to lighten the neural network model and lighten the neural network model based on the received data.

[0135] Meanwhile, embodiments according to the present invention may be implemented in the form of a computer program that can be executed through various components on a computer, and such a computer program may be recorded on a computer-readable medium. At this time, the medium may include magnetic media such as hard disks, floppy disks, and magnetic tapes, optical recording media such as CD-ROMs and DVDs, magneto-optical media such as floptical disks, and hardware devices specifically configured to store and execute program instructions, such as ROMs, RAMs, and flash memories.

[0136] Meanwhile, the computer program may be specifically designed and constructed for the present invention, or may be one known and available to those skilled in the computer software field. Examples of computer programs may include not only machine language code, such as that generated by a compiler, but also high-level language code that can be executed by a computer using an interpreter or the like.

[0137] According to one embodiment, the method according to various embodiments of the present disclosure may be provided as included in a computer program product. The computer program product may be traded as a product between a seller and a buyer. The computer program product may be distributed in the form of a machine-readable storage medium (e.g., compact disc read-only memory (CD-ROM)), or may be distributed online (e.g., downloaded or uploaded) through an application store (e.g., Play Store™) or directly between two user devices. In the case of online distribution, at least a portion of the computer program product may be temporarily stored or temporarily generated in a machine-readable storage medium, such as the memory of a manufacturer's server, an application store's server, or a relay server.

[0138] Unless the steps constituting the method according to the present invention are explicitly described in a specific order or are otherwise described in a different order, the steps may be performed in any appropriate order. The present invention is not necessarily limited to the order in which the steps are described. The use of all examples or exemplary terms in the present invention is merely intended to illustrate the present invention in detail, and the scope of the present invention is not limited by the examples or exemplary terms unless otherwise defined by the claims. Furthermore, those skilled in the art will appreciate that various modifications, combinations, and variations can be configured according to design conditions and factors within the scope of the appended claims or their equivalents.

[0139] Therefore, the idea of ​​the present invention should not be limited to the embodiments described above, and not only the scope of the patent claims described below but also all scopes equivalent to or equivalently modified from the scope of the patent claims are considered to fall within the scope of the idea of ​​the present invention.

Claims

1. A step of generating a plurality of sample data by randomly extracting a frame having a first data length from raw data including biosignal data; A step of generating a plurality of unit data by extracting frames having a second data length from the sample data in time order; A step of generating a learning data set based on the plurality of unit data; and A method for detecting an abnormal biosignal, comprising: a step of generating an abnormal biosignal detection model by training an artificial intelligence model with the above learning data set.

2. In paragraph 1, The above plurality of sample data and the above plurality of unit data, It is generated for each of the normal bio-signal data, abnormal bio-signal data and noise data included in the above raw data. The above learning data set is, A method for detecting an abnormal biosignal, comprising normal biosignal learning data generated from the normal biosignal data, abnormal biosignal learning data generated from the abnormal biosignal data, and noise signal learning data generated from the noise data.

3. In paragraph 1, The step of generating the above multiple unit data is: A method for detecting an abnormal biosignal, comprising: a step of sequentially extracting frames having the second data length so that the plurality of unit data overlap each other by a third data length.

4. In paragraph 1, The step of generating the above multiple unit data is: A method for detecting an abnormal biosignal, comprising: a step of converting the plurality of generated unit data into frequency components.

5. In paragraph 4, The steps for generating the above learning data set are: A method for detecting an abnormal biosignal, comprising: a step of synthesizing the plurality of converted unit data in the extracted order.

6. In paragraph 5, The above raw data is, Contains in-phase raw data and phase-shifted raw data, The step of generating the above multiple sample data is: A step of generating a plurality of in-phase sample data by randomly extracting a frame having the first data length from the in-phase raw data; and A method for detecting an abnormal biosignal, comprising: a step of extracting a frame having the first data length from the moving phase raw data so as to correspond to a signal of each of the plurality of in-phase sample data, thereby generating a plurality of moving phase sample data.

7. In paragraph 6, The steps for generating the above learning data set are: A method for detecting an abnormal biosignal, comprising: a step of cross-synthesizing a plurality of isophase unit data corresponding to the plurality of isophase sample data and a plurality of moving phase unit data corresponding to the moving phase sample data.

8. In paragraph 1, The step of generating the above abnormal biosignal detection model is: A step of adjusting the parameters of the artificial intelligence model using the above learning data set; and A method for detecting an abnormal biosignal, comprising: a step of generating an abnormal biosignal detection model based on a loss function according to the number of times (epoch number) in which the above adjustment is performed and the accuracy of the artificial intelligence model.

9. In paragraph 1, The above abnormal biosignal detection model is, Implemented as a neuromorphic device, The above neuromorphic device, At least one processor driving the above abnormal biosignal detection model; and A crossbar array circuit for receiving commands from at least one processor and performing operations in memory, At least one processor of the above, Input data collected by the sensor is input into the above abnormal biosignal detection model, A method for detecting an abnormal biosignal, wherein the abnormal biosignal is detected based on the output of the above abnormal biosignal detection model.

10. Memory containing at least one program; A processor that performs a calculation by executing at least one program; The above processor, Generating multiple sample data by randomly extracting a frame having a first data length from raw data including biosignal data, A plurality of unit data are generated by extracting frames having a second data length from the above sample data in time order, and Generate a learning data set based on the above multiple unit data, A device for detecting abnormal biosignals, which creates an abnormal biosignal detection model by training an artificial intelligence model with the above learning data set.

11. A computer-readable recording medium having recorded thereon a program for executing the method of Article 1 on a computer.

Images