KR20200034505A - Method for denoising of periodic vital signal and apparatus thereof - Google Patents

Abstract

One embodiment of the present invention relates to a noise removal method capable of effectively removing a noise included in a vital signal having a predetermined period and having first and second feature points within the period. More specifically, disclosed is a method of effectively removing a noise by inputting a vital signal as time series data. The method comprises: a sampling step; a group separation step; a learning step; and a test input step.

Description

Method for removing noise of periodic biosignal and apparatus therefor {Method for denoising of periodic vital signal and apparatus thereof}

The present invention relates to a method and apparatus for removing noise of a periodic biosignal, and more specifically, to effectively remove noise of a biosignal having a constant period among biosignals that can be measured through a device from a person or an animal. It relates to a method and apparatus.

2. Description of the Related Art Businesses that measure personal bio signals and provide personalized medical services with information obtained from bio signals are widely known. In order for the above-mentioned customized medical service to be effectively provided to a visitor who visits a clinic or a hospital, it is obvious that a biosignal that can effectively display information about a visitor's disease and disease by minimizing noise must be measured.

However, the types of bio-signals that can be measured from a person are not only various, but noise is inevitably added to the bio-signals according to adjacent organs or measurement environments in the process of measuring bio-signals for a specific organ or region. A method of effectively removing noise from a biosignal and minimizing loss of unique characteristic information possessed by the biosignal has been studied for a long time.

As an example of a method of removing noise of a biosignal, there is a method of removing only a specific frequency component or leaving only a specific frequency component using a frequency filter. While the above method has an advantage that it is possible to simply leave only the necessary frequency components, a phase shift occurs for all or part of the biosignal, so even if it is restored as a time domain signal, it is specific to the original biosignal. In addition to not being able to accurately know the information on the viewpoint, when the noise frequency falls within the frequency range of the biosignal, there is a limitation in that it is logically impossible to remove the noise.

In addition, as another example of a method of removing noise of a biosignal, there is a method using a denoising autoencoder. The noise canceling automatic encoder is a module (or program) that operates with an algorithm created using the structure of an artificial neural network (ANN), and has the feature of learning so that the input and output of the internally implemented artificial neural network are the same. The noise canceling automatic encoder includes an artificial neural network consisting of an input layer, a hidden layer, and an output layer, trained to reduce the difference between input and output, and repeats training. As you do, you will grasp the characteristics of the input. The noise canceling auto-encoder outputs the bio-signal with the noise removed when inputting the bio-signal with the noise after the training is completed for a specific bio-signal.

However, since the automatic noise canceling encoder operates on the basis of a general artificial neural network, it receives a parameter of a fixed point in time and estimates the output value corresponding to the time, and outputs it. There is a limitation in that it cannot be input to an automatic noise canceling encoder without going through the process. This is because, in general, a biosignal is not simply data that changes with the passage of time, but a signal value at a certain past time point is expressed in a form that affects a signal value at a future time point rather than the time point.

1. U.S. Patent Registration No. 5339818 (registered on August 23, 1994) 2. United States Registered Patent Publication No. 9767557 (Registration on September 19, 2017) 3. US Published Patent Publication No. 2017-0273632 (released on September 28, 2017)

1.Innovative continuous non-invasive cufless blood pressure monitoring based on photoplethysmography technology. Intensive Care Med, 39: 1618-1625. Ruiz_rodriguez, et. al. (2013). 2. Improving photoplethysmographic measurements under motion artifacts using artificial neural network for personal healthcare. IEEE Transactions on Instrumentation and Measurement, doi: 10.1109 / TIM. 2018.2829488 Roy et al. (2018).

The technical problem to be solved by the present invention is to provide a method and apparatus for effectively removing noise contained in a biosignal having a certain period among biosignals and having at least two or more feature points in one period. To do it.

A method according to an embodiment of the present invention for solving the technical problem includes: a sampling step of obtaining a plurality of sampled signals by sampling a biosignal of a user having a constant period based on a preset sampling rate; A group separation step of separating the sampled signals into a learning group and a test group, and adding noise to a sampling signal included in the learning group; The noise-added sampling signal is sorted in the order of signal generation time, and the alignment is arranged in a denoising autoencoder module that sequentially performs encoding and decoding based on an artificial neural network (ANN). A learning step of inputting a sampling signal in a time series to train the noise canceling automatic encoder module; And a test input step of adding noise to the sampling signal included in the test group and inputting the noise to the noise removal automatic encoder module in time series to remove noise added to the sampling signal included in the test group.

Apparatus according to another embodiment of the present invention for solving the above technical problem is a sampling unit for obtaining a plurality of sampled signals by sampling a biosignal of a user having a constant period based on a predetermined sampling rate; A group separation unit separating the sampled signals into a learning group and a test group, and adding noise to a sampling signal included in the learning group; The sampling signal to which the noise is added is arranged in the order of signal generation time, and the alignment is performed in a denoising autoencoder module that sequentially performs encoding and decoding based on an artificial neural network (ANN). A module learning unit that inputs the sampled signal in a time series to learn the automatic noise canceling encoder module; And a test input unit that adds noise to the sampling signal included in the test group and inputs the time-series to the noise removal automatic encoder module to remove noise added to the sampling signal included in the test group.

One embodiment of the present invention provides a computer-readable recording medium storing a program for implementing a method for removing noise of a periodic biosignal.

According to the present invention, it is possible to effectively remove noise included in a biosignal having a constant period among biosignals and having at least two or more feature points in one period.

1 is a block diagram of an example of an apparatus for removing noise in a biosignal according to the present invention.
2 is a diagram schematically showing an example of a user's biosignal that is a sampling target of the sampling unit.
FIG. 3 is a diagram for explaining a process in which the module learning unit learns the noise canceling automatic encoder module.
4 to 8 are diagrams schematically showing noise added to the sampling signal included in the test group.
9 is a diagram schematically showing the operating characteristics of the bidirectional circulating neural network automatic encoder.
10 is a diagram schematically showing the noise reduction performance according to the bidirectional circulatory network noise removal automatic encoder.
11 is a diagram schematically showing an example of constructing a signal without noise through a sliding window method.
12 is a flowchart illustrating an example of a method of removing noise of a periodic biosignal according to the present invention.

The present invention can be applied to a variety of transformations and may have various embodiments, and specific embodiments will be illustrated in the drawings and described in detail in the detailed description. Effects and features of the present invention, and methods for achieving them will be clarified with reference to embodiments described below in detail with reference to the drawings. However, the present invention is not limited to the embodiments disclosed below, but may be implemented in various forms.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings, and the same or corresponding components will be denoted by the same reference numerals and redundant description will be omitted when described with reference to the drawings. .

In the following examples, terms such as first and second are not used in a limited sense, but for the purpose of distinguishing one component from other components.

In the following embodiments, singular expressions include plural expressions, unless the context clearly indicates otherwise.

In the following examples, terms such as include or have are meant to mean that features or components described in the specification exist, and do not preclude the possibility of adding one or more other features or components.

When an embodiment can be implemented differently, a specific process order may be performed differently from the described order. For example, two processes described in succession may be performed substantially simultaneously, or may be performed in an order opposite to that described.

1 is a block diagram of an example of an apparatus for removing noise in a biosignal according to the present invention.

Referring to FIG. 1, the apparatus 10 for removing noise of a biosignal according to the present invention includes a sampling unit 110, a group separation unit 130, a module learning unit 150, a test input unit 170, and a test verification unit It can be seen that it includes 190 and the database 195. Depending on the embodiment, the test verification unit 190 and the database 195 may be omitted.

The sampling unit 110 obtains a plurality of sampled signals by sampling a user's biosignal based on a preset sampling rate. The biological signal of the user sampled by the sampling unit 110 may include various signals, and is preferably a signal that satisfies a plurality of preset conditions in order to maximize a learning effect according to a circulatory neural network described later.

For example, the user's biological signal sampled by the sampling unit 110 must have a certain period. The fact that the user's biosignal has a certain period means that the part that can be regarded as a feature point in the biosignal appears at a predetermined time, and thus, by using these characteristics, noise is removed, and the damaged biosignal is removed in the process of removing noise. It is easy to recover.

Here, the predetermined time may be a fixed point of time, or a predetermined range of time. For example, a biosignal having a period of 10 seconds is a signal capable of removing noise according to the present invention, and a biosignal whose period changes with a value between 8 and 12 seconds has a basic condition for effectively removing noise. Considered to be satisfied, noise can be eliminated by the present invention.

In addition, the biosignal of the user may include a first feature point and a second feature point for each period. In the biosignal including the first feature point and the second feature point for each period, it is preferable that the time interval between the first feature point and the second feature point is within a preset range.

2 is a diagram schematically showing an example of a user's biosignal that is a sampling target of the sampling unit.

More specifically, FIG. 2 shows an example of a PPG signal according to a photoplethysmogram test, and the PPG signal has a first characteristic point and a second characteristic point with a systolic peak and an aortic dip. It contains as. The signal according to FIG. 2 is a signal segment obtained after only one period from the PPG signal according to the light volume pulse wave test, and the PPG signal of the actual user shows a pattern in which the signal segment shown in FIG. 2 continues to be repeated.

The systolic peak 210 is the highest point in the PPG signal, and the PPG signal has a characteristic that after reaching the systolic peak 210, the systolic peak 210 does not reach a higher point than the systolic peak 210.

Diastolic Peak 250 is a characteristic point of the PPG signal observed after the Systolic Peak 210 appears, and the PPG signal continues to descend to the Dicrotic notch 230 after reaching the Systolic Peak 210, and then the Diastolic Peak (250) ) Until it reaches the signal characteristics.

The reason why the PPG signal must have the above signal characteristics is that when the ventricle contracts, when the pressure is lower than the pressure of the aorta, blood cannot open the meniscus, and pressure is applied to the atrium, and when the ventricle is relaxed This is because a diptic notch is forced to appear due to recoil as the moment closes. There is only a difference between x (280) and y (290) depending on the measurement object, and the PPG signal waveform is Systolic due to the atrium constituting the heart, the mechanical placement and atrium of the ventricle, and the pressure of blood flow to and from the ventricle. Peak (210), Dicrotic notch (230), Diastolic Peak (250) are sequentially observed, Systolic Peak (210), Dicrotic notch (230), Diastolic Peak (250) is the first feature point or the present invention 2 It can be a feature point. Here, x (280) is the PPG signal value at the Systolic Peak 210, y (290) is the PPG signal value of the Dicrotic notch (230) or Diastolic Peak (250).

In addition, Width (270), which is a temporal gap between Systolic Peak (210) and Dicrotic notch (230), can be used to recover the signal when one of Systolic Peak (210) and Dicrotic notch (230) is not observed due to noise. It means a possible time interval, and is an important factor for determining a user's biosignal sampled by the sampling unit 110. By the sampling unit 110, the time interval between the Systolic Peak 210 and the Dicrotic notch 230 is limited to a value within a preset range.

In FIG. 2, for convenience of description, the PPG signal according to the light volume pulse wave test is exemplarily described, but as described above, as a periodic biosignal including the first feature point and the second feature point, the first feature point And it is obvious that any biological signal can be a sampling target if the time interval between the second feature points is within a predetermined range by the sampling unit 110. For example, the biosignals that can be sampled by the sampling unit 110 include electrocardiogram (ECG), electrocardiogram (ECG), seismocardiogram, impedance cardiogram (BCG), ballistocardiogram (BCG), and paper application Signals according to plethysmogram, ultrasound, and electroephalography may all be included.

The sampling unit 110 acquires a plurality of sampled signals by sampling the bio-signals at a preset sampling rate. Here, the sampling rate may be 125 Hz. The plurality of sampled signals means each signal segment constituting the biosignal. For example, if the total length of the biosignal is 10 seconds and the sampling rate is 125 Hz, the number of signal segments of the biosignal is 1250. The length and sampling rate of the biosignal as described above are not fixed values and may vary according to embodiments. Hereinafter, the sampled signal and the signal segment will be used in the same sense.

The group separation unit 130 separates the sampled signals into a learning group and a test group, and adds noise to the sampling signal included in the learning group. The learning group includes signal segments for training the neural network included in the noise canceling automatic encoder described later. The test group includes signal segments to test whether the training is actually enabled after the neural network included in the noise canceling automatic encoder is trained by the signal segments included in the learning group.

The group separation unit 130 additionally undergoes an amplitude-normalized process by subtracting the minimum value of the signal belonging to the group and then dividing it by the maximum value of the signal belonging to the group. Can be processed. This normalization process can greatly improve the learning efficiency of the noise canceling automatic encoder.

Various noises may be included in the noise added by the group separator 130 to the sampled signals included in the learning group. As an example, the group separator 130 may add random noise to sampled signals included in the learning group. As another example, the group separator 130 may include white noise. At least one of noise, sloping noise, and saturation noise may be added to the sampled signals included in the learning group. According to an embodiment, noise in which two or more of white noise, slope noise, and saturation noise are overlapped may be treated as random noise as combined noise. The sampled signals included in the learning group may be copied and stored in the database 195 before noise is added, and then transmitted to the module learning unit 150 by a call to the module learning unit 150 described later.

The module learning unit 150 arranges the sampling signals to which noise is added, in order of signal generation time, and samples arranged in an automatic noise canceling encoder module that sequentially performs encoding and decoding based on the artificial neural network. The signal is input in time series to train the automatic noise canceling encoder module. The module learning unit 150 includes a noise canceling automatic encoder module physically or logically, or is connected to the noise canceling automatic encoder module by wire or wirelessly.

Here, when the same signal is input to the input and output as described in the background of the noise canceling automatic encoder module, the characteristics of the signal are trained by the ANN included in the noise canceling automatic encoder module, and the training is completed (learned state). In the above, it means a module that has the function of outputting a signal with no noise when the signal with noise is input to the input of the automatic noise canceling encoder.

FIG. 3 is a diagram for explaining a process in which the module learning unit learns the noise canceling automatic encoder module.

Referring to FIG. 3, it can be seen that the artificial neural network included in the noise reduction automatic encoder module is composed of an Input Layer 330, a Forward Layer 350, a Backward Layer 370, and an Output Layer 390. In the input layer 330, a segment of a bio-signal is input in time series, and in FIG. 3, this time-series input operation is expressed through Time steps 310.

The module learning unit 150 sorts the noise-added sampling signals in the order of signal generation time, and inputs them serially to the input layer 330 of the artificial neural network of the noise reduction automatic encoder module to forward layer 350 and backward layer. Through 370, the sampling signal added to the learning group is trained.

For example, if the biosignal is sampled at the sampling unit 110 at 125 Hz, and the signal segments included in the learning group are sampled at a length corresponding to 2 seconds of the user's biosignal, the signal segments are 250, and the past time point The signal segment at the time t-249 from the signal segment at time t-249 is sequentially input to the input layer 330. Due to the basic operation characteristics of the noise canceling automatic encoder module, the signal segments (250) of the learning group before the noise is mixed by the group separator 130 are input to the Output Layer 390 in time series.

In FIG. 3, the Forward Layer 350 and Backward Layer 370 are each composed of 40 nodes, and while training signal segments to which noise input in time series is added to the Input Layer 330, finally, the Output Layer ( Like the signal segments input to 390), the signal segments from which noise has been removed are repeatedly learned to be output. In FIG. 3, the number of nodes of the Forward Layer 350 and Backward Layer 370 is not limited to 40, so it may be 20 or 60 nodes.

The Forward Layer 350 and Backward Layer 370 constitute a Hidden Layer (not shown) and perform training for distinguishing between a noise and a non-noisy biosignal among time-sequentially input signal segments. According to FIG. 3, signal segments of various viewpoints are trained by being input and trained in time in the input layer 330 and the output layer 390 with or without noise added, thereby circulating the neural network rather than a simple artificial neural network (ANN). It works as a principle of (RNN). RNN is an artificial neural network with a memory function, and it is possible to learn a plurality of signal segments input in chronological order rather than a signal segment of one time point, so that the signal values of the past time point affect the signal values of the future time point. It is very suitable for processing periodic biosignals.

In FIG. 3, the number of nodes of the Input Layer 330 and the Output Layer 390 is limited to 250, but since these are only exemplary values, according to an embodiment, the nodes of the Input Layer 330 and the Output Layer 390 The number may be less or more than 250.

The module learning unit 150 ends the iterative learning process for the signal segments included in the learning group when the signal segments in which the noise is removed by a certain level or more are output by inputting the signal segments to which noise is added to the noise reduction automatic encoder. Is done.

The test input unit 170 adds noise to the sampling signal included in the test group, inputs the signal segments to which the noise has been added, in time-series to the noise canceling automatic encoder module, and applies the noise added to the sampling signal included in the test group. Remove it. Specifically, when noise is added to the sampling signal included in the test group, the test input unit 170 clockwise inputs the signal segments to which the noise has been added to the noise reduction automatic encoder module to remove noise from the noise removal automatic encoder module. Signal segments are output.

4 to 8 are diagrams schematically showing noise added to the sampling signal included in the test group.

Before describing FIGS. 4 to 8, it is assumed that the sampling signal included in the test group is a signal measured when a user performs a light volume pulse wave (PPG) test.

FIG. 4 is a diagram schematically showing one selected PPG segment when a PPG signal is sampled at a specific sampling rate and separated into a plurality of PPG segments. Referring to FIG. 4 with reference to FIG. 2, it is understood that FIG. 4 shows an original PPG segment including two systolic peaks and three aortic notches. You can.

5 schematically shows a result of applying white noise to the original signal PPG segment. More specifically, FIG. 5 shows a shape in which a waveform having a large vibration rate is applied to the original signal PPG segment of FIG. 4 and changed to a signal segment of a different aspect from the original signal PPG segment according to the characteristic of white noise.

FIG. 6 schematically shows the result of applying the slipping noise to the original signal PPG segment. More specifically, FIG. 6 schematically shows the original signal PPG segment in which the section showing the contraction peak and the aortic pimple disappeared due to the intrinsic characteristic of the slope noise.

7 schematically shows a result of applying saturation noise to the original signal PPG segment. More specifically, FIG. 7 schematically shows the original signal PPG segment that maintains the saturation value by rapidly increasing to a saturation value after the second aortic dip due to the characteristic characteristic of saturation noise. In FIG. 7, after reaching the saturation value, information about the original signal PPG segment is completely lost.

8 schematically shows the result of applying the complex noise to the original signal PPG segment. More specifically, the complex noise of FIG. 8 refers to noise in which at least two or more of the white noise, the slope noise, and the saturated noise described through FIGS. 5 to 7 are mixed, and the characteristics of the noises included in the complex noise are By this, a completely different waveform from the original signal PPG segment is displayed.

The test input unit 170 generates a signal segment to which noise is added by adding noise of at least one of white noise, sloped noise, and saturated noise to the signal segment shown in FIG. 4, and the generated signal segments Time-series input to the input layer 330 of the noise cancellation automatic encoder module learned by the module learning unit 150.

As an optional embodiment, when the test input unit 170 determines that the noise added to the sampling signal included in the test group exceeds the preset ratio by the learned noise removal automatic encoder module, it is determined from the sampling unit 110. The sampled target biosignal may be additionally received, and noise may be removed by inputting the received target biosignal to the trained noise canceling automatic encoder module in time series.

In this optional embodiment, the noise removal automatic encoder module is removed by the module learning unit 150 that the noise added to the sampling signal included in the test group is removed in excess of a preset ratio by the learned noise removal automatic encoder module. This means that it has been properly trained. Here, the preset ratio is a value that is stored in the test input unit 170 in advance or stored in the database 195 and then transferred to the test input unit 170 in the process of the logical input of the test input unit 170. The preset ratio may be increased or decreased depending on how strictly the level of learning of the automatic noise canceling encoder module is set.

In addition, the target biological signal is a noise removal automatic encoder module, which has been trained by the module learning unit 150, and after validation is verified by the test input unit 170 according to the above-described process, it is actually noise. In order to remove the data, it is stored in the database 195 in advance, or refers to a newly received biosignal after validity is verified. After the target biological signal is properly sampled through the sampling unit 110, the noise can be sufficiently removed by being input in time series to the noise canceling automatic encoder module.

The test verification unit 190 outputs an output signal from the noise canceling automatic encoder module that receives a signal segment to which noise is added by the test input unit 170 as an input value, and the sampling signal and output signal included in the test group. The test process is verified through a method of grasping the degree of identity. The output signal output by the noise canceling auto-encoder module is a signal segment corresponding to the input signal segment and has the same length of time, and the test verification unit 190 is configured to perform an RMSE (SMS) between the sampling signal included in the test group and the output signal. The verification process can be performed by calculating the Root Mean Square Error or Correlation and comparing it with a preset value. According to an embodiment, the output signal output from the noise canceling automatic encoder module may be a signal combining signal segments.

As described above, the test verification unit 190, the signal segments included in the test group by the test input unit 170 is input to the noise canceling automatic encoder module in time series, when the noise included in the signal segments is removed, the noise It performs the function of verifying how well it was removed and outputting the verified result to the user. The manager managing the apparatus 10 for removing noise of a biosignal according to the present invention observes the results verified by the test verification unit 190 and is included in each layer of the neural network included in the noise removal automation encoder module. The number of nodes may be adjusted, or parameters for improving the learning algorithm of the module learning unit 150 may be more finely adjusted.

While the database 195 communicates with the sampling unit 110, the group separation unit 130, the module learning unit 150, the test input unit 170, and the test verification unit 190, the noise canceling automatic encoding module is learned. It performs the function of temporarily storing and returning various data transmitted and received in the process of testing and verifying the signal segments included in the test group.

As described above, the noise canceling apparatus 10 according to the present invention is an apparatus for effectively using a noise canceling automatic encoder module based on a conventionally known ANN, having a certain period and having a first feature point and a second feature point within a period. It is possible to remove the noise of the bio-signal containing. More specifically, the present invention is a configuration that allows the operation of the noise canceling automatic encoder module to be performed by RNN without limiting it to ANN, and measures biosignal in time series data format rather than inputting it to the input layer of the neural network at once. We propose a device having a structure for inputting at least one sample of a biosignal every time step (time step).

In particular, when the magnitude of the noise included in the biosignal is large, the feature points of the biosignal may be buried. However, waveform feature accentuation that can be reconstructed by predicting where these feature points are generated using temporal information The same function, as in the present invention, has an advantage that can be used in a noise canceling automatic encoder incorporating RNN.

Hereinafter, additional embodiments of the noise removing apparatus 10 according to the present invention will be described.

As an exemplary embodiment, if the biosignal sampled by the sampling unit 110 is a PPG signal, the first feature point may be an aortic dent, and the second feature point may be a contraction peak. According to the present exemplary embodiment, when the biosignal is a PPG signal, it is assumed that the aortic cavities are positioned earlier in time than the contraction peak, and the learning and testing steps of the circulatory neural network of the noise canceling automatic encoder module are performed. When the verifying unit 190 observes a signal segment from which noise has been removed, and when a contraction peak is not found in a specific period and only aortic pimples are found, a signal component corresponding to the contraction peak is omitted in the process of removing noise. You can easily grasp things. That is, in the present exemplary embodiment, when the noise removal algorithm is executed with the first feature point and the second feature point as the contraction peak and the aortic concave, respectively, the noise is additionally removed when the result is not satisfactory. It may be a method performed to acquire a signal.

As another optional embodiment, the module learning unit 150 not only inputs the signal segments arranged in chronological order into the input layer 330 of the noise canceling automatic encoder module, but also semi-timely. Through this, it is possible to train an automatic noise canceling encoder module. The device implemented according to the present exemplary embodiment is a bidirectional circulating neural network noise reduction automatic encoder (BRDAE: BRDAE), rather than simply inputting a signal segment in one direction, and inputting a signal segment in both directions to the input layer 330. Bidirectional Recurrent Denoising AutoEncoder).

9 is a diagram schematically showing the operating characteristics of the bidirectional circulating neural network automatic encoder.

First, in the case of a recurrent denoising autoencoder (RDAE), signal segments arranged in chronological order are input to the noise canceling autoencoder in order from 0 to i (time series) to be learned. At this time, it is assumed that 0 in the signal segment means the most advanced time point.

Subsequently, in the bidirectional cyclic neural network noise reduction automatic encoder according to the present exemplary embodiment, not only the signal segments sorted in chronological order are input to the noise reduction automatic encoder in a sequence from 0 to i (time series), It is entered and learned in the order of i to 0 (semi-serial).

As described above, according to the bidirectional circulatory network noise canceling automatic encoder, not only information that is subsequently input but also information that is input on the contrary can be applied and used. From the viewpoint of removing the noise of the biosignal, if the large noise is located at the front of the signal segment, the noise is removed for the remaining parts included in the signal segment by the circulatory network noise canceling automatic encoder. Performance may drop dramatically. This is because the noise located at the very front disturbs the time series information of the signal segment, so that it is impossible to identify the feature points of the signal segment.

As an example, when a very large amount of noise is mixed from the first sample to the systolic peak in the signal segment of the light volume pulse wave, the contraction peak itself cannot be identified, and time-series information before and after the contraction peak is meaningless. As it becomes obsolete, it may become impossible to rebuild the contraction peak with the existing unidirectional circulatory network noise canceling automatic encoder.

In the above situation, according to the bidirectional circulatory network noise canceling automatic encoder, as the cyclic neural network learns signal segments in both directions, when a certain signal segment is forward in time, aortic artery after a certain time has elapsed after the contraction peak appears. In addition to not only the appearance of pimples, information that a contraction peak appears after the aortic pimples appear in the reverse direction can be considered in restoring and reconstructing the signal segment. In the above-described embodiment, for convenience of description, the signal segment is described as a PPG segment, and the first characteristic point and the second characteristic point are described as contraction peaks and aortic pimples. It will be apparent to those skilled in the art that it can be a segment of a signal.

10 is a diagram schematically showing the noise reduction performance according to the bidirectional circulatory network noise removal automatic encoder.

Referring to FIG. 10, the noise-removed output signal 1010 is a signal according to an irregularly vibrating light volume pulse wave inspection (PPG), and the noise-removed output signal 1030 is a complex noise similar to FIG. 8. You can see that this is an added form of the signal.

Referring to FIG. 10, although there is almost no external similarity between the output signal 1010 from which noise is removed and the output signal 1030 before noise is removed, noise is generated in the output signal 1030 before noise is removed. Not only does this feature point of the removed output signal 1010 be included, but also according to the learned information of the bidirectional cyclic neural network noise removal automatic encoder, as the noise is removed from the output signal 1030 before the noise is removed, finally An output signal 1010 from which noise is removed can be obtained.

The test verification unit 190 may verify the similarity between the output signal 1010 from which noise is removed and the output signal 1030 before noise is removed according to an exemplary embodiment. For example, when the biosignal is a PPG signal according to a light volume pulse wave test, the test verification unit 190 calculates a heart rate (HR) based on the PPG signal from which noise is removed, and according to the calculated PPG signal. The noise reduction rate for the biosignal of the present invention can be verified by comparing the heart rate with a heart rate according to an ECG (ECG) stored in advance and calculating a correlation and a root mean square error (RMSE). .

Denoising Method None Wavelet DAE RDAE BRDAE Correlation 0.408 ± 0.301 0.406 ± 0.304 0.558 ± 0.238 0.620 ± 0.254 0.686 ± 0.147 RMSE 14.8 ± 15.7 14.4 ± 15.0 6.2 ± 4.9 7.5 ± 10.7 6.1 ± 8.2

Table 1 is a table showing the results of removing the noise of the PPG signal to which the test verifier 190 added noise according to various noise removal methods.

Referring to Table 1, the noise reduction was not performed at all (None) or the method according to the present invention (RDAE, BRDAE) than when the noise of the PPG signal was removed by a known method (Wavelet, DAE: Denoising AutoEncoder). Therefore, it can be seen that the case where the noise is removed has a higher correlation and a lower RMSE. In other words, when denoising a periodic biosignal as a method of using information from a past point in time, such as a circulatory neural network, the effect of removing noise due to denoising is superior to that of a conventionally known method. In addition, it can be seen that when the noise is removed according to the bidirectional circular neural network noise removal automatic encoder rather than the unidirectional, the higher correlation and lower RMSE are displayed than when the noise is removed using the unidirectional circular network noise removal automatic encoder. .

As another optional embodiment, the test input unit 170 is included in the test group through a sliding window method based on an input sampling rate lower than the sampling rate used by the sampling unit 110 for sampling. The inputted sampling signal is input in time series to the noise canceling automatic encoder module, and the test verification unit 190 restores the output signal generated by overlapping based on the input sampling rate and compares the generated output signal with the sampling signal included in the test group. Can be verified.

11 is a diagram schematically showing an example of constructing a signal without noise through a sliding window method.

Even in the bi-directional circulatory network noise canceling automatic encoder, there may be a case where noise for a biosignal is not effectively removed. In this case, the signal segments are superimposed in a sliding window method, input to the input layer of the neural network of the noise canceling automatic encoder module, and the output signal is calculated by averaging after outputting the output according to the input. High noise reduction effect can be expected.

Referring to FIG. 11, the PPG signal of 125 Hz is input to the bidirectional circulating neural network noise reduction automatic encoder (BRDAE) at a time of 2 seconds (250 samples) at a time, and the sliding window is sampled by 50 to remove the PPG. You can acquire a signal. Since the input signal segment is 250 samples and the sliding window is 50 samples, the overlapping part between the input signal segments corresponds to a maximum of 200 samples, and the noise is removed through the method of averaging and considering the overlapping part The signal can be obtained.

11 is a bidirectional cyclic neural network noise removal automatic encoder, by repeating a method in which the noise-removed segments 1110 only take an average value of overlapping portions through a sliding window method, the biosignal according to the average result 1130 The resulting process is shown schematically. In FIG. 11, the biosignal according to the average result 1130 is a PPG signal according to an input signal, but if it is a biosignal having a certain period and having a first feature point and a second feature point for each period, the biosignal is not a PPG signal. It will be apparent to those skilled in the art that a method of calculating a biosignal from which noise is removed may also be applied through a sliding window method.

12 is a flowchart illustrating an example of a method of removing noise of a periodic biosignal according to the present invention.

The method according to FIG. 12 may be implemented by the apparatus 10 for removing the noise of the periodic biosignal of FIG. 1, and will be described with reference to FIG. 1, and will be described below with reference to FIGS. 1 to 11 Duplicate description will be omitted.

First, the sampling unit 110 samples a periodic biosignal having a first feature point and a second feature point to obtain a plurality of sampled signals (S1210).

The group separation unit 130 separates the sampled signals into a learning group and a test group, and adds noise to the signals included in the learning group (S1230).

The module learning unit 150 trains the bio-signal by inputting the signals included in the learning group in a time-series manner to the automatic noise canceling encoder module (S1250).

The test input unit 170 adds noise to the signal included in the test group, and inputs it in time series to the noise removal automatic encoder module (S1270).

The test verification unit 190 verifies the identity by comparing the output signal output from the noise canceling automatic encoder module with the sampling signal included in the test group (S1290).

The embodiment according to the present invention described above may be implemented in the form of a computer program that can be executed through various components on a computer, such a computer program can be recorded on a computer-readable medium. At this time, the medium is a hard disk, a magnetic medium such as a floppy disk and a magnetic tape, an optical recording medium such as a CD-ROM and DVD, a magneto-optical medium such as a floptical disk, and a ROM , Hardware devices specially configured to store and execute program instructions, such as RAM, flash memory, and the like.

Meanwhile, the computer program may be specially designed and configured for the present invention or may be known and available to those skilled in the computer software field. Examples of computer programs may include machine language codes such as those produced by a compiler, as well as high-level language codes that can be executed by a computer using an interpreter or the like.

The specific implementations described in the present invention are examples, and do not limit the scope of the present invention in any way. For brevity of the specification, descriptions of conventional electronic configurations, control systems, software, and other functional aspects of the systems may be omitted. In addition, the connection or connection members of the lines between the components shown in the drawings are illustrative examples of functional connections and / or physical or circuit connections. In the actual device, alternative or additional various functional connections, physical It can be represented as a connection, or circuit connections. In addition, unless specifically mentioned, such as “essential”, “importantly”, etc., it may not be a necessary component for application of the present invention.

In the specification (particularly in the claims) of the present invention, the use of the term “above” and similar indication terms may be in both singular and plural. In addition, in the case of describing a range in the present invention, as including the invention to which the individual values belonging to the range are applied (if there is no contrary description), each individual value constituting the range is described in the detailed description of the invention. Same as Finally, unless there is an explicit or contradictory description of the steps constituting the method according to the invention, the steps can be done in a suitable order. The present invention is not necessarily limited to the order of description of the above steps. The use of all examples or exemplary terms (eg, etc.) in the present invention is merely for describing the present invention in detail, and the scope of the present invention is limited due to the examples or exemplary terms, unless it is defined by the claims. It does not work. In addition, those skilled in the art can recognize that various modifications, combinations, and changes can be configured according to design conditions and factors within the scope of the appended claims or equivalents thereof.

Claims (19)

A sampling step of obtaining a plurality of sampling signals by sampling a biosignal of a user having a constant period based on a preset sampling rate;
A group separation step of separating the sampling signals into a learning group and a test group and adding noise to a sampling signal included in the learning group;
The noise-added sampling signal is sorted in the order of signal generation time, and the alignment is arranged in a denoising autoencoder module that sequentially performs encoding and decoding based on an artificial neural network (ANN). A learning step of inputting a sampling signal in a time series to train the noise canceling automatic encoder module; And
A periodic living body comprising a test input step of adding noise to the sampling signal included in the test group and inputting the noise to the noise removal automatic encoder module in time series to remove noise added to the sampling signal included in the test group. How to remove signal noise. According to claim 1,
The above method,
And a test verification step of verifying the identity of the output signal output from the noise removal automatic encoder module and the sampling signal included in the test group. According to claim 1,
The noise is
A method for removing noise of a periodic biosignal characterized by at least one of white noise, sloping noise, and saturation noise. According to claim 1,
The biosignal,
A method of removing noise of a periodic biosignal, characterized in that it is a PPG (photoplethysmogram) signal calculated as a result of performing a light volume pulse wave test for the user. According to claim 4,
The biosignal includes a first feature point and a second feature point in a period,
The first feature point is a systolic peak,
The second feature point is a method of removing the noise of a periodic biosignal, characterized in that the aortic dig (dicrotic notch). According to claim 4,
The biosignal includes a first feature point and a second feature point in a period,
The first feature point is an aortic dip (dicrotic notch),
The second feature point is a method for removing noise of a periodic biosignal, characterized in that the systolic peak (systolic peak). According to claim 1,
The learning step,
In addition to inputting the sorted sampling signal in time series, the input of the sorted sampling signal in a reverse time series removes the noise of the periodic biosignal characterized by learning the automatic noise canceling encoder module. How to. According to claim 2,
The test input step,
Through the sliding window method based on the input sampling rate lower than the sampling rate, the sampling signal included in the test group is input in time to the noise reduction automatic encoder module,
The test verification step,
A method of removing noise of a periodic biosignal that verifies the identity by comparing the output signal generated by overlapping restoration based on the input sampling rate with a sampling signal included in the test group. According to claim 1,
The test input step,
If it is determined that the noise added to the sampling signal included in the test group has been removed by exceeding a preset ratio by the learned noise cancellation automatic encoder module, the target biological signal sampled in the sampling step is additionally received,
And removing the noise by inputting the received target biosignal in time series to the learned noise canceling automatic encoder module. A computer-readable recording medium storing a program for executing the method according to any one of claims 1 to 9. A sampling unit to obtain a plurality of sampling signals by sampling a biosignal of a user having a constant period based on a preset sampling rate;
A group separation unit separating the sampling signals into a learning group and a test group, and adding noise to a sampling signal included in the learning group;
The sampling signal to which the noise is added is arranged in the order of signal generation time, and the alignment is performed in a denoising autoencoder module that sequentially performs encoding and decoding based on an artificial neural network (ANN). A module learning unit that inputs the sampled signal in a time series to learn the automatic noise canceling encoder module; And
A periodic biosignal including a test input unit that adds noise to the sampling signal included in the test group and inputs the noise removal automatic encoder module in time series to remove noise added to the sampling signal included in the test group. Device to remove noise. The method of claim 11,
The device,
The apparatus for removing noise of a periodic bio-signal, further comprising a test verification unit verifying the identity of the output signal output from the noise removal automatic encoder module and the sampling signal included in the test group. The method of claim 11,
The noise is
A device for removing noise of a periodic biosignal, characterized in that it is at least one of white noise, sloping noise, and saturation noise. The method of claim 11,
The biosignal,
A device for removing noise of a periodic biosignal, characterized in that it is a PPG (photoplethysmogram) signal calculated as a result of performing a light volume pulse wave test for the user. The method of claim 14,
The biosignal includes a first feature point and a second feature point in a period,
The first feature point is a systolic peak,
The second feature point is a device for removing noise of a periodic bio-signal, characterized in that the aortic dig (dicrotic notch). The method of claim 14,
The biosignal includes a first feature point and a second feature point in a period,
The first feature point is an aortic dip (dicrotic notch),
The second feature point is a systolic peak (systolic peak) device for removing the noise of the periodic biosignal. The method of claim 11,
The module learning unit,
In addition to inputting the sorted sampling signal in time series, the input of the sorted sampling signal in a reverse time series removes the noise of the periodic biosignal characterized by learning the automatic noise canceling encoder module. Device. The method of claim 12,
The test input unit,
Through the sliding window method based on the input sampling rate lower than the sampling rate, the sampling signal included in the test group is input in time to the noise reduction automatic encoder module,
The test verification unit,
An apparatus for removing noise of a periodic biosignal that verifies the identity by comparing the output signal generated by restoring overlappingly based on the sampling rate with a sampling signal included in the test group. The method of claim 18,
The test input unit,
If it is determined that the noise added to the sampling signal included in the test group has been removed by exceeding the preset ratio by the learned noise canceling automatic encoder module, the target biological signal sampled from the sampling unit is additionally received,
And removing the noise by inputting the received target biosignal in time series to the learned noise canceling automatic encoder module.

Images