KR20240048127A - Signal processing device and signal processing method - Google Patents

Abstract

The device according to the present invention is a device for obtaining a prediction model from a plurality of previously detected sensing signals and correcting the reference setting value of the input sensing signal using the prediction model. The device includes a prediction model acquisition unit (100) ) and a signal processing unit 200, wherein the prediction model acquisition unit includes a data conversion module 110 that converts the sensing signal into a digital signal, and a covariance matrix acquisition module that obtains a covariance matrix based on the converted digital data. (120), a distribution area setting unit 130 that calculates the separation distance between data and sets a data distribution area using the obtained covariance matrix, and a prediction model that obtains a prediction model based on the set distribution area. Includes an acquisition unit 140.
The method for processing a sensing signal according to the present invention includes performing data conversion of the sensing signal (S100), obtaining a covariance matrix based on the converted sensing data (S200), and determining the separation distance between data. It includes calculating and setting the distribution area (S300), acquiring a prediction model using linear regression (S400), and correcting the input sensing signal using the obtained prediction model (S500). , the separation distance between the data may be a stochastic separation distance away from the distribution center of the sensing data.

Description

Signal processing device and signal processing method {SIGNAL PROCESSING DEVICE AND SIGNAL PROCESSING METHOD}

The present invention is a signal processing device and signal processing for improving the sensing performance of the sensor by obtaining a prediction model based on sensing signals detected from sensors installed separately from each other and readjusting the initial setting value of the sensor using the prediction model. It's about method.

Sensors are used in electrical and electronic devices to detect physical and chemical changes. Sensors detect various changes such as temperature, pressure, flow rate, sound waves, and vibration, and convert them into electronic signals and output them to the control element.

Sensors manufactured in the manufacturing process must sense the same amount of change. However, the actual detection value may differ slightly from sensor to sensor. These differences can cause incorrect control operations in devices that require precise measurement values. For example, in a control device where one or more sensors installed at different points are connected, if the difference between signals detected from these sensors must be detected, incorrect sensing information may be provided. Therefore, it is necessary to correct the measurement deviation so that all sensors connected to one control device have the same measurement performance.

Additionally, the sensor's sensing signal may be output with noise included. Most of these noises are generated and introduced from the external environment. In order to obtain a high-quality detection signal, noise must be effectively removed.

Registered Patent Publication No. 10-0959155, Automatic sensor correction method Publication of Patent No. 10-1995-0008791, Method for correcting error in optical sensor of washing machine Registered Patent Publication No. 10-1905829, Sensitivity correction device and method for sensor node Publication of Patent No. 10-2020-0054370, integrated sensor automatic correction device and method for autonomous vehicle

The present invention was devised to solve the above problems, and its purpose is to provide a signal processing device and a signal processing method that allow sensors connected to one controller to have the same sensing performance.

Another object of the present invention is to provide a signal processing device and a signal processing method for removing noise included in a detected sensing signal.

The problems to be solved by the present invention are not limited to the problems mentioned. Other technical problems not mentioned will be clearly understood by those skilled in the art from the following description.

The device according to the present invention is a device for obtaining a prediction model from a plurality of previously detected sensing signals and correcting the reference setting value of the input sensing signal using the prediction model. The device includes a prediction model acquisition unit (100) ) and a signal processing unit 200, wherein the prediction model acquisition unit includes a data conversion module 110 that converts the sensing signal into a digital signal, and a covariance matrix acquisition module that obtains a covariance matrix based on the converted digital data. (120), a distribution area setting unit 130 that calculates the separation distance between data and sets a data distribution area using the obtained covariance matrix, and a prediction model that obtains a prediction model based on the set distribution area. Includes an acquisition unit 140.

The method for processing a sensing signal according to the present invention includes performing data conversion of the sensing signal (S100), obtaining a covariance matrix based on the converted sensing data (S200), and determining the separation distance between data. It includes calculating and setting the distribution area (S300), acquiring a prediction model using linear regression (S400), and correcting the input sensing signal using the obtained prediction model (S500). , the separation distance between the data may be a stochastic separation distance away from the distribution center of the sensing data.

According to the present invention, sensing data that deviates from the data distribution area is judged to be an outlier, thereby increasing the reliability of the sensing data due to noise.

Additionally, the present invention can obtain a prediction model using a previously detected sensing signal, and correct errors in input sensing data using the obtained prediction model.

1 is a configuration diagram of a signal processing device according to the present invention.
Figure 2 shows a method of obtaining a prediction model from the prediction model acquisition unit according to the present invention.

Hereinafter, embodiments of the present invention will be described in detail with reference to the attached drawings. Terms such as device, unit, and module used in the embodiments may be a combination of hardware and software. Hardware may be a data processing device including a CPU or processor, and software may refer to a program that runs on hardware.

The present invention relates to a signal processing device and a signal processing method for readjusting the initial setting value of a sensor and improving sensing performance using a prediction model based on a sensing signal.

1 is a configuration diagram of a signal processing device according to the present invention.

The signal processing device according to the present invention consists of a prediction model acquisition unit 100 and a signal processing unit 200. The signal processing unit processes the sensing signal using the prediction model obtained from the prediction model acquisition unit.

The prediction model acquisition unit 100 acquires a prediction model from a plurality of previously detected sensing signals. The plurality of sensing signals may be signals detected from a temperature sensor, a humidity sensor, a fine dust sensor, a vibration sensor, a current sensor, etc.

The prediction model acquisition unit 100 according to the present invention may be comprised of a data conversion module 110, a covariance matrix acquisition module 120, a distribution area setting module 130, and a prediction model acquisition module 140.

The data conversion module 110 gives priority to removing junk areas included in the sensing signal. Next, the data conversion module performs data conversion on the previously detected sensing signal for data processing in the subsequent module. The data conversion may be digital signal conversion. The sensing signal detected by the sensor may be an analog signal or a digital signal. The data conversion module converts the sensing signal into digital data.

The covariance matrix acquisition module 120 acquires a covariance matrix based on the converted sensing data. The covariance matrix may be composed of the variance of the relationship between two variables and the covariance between the variables. As an example, the covariance matrix may be generated from signals sensed from a sensor installed at a specific point and sensors installed at a plurality of different points. Here, the sensing data may be temperature, humidity, concentration, vibration signal, current signal, etc.

The distribution area setting module 130 calculates the separation distance between data from the generated covariance matrix and sets the distribution area. The distribution area may be a cut-off region divided by a weight.

The data separation distance (d) is calculated using a covariance matrix. Next, at least one cutoff area is set using the weight. Weights can be adjusted depending on the type of sensor and sensing data. The weight can use values from 1 to 3. Preferably, it can be set to 2.5 to 3.0.

In general, most data has a normal distribution. Therefore, the sensing data detected from the sensor may be located within the cutoff area. Sensing data that falls outside the cut-off region can be judged as an outlier, that is, noise.

The prediction model acquisition module 140 acquires a prediction model using linear regression.

Before this step, variables according to the type of sensing data can be set. For example, in the case of a vibration sensor, the set variables may be time, amplitude, and frequency, and in the case of a temperature sensor, the set variables may be time and temperature. .

The prediction model acquisition module models the prediction model. The prediction model according to the present invention can be obtained using a linear regression equation. The regression equation can use the mean square error method to minimize the loss value.

The predicted value obtained by the prediction model is calculated in Equation 1 below.

〈Formula 1〉

here, is the value of each sensor n, is the average value of each sensor n, is the covariance matrix, Silver external ear section, is the regression coefficient.

The prediction model obtained in this way is applied to processing the sensing signal of the sensor input to the signal processing unit 200.

The signal processing unit 200 according to the present invention receives a sensing signal from a sensor, converts the input sensing signal into sensing data, and then uses the obtained prediction model to correct the standard setting value of the sensor.

Hereinafter, a method of acquiring a prediction model from the prediction model acquisition unit according to the present invention will be described.

Figure 2 shows a method of obtaining a prediction model from the prediction model acquisition unit according to the present invention.

The method of obtaining a prediction model according to the present invention includes performing data conversion of a sensing signal (S100), obtaining a covariance matrix based on the converted sensing data (S200), and calculating the separation distance between data. It includes a step of setting a distribution area (S300), a step of acquiring a prediction model using linear regression (S400), and a step of correcting the input sensing signal using the obtained prediction model (S500).

Before explaining, the sensor may be comprised of first to nth sensors. The first to nth sensors are connected to a controller, and these sensors may be installed spaced apart from each other. All sensing signals detected from the sensor at the same time or at different times may be transmitted to the controller. Sensing data transmitted to the controller is transmitted to a signal processing device.

1. Performing data conversion of the sensing signal (S100);

This step is performed in the data conversion module. One or more sensing signals are input to the data conversion module, and the input sensing signals are converted into digital data. A junk region included in the sensing signal (or sensing data) may be removed before or after the digital data conversion. The junk area refers to a signal or data area that is unnecessary for data processing.

2. Obtaining a covariance matrix based on the converted sensing data (S200);

This step is performed in the covariance matrix acquisition module. The covariance matrix acquisition module acquires a covariance matrix based on sensing data detected from sensors installed separately from each other. The covariance matrix is a technique for analyzing multivariate data, and its components include the variance of multivariate variables and the covariance between variables. For example, the covariance matrix may be calculated based on the square of the difference value for each sensed data spaced apart from the average value of the sensed data. The obtained covariance matrix is used to calculate the separation distance between data, which will be described later.

3. Calculating the separation distance between data and setting the distribution area (S300);

This step is performed in the distribution area setting module. The distribution area setting module calculates the separation distance between data. The covariance matrix obtained in step S200 is used as basic data for calculating the separation distance between data. Here, the separation distance between data refers to the probabilistic separation distance from the center of the distribution of the sensed data. As an example, the sensing data may be displayed inside an oval shape on probability coordinates. Most of the sensing data is contained within a virtual ellipse having an elliptical long axis and a short axis, and the distribution area of the sensing data can be set in the virtual ellipse according to a preset weight. As an example, sensing data that falls outside the distribution area is determined to be an outlier.

4. Obtaining a prediction model using linear regression (S400);

This step is performed in the prediction model acquisition module. In this step, a linear regression equation can be created based on the set distribution area, and a prediction model can be obtained based on the linear regression equation. The linear regression function can be generated using the mean square error method to minimize the loss value.

5. Calibrating the input sensing signal using the obtained prediction model (S500);

This step is performed in the signal processing unit 200. Sensing signals detected from installed sensors are first input into the prediction model. The prediction model determines whether the sensing data is an outlier, and if not, corrects the sensing value of the sensing data according to the location of the distribution area in which the sensing data is preset. Correction of the sensing value may be performed in a manner that minimizes the error value of the sensing data.

Operations by the sensing data processing method according to the embodiments described above may be implemented as a computer program and recorded on a computer-readable recording medium. The computer-readable recording medium may include a recording device that stores data that can be read by a computer, and examples of the computer-readable recording medium may include ROM, RAM, and an optical data storage device. there is.

Above, the present invention has been described in detail through specific embodiments, but the present invention is not limited to the above embodiments, and various modifications can be made by those skilled in the art within the scope of the technical idea of the present invention.

100: Prediction model acquisition unit
110: Data conversion module
120: Covariance matrix acquisition module
130: Distribution area setting module
140: Prediction model acquisition module
200: signal processing unit

Claims (3)

A device for obtaining a prediction model from a plurality of previously detected sensing signals and correcting the standard setting value of the input sensing signal using the prediction model,
The device is,
It consists of a prediction model acquisition unit 100 and a signal processing unit 200,
The prediction model acquisition unit uses a data conversion module 110 to convert a sensing signal into a digital signal, a covariance matrix acquisition module 120 to obtain a covariance matrix based on the converted digital data, and the obtained covariance matrix. Thus, it is characterized in that it includes a distribution area setting unit 130 that calculates the separation distance between data and sets the data distribution area, and a prediction model acquisition unit 140 that obtains a prediction model based on the set distribution area. Device.
As a method for processing a signal detected from a sensor,
The method is:
A step of performing data conversion of the sensing signal (S100),
Obtaining a covariance matrix based on the converted sensing data (S200),
Calculating the separation distance between data and setting the distribution area (S300),
Obtaining a prediction model using linear regression (S400),
A method comprising the step of correcting the input sensing signal using the obtained prediction model (S500).
In claim 2,
A method characterized in that the separation distance between the data is a probabilistic separation distance away from the distribution center of the sensing data.

Images