JP2014176427A - Data analysis device and data analysis program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a data analysis device that realizes a precisely corresponding wavelet transformation for a biometric signal which varies from one targeted living body to another and has complicated constitutions, and to provide a program therefor.SOLUTION: A data analysis device includes: a derivation unit which executes autocorrelation processing on a biometric waveform detected from a living body to produce an autocorrelation waveform; and a generator which obtains a cut waveform by cutting out a part of the autocorrelation waveform, generates a basic wavelet waveform on the basis of the cut waveform, and then applies a wavelet transformation to the biometric waveform on the basis of the basic wavelet waveform.

Description

  The present invention relates to a data analysis apparatus and a data analysis program for analyzing feature quantities related to biological signals such as heartbeats and respiratory waveforms based on biological signal data related to body movements and heartbeats acquired from a living body.

  Conventionally, a technique for analyzing a heartbeat waveform included in electrocardiogram data by subjecting the electrocardiogram data of the subject to wavelet transform is known (Patent Document 1). If the wavelet transform is used, an accurate frequency analysis can be realized even when the analysis target data does not have complete periodicity as in the case where the electrocardiogram data includes an arrhythmia waveform.

  The electrocardiogram signal processing method described in Patent Document 1 amplifies and records electrocardiogram data. Then, the data of the section where the heartbeat waveform exists is cut out from the recorded electrocardiogram data, and wavelet transform is performed on the cut-out waveform. In this electrocardiogram signal processing method, the appearance time of the heartbeat waveform is detected based on the phase change as a result of the wavelet transform.

  In the wavelet transform, the basic wavelet serving as a conversion reference is scaled, and the homology between the wavelet waveform and the analysis target data is referenced while moving the scaled wavelet waveform in the time axis direction. According to the wavelet transform, it is possible to simultaneously analyze time information and frequency information of a wavelet waveform included in analysis target data.

  An advantage of wavelet transform is that time domain information is not lost when determining frequency characteristics. Therefore, the electrocardiogram signal processing method described in Patent Document 1 can accurately detect the appearance time of the heartbeat waveform included in the electrocardiogram data by using wavelet transform.

  In addition, as a prior art of this invention, the biological vibration frequency detection apparatus described in patent document 2 and the signal peak measurement system described in patent document 3 are also mentioned.

JP-A-9-28687 JP 2009-55997 A JP 2010-286268 A

  In the electrocardiogram signal processing method of Patent Document 1, a typical Gabor function is adopted as a basic wavelet. In addition, Mexican hat functions, American hat functions, French hat functions, modified Gaussians, Morlet functions, Daubechies functions, and the like are well known as functions that can be adopted as basic wavelets.

  As described above, the wavelet transform is a technique for analyzing the waveform component included in the analysis target data based on the homology with the basic wavelet. For this reason, the more similar the waveform component included in the analysis target data is with the shape of the basic wavelet, the more accurate the information extraction result by the wavelet transform.

  In general, the heartbeat waveform included in the heartbeat data is a complicated waveform as shown in FIG. The heartbeat waveform is composed of a complex waveform that combines a P wave indicating atrial excitement, a QRS wave indicating ventricular excitement, and a T wave indicating a state when the excitement has subsided.

  However, the shape of the well-known function generally adopted as a basic wavelet is not sufficiently similar to a heartbeat waveform having a complicated configuration. Therefore, it is difficult to say that the wavelet transform using such a well-known function is an optimal process for the purpose of accurately extracting the information of the heartbeat waveform included in the heartbeat data.

  In addition, the heartbeat waveform included in the heartbeat data not only changes in shape depending on the measurement (derivation) site, body movement, or posture, but also slightly differs depending on individual subjects from the existence of individual differences. Even if a specific waveform is suitable as a basic wavelet waveform for analyzing heart rate data of a subject, the specific waveform may not be suitable as a basic wavelet waveform for analyzing heart rate data of another subject. Absent.

  Therefore, as in Patent Document 1, when wavelet transform is performed using a uniform function as a basic wavelet, accurate information extraction is performed in response to a heartbeat waveform that is different for each subject and has a complicated configuration. It is difficult.

  The present invention has been made in order to solve the above-described problems, and an object of the present invention is to provide an accurate wavelet for a biological signal, for example, a heartbeat waveform, which is different for each target living body and has a complicated configuration. An object of the present invention is to provide a data analysis apparatus and a data analysis program that execute conversion and further accurately analyze a feature amount.

  In an aspect of the present invention, a derivation unit that performs autocorrelation processing on a biological waveform detected from a living body to create an autocorrelation waveform, cuts out a part of the autocorrelation waveform, acquires a cutout waveform, and Provided is a data analysis device that includes a basic wavelet waveform based on a waveform, and further includes a conversion unit that wavelet transforms the biological waveform based on the basic wavelet waveform.

  In an aspect of the present invention, a computer performs autocorrelation processing on a biological waveform detected from a living body to create an autocorrelation waveform, and cuts out a part of the autocorrelation waveform to obtain a cutout waveform. A data analysis program that generates a basic wavelet waveform based on the cut-out waveform and further functions as a conversion unit that performs wavelet transform on the biological waveform based on the basic wavelet waveform is provided.

  The data analysis apparatus according to the present invention can accurately analyze the feature amount of a biological signal that is in an unstable state depending on individual differences, measurement sites, or measurement conditions.

The block diagram which shows the structure of the data analyzer of 1st Embodiment Body surface displacement waveform analysis flowchart A flow chart summarizing the waveform processing flow of the data analyzer Graph showing examples of body surface displacement waveforms Graph showing filter passing waveform Graph showing an example of a cut-out waveform Graph showing basic wavelet waveform Graph showing wavelet transform waveform A graph showing the heartbeat waveform generated by the feature value analysis extraction unit of the wavelet analysis unit Graph showing examples of heart rate waveforms included in heart rate data

  Hereinafter, a data analysis apparatus according to a first embodiment will be described with reference to the accompanying drawings.

The data analysis apparatus according to the present embodiment is a data analysis apparatus that analyzes a biological waveform (biological signal) emitted from a living body, and in particular, only changes occur due to the body part, the body movement of the living body, or the posture of the living body. It is a data analysis device that analyzes a biological waveform (biological signal) that varies due to individual differences due to individual living bodies. Hereinafter, the data analysis device according to the present embodiment will be described as a data analysis device that analyzes heartbeat data of a subject as an example.

[1-1. Configuration of data analysis device]
With reference to FIG. 1, the structure of the data analysis apparatus of this embodiment is demonstrated. As shown in FIG. 1, the data analysis apparatus 1 includes a sensor 2, a control unit 3, a RAM 4, a ROM 5, an input control IF 6, an output control IF 7, a keyboard 8, a mouse 9, a monitor 10, and a bus 20.

  The sensor 2 is a sensor that detects the movement of the body surface of the subject using microwaves. The sensor 2 irradiates a subject's body surface (for example, a chest part where the heart is located) with a microwave having a predetermined frequency and intensity, and detects a change in the microwave reflected on the subject's body surface. The sensor 2 is a sensor that detects, for example, a frequency change, a phase change, or an intensity change of the reflected microwave. The sensor 2 detects the movement of the subject's body surface by detecting the amount of change in the frequency, phase, or intensity of the reflected microwave. The sensor 2 outputs the detected movement of the body surface as an analog displacement signal.

  The sensor 2 uses the microwave radar to pass through the subject's clothing and capture minute movements on the body surface accompanying the heartbeat. By using such a sensor 2, the data analysis device 1 can acquire heart rate data without touching the subject's body surface at all even during driving or sleeping. Thereby, the data analysis apparatus 1 can be suitably used for acquisition and analysis of heartbeat data in a daily environment.

  The control unit 3 is a control circuit that controls the overall control operation of the data analysis apparatus 1. The control unit 3 is realized by a microprocessor or the like. The control unit 3 executes the control operation of the data analysis device 1 based on the control program stored in the ROM 5.

  The RAM 4 is a random access memory that stores heartbeat data, various measurement data, calculation data, temporary calculation results, and the like under the control of the control unit 3.

  The ROM 5 is a read-only memory that stores the aforementioned control program.

  The input control IF 6 is an interface circuit that controls input signals from the sensor 2 and input signals from other input devices. A keyboard 8 and a mouse 9 are connected to the input control IF 6.

  The output control IF 7 is an interface circuit that controls the output of signals to an external device. A monitor 10 is connected to the output control IF 7.

  The keyboard 8 or the mouse 9 is an input device that accepts an instruction input from the user regarding, for example, acquisition, analysis, and display of heartbeat data.

  The monitor 10 is a display monitor that displays information such as heartbeat data, analysis data thereof, and a guidance message for the user.

  The bus 20 is a bus line that electrically connects each member inside the data analysis apparatus 1 and executes transmission and reception of signals.

  The control unit 3 includes an A / D conversion unit 11 and a wavelet analysis unit 12.

  The A / D converter 11 is a signal converter that converts an analog body surface displacement signal output from the sensor 2 into a digital body surface displacement waveform W1 (see FIG. 3) based on a predetermined sampling rate. This sampling rate is set to 1 kHz, for example. The A / D converter 11 outputs a digital body surface displacement waveform W1 to the wavelet analyzer 12.

  The wavelet analysis unit 12 is a functional block that extracts the feature quantity of the heartbeat waveform by performing wavelet analysis on the body surface displacement waveform W1 output from the A / D conversion unit 11. The wavelet analysis unit 12 includes a filter 13, an autocorrelation derivation unit 14, a wavelet transform unit 15, and a feature amount analysis extraction unit 16.

  The filter 13 is a digital filter that allows signals in a predetermined range to pass through the body surface displacement waveform W1 output from the A / D converter 12 and blocks other signals. For example, the filter 13 may be a band-pass filter that allows a signal in a specific frequency band in the body surface displacement waveform W1 to pass and blocks a signal in another frequency band. The filter 13 outputs the passed filter passage waveform W2 (see FIG. 3) to the autocorrelation deriving unit 14.

  The autocorrelation deriving unit 14 is a functional block that obtains an autocorrelation waveform W3 (see FIG. 3) from the filter passage waveform W2. Details of the process for obtaining the autocorrelation waveform W3 will be described later.

The wavelet transform unit 15 is a functional block that obtains a wavelet transform waveform W6 (see FIG. 3) from the autocorrelation waveform W3 obtained by the autocorrelation deriving unit 14.

  The feature amount analysis extraction unit 16 accurately extracts the feature amount of the heartbeat waveform based on the wavelet transform waveform W6 that is the conversion result by the wavelet transform unit 15. The feature amount of the heartbeat waveform includes the appearance time of the heartbeat waveform, the time interval between the heartbeat waveforms, the frequency of the heartbeat waveform, and the like.

  The feature amount analysis extraction unit 16 specifies, for example, an R wave of the heartbeat waveform from the extraction results of these feature amounts. The analysis result of the feature value of the wavelet transform result is time accurate information. Thereby, the data analyzer 1 can effectively detect the presence of arrhythmia and the like.

  The feature amount analysis extraction unit 16 newly generates a heartbeat waveform W7 based on the extraction result of the feature amount of the heartbeat waveform. By displaying this heartbeat waveform W7 on the monitor 10, the user can confirm the heartbeat waveform W7 as one of the final analysis results.

[1-2. Operation flow of data analyzer]
Next, the operation flow of the data analysis apparatus 1 for heart rate data according to the present embodiment will be described for each process using the flowchart of FIG.

  The sensor 2 irradiates the subject's body surface (for example, the chest part where the heart is located) with microwaves, detects a change in the reflected wave, and calculates the displacement of the subject's body surface from the detected change as an analog body surface displacement. Derived as a signal (S11). When irradiating the subject with microwaves from the sensor 2, it is preferable to arrange the sensor 2 so that the reflected waves received are stabilized.

  The A / D converter 12 converts the subject's body surface displacement derived in S11 and output as an analog signal from the analog signal to a digital signal (S12).

  The filter 13 passes, for example, only the region frequency (for example, 0.6 to 2 Hz) of the heartbeat waveform in the waveform W1 composed of the digital signal converted in S12 (S13).

  The body surface displacement signal obtained by using the sensor 2 includes many signals other than the body surface displacement caused by the heartbeat. Examples of the cause of displacement of the subject's body surface for reasons other than the heartbeat include rhythm of the body surface due to respiration.

  The movement of the body surface due to the heartbeat is about 0.1 to 0.2 mm, whereas the movement of the body surface due to breathing is about several centimeters. Therefore, the information on the body surface displacement due to the heartbeat is easily buried in the information on the body surface displacement due to respiration. In addition, when the subject is not stationary, body movement also appears as noise (artifact).

  In the above, 0.6 to 2 Hz is exemplified as the region frequency including information on the heartbeat waveform. However, when it is desired to reduce noise by narrowing down to a narrower frequency band, a person with a high heart rate or a sports player with a low heart rate. Assuming a wide range of subjects, the filter 13 may be set so as to pass only signals of 30 to 180 beats (0.5 to 3 Hz) per minute.

  By allowing the filter 13 to pass, noise caused by breathing and body movement other than heartbeat can be considerably removed from the body surface displacement waveform W1 output from the A / D conversion unit 12.

  The autocorrelation deriving unit 14 takes the autocorrelation of the filter passage waveform W2 that has passed through the filter 13, and creates an autocorrelation waveform W3 that becomes sample data of the heartbeat waveform (S14).

  Auto-correlation processing has a function of extracting random components from the original signal and reducing random noise. The heartbeat waveform included in the filter passing waveform W2 includes periodicity. Therefore, the autocorrelation waveform W3 obtained from the filter passing waveform W2 is obtained by extracting the heartbeat waveform component, and becomes favorable sample data of the heartbeat waveform.

  The data analysis apparatus 1 can always obtain accurate sample data of a heartbeat waveform even if the heartbeat waveform is complicated and is different for each subject. Due to the random noise reduction function of the autocorrelation process, the autocorrelation waveform W3 is not greatly affected by body movement or noise components.

When the autocorrelation derivation unit 14 obtains the autocorrelation waveform W3 from the filter passage waveform W2, assuming that the values of the filter passage waveform W2 in a predetermined section are f ₀ , f ₁ ,..., F _N , the autocorrelation signal ρ ( j) is expressed by the following equation (1).

In order to make the shape of the autocorrelation waveform W3 indicated by the autocorrelation signal ρ (j) a stable waveform, the digital signals f ₀ , f ₁ ,. The data length of f _n is preferably set to a period of several times to several tens of times the length of the assumed reference waveform.

In general, by setting the data length of the digital signals f ₀ , f ₁ ,..., F _n of the filtered waveform W2 to be subjected to autocorrelation processing to be long, the amount of data information is increased, and the autocorrelation signal It is possible to stabilize the waveform of ρ (j).

However, _if the data lengths of the digital signals f ₀ , f ₁ ,..., F _n of the filter passage waveform W2 to be subjected to autocorrelation processing are set too long, the influence of respiration and body movements having a frequency lower than that of the heartbeat may occur. It becomes stronger and the information accuracy of heartbeat deteriorates. If the data length is set too long, the data will be averaged over time, the wavelet transform response will be slow, and the heartbeat waveform will be difficult to capture in real time.

Therefore, the data length is not excessively long, and is based on, for example, digital signals f ₀ , f ₁ ,..., F _n for a period of several to several tens of times the length of the assumed reference waveform. When the autocorrelation waveform W3 indicated by the autocorrelation signal ρ (j) is created, the waveform can be stabilized.

  The autocorrelation waveform W3 indicated by the autocorrelation signal ρ (j) obtained by the equation 1 is a waveform that decays with time. The wavelet transform unit 15 performs processing to cut out a representative waveform portion from the autocorrelation waveform W3 to obtain a cut-out waveform W4 (see FIG. 3). Here, the representative waveform portion is, for example, a waveform portion from one maximum value peak to the next maximum value peak in the autocorrelation waveform W3.

  The wavelet transform unit 15 generates a basic wavelet waveform W5 (see FIG. 3) by interpolating or fitting a cut-out waveform W4 composed of discrete digital data to a given waveform function (see FIG. 3). S15).

  For example, the wavelet transform unit 15 prepares in advance a parameter function that combines waveforms simulating P waves, QRS waves, and T waves of the heartbeat waveform, and adjusts each parameter according to the cutout waveform W4. By such parameter adjustment, a basic wavelet waveform W5 corresponding to the cut-out waveform W4 can be generated.

  In this manner, the data analysis apparatus 1 generates an accurate basic wavelet waveform W5 having high similarity to the heartbeat waveform of the subject, and executes wavelet transform using this. That is, it is possible to generate accurate sample data as the basic wavelet waveform W5 even if the heartbeat waveform of the subject is complicated or is a unique heartbeat waveform different from that of a general subject.

  The wavelet transform unit 15 performs wavelet transform on the filtered waveform W2 based on the basic wavelet waveform W5 (S16).

  Thereby, the wavelet transform unit 15 obtains a wavelet transform waveform W6 (see FIG. 3). The wavelet transform waveform W6 accurately reflects the feature quantity of the heartbeat waveform.

  The feature amount analysis extraction unit 16 extracts the feature amount of the heartbeat waveform from the wavelet transform waveform W6 (S17). Specifically, the feature amount analysis extraction unit 16 detects the P wave based on the wavelet transform waveform W6, the interval between the Q wave and the T wave (QT interval), and the interval between the S wave and the T wave (ST interval). Measure the equipotential level of the baseline.

  The data analysis apparatus 1 can detect fine temporal changes and fluctuations in the waveform, and can be applied to diagnosis of heart rate variability index, arrhythmia and the like.

[1-3. Example of waveform processing]
FIG. 3 is a flowchart summarizing the flow of waveform processing (W1 to W7) of the data analysis apparatus 1 of the present embodiment.

  From the body surface displacement waveform W1 output from the A / D conversion unit 12, a filter passage waveform W2 subjected to the filter processing by the filter 13 is obtained.

  FIG. 4 is a graph showing an example of the body surface displacement waveform W1. As shown in the figure, the body surface displacement waveform W1 output from the A / D converter 12 has a complex waveform including high-frequency noise. The body surface displacement waveform W1 is greatly affected by respiration and body movement as described above, and effective analysis cannot be performed as it is.

  On the other hand, FIG. 5 is a graph showing a filter passage waveform W2 corresponding to the body surface displacement waveform W1 of FIG. As shown in the figure, high-frequency noise is removed and a relatively smooth waveform is obtained. In the filter passing waveform W2, the components derived from respiration and body movement are greatly reduced.

  The autocorrelation deriving unit 14 performs the above-described autocorrelation processing on the filter passing waveform W2 to obtain an autocorrelation waveform W3. The wavelet transform unit 15 obtains a cutout waveform W4 by cutting out a representative waveform portion from the autocorrelation waveform W3 (see FIG. 3).

  FIG. 6 is a graph showing an example of the cut-out waveform W4. As described above, the cutout waveform W4 is obtained by cutting out a typical waveform portion from the autocorrelation waveform W3 that decays with time, that is, a suitable waveform portion as sample data of the heartbeat waveform. The cut-out waveform W4 shows a different shape for each subject.

  The wavelet transform unit 15 generates a basic wavelet waveform W5 by interpolating or fitting a cutout waveform W4 composed of discrete digital data to a given waveform function (see FIG. 3).

  FIG. 7 is a graph showing the basic wavelet waveform W5. The basic wavelet waveform W5 can also be expressed by a continuous function, for example.

  The wavelet transform unit 15 performs wavelet transform on the filtered waveform W2 based on the basic wavelet waveform W5 to obtain a wavelet transform waveform W6 (see FIG. 3).

  FIG. 8 is a graph showing the wavelet transform waveform W6. The feature amount analysis extraction unit 16 performs feature amount analysis based on the wavelet transform waveform W6.

  The feature amount analysis extraction unit 16 generates a heartbeat waveform W7 included in the filter passage waveform W2 based on the result of the feature amount analysis.

  FIG. 9 is a graph showing an example of the heartbeat waveform W7 generated by the feature amount analysis extraction unit 16. It can be seen that the data analysis apparatus 1 reproduces a natural heartbeat waveform even though the heartbeat data is acquired without contact.

  In the graphs of FIGS. 4 to 8, the horizontal axis indicates time, but the absolute value of the horizontal axis does not have a meaning corresponding to real time, and is described as a mere point.

  According to the data analysis apparatus 1 described above, the wavelet transform that is different for each subject and has a complicated configuration and that accurately corresponds to the heartbeat waveform is executed, and the heartbeat waveform information included in the heartbeat data is analyzed more accurately. Can do.

  Note that the data analysis apparatus of the present embodiment may be realized by a computer. In this case, a data analysis program that causes the data analysis apparatus to be realized by the computer by operating the computer as the control unit is also included in the scope of the present invention.

  That is, in the above-described embodiment, each unit (each block) provided in the data analysis device 1, in particular, the control unit 3 may be realized by software using a processor such as a CPU. In this case, the data analysis apparatus 1 includes a CPU that executes instructions of a control program that implements each function, a ROM that stores the program, a RAM that expands the program, a storage device (recording medium) that stores the program and various data. ) Etc.

  An object of the present invention is to record a program code (execution format program, intermediate code program, source program) of a control program (data analysis program) of the data analysis apparatus 1 which is software that realizes the above-described functions in a computer-readable manner. This is also achieved by supplying a recording medium to the data analysis apparatus 1 and reading and executing the program code recorded on the recording medium by the computer (or CPU or MPU).

  Examples of the recording medium include tape systems such as magnetic tapes and cassette tapes, disk systems including magnetic disks such as hard disks and optical disks such as CD-ROM / MO / MD / DVD / CD-R, and IC cards (memory). A card system such as an optical card) or a semiconductor memory system such as a mask ROM / EPROM / EEPROM / flash ROM.

  When the data analysis apparatus 1 is configured to be connectable to a communication network, the program code may be supplied via the communication network.

  Each block of the data analysis apparatus 1 is not limited to that realized using software. Each block may be configured by hardware logic. Each block may be a combination of hardware that performs a part of the processing and arithmetic means that executes software that performs hardware control and residual processing.

  In addition, in the above, the data-analysis apparatus 1 which detects and analyzes the motion of a test subject's body surface by non-contact using the sensor 2 using a microwave was illustrated. However, the present invention is not limited to such non-contact analysis of heart rate data, and can also be applied to a data analysis apparatus that acquires heart rate data by contacting the subject's body.

  Furthermore, the present invention can be applied not only to a data analysis apparatus that detects and analyzes heartbeat data, but also to a data analysis apparatus that detects and analyzes a biological signal emitted from a living body.

1 Data Analysis Device 2 Sensor 3 Control Unit 4 RAM
5 ROM
6 Input control IF
7 Output control IF
8 Keyboard 9 Mouse 10 Monitor 11 A / D Converter 12 Wavelet Analyzer 13 Bandpass Filter 14 Autocorrelation Deriving Unit 15 Wavelet Transformer 16 Feature Analysis Extractor

Claims (6)

A derivation unit that creates an autocorrelation waveform by performing autocorrelation processing on a biological waveform detected from a living body;
A part of the autocorrelation waveform is cut out to obtain a cutout waveform, a basic wavelet waveform is generated based on the cutout waveform, and further, a conversion unit that wavelet transforms the biological waveform based on the basic wavelet waveform; Including data analysis equipment. The data analysis apparatus according to claim 1, wherein the conversion unit generates a basic wavelet waveform by interpolating the cut-out waveform or fitting to a predetermined waveform function. Furthermore, a sensor for detecting a biological waveform from a living body is included,
The data analysis apparatus according to claim 1, wherein the derivation unit performs autocorrelation processing on the biological waveform detected by the sensor to create an autocorrelation waveform. Furthermore, the data analysis apparatus of Claim 1 or 2 containing the analysis part which extracts the feature-value of the said biological waveform by which the said wavelet transform was carried out. The data analysis apparatus according to claim 4, wherein the feature quantity of the biological waveform extracted by the analysis unit is an appearance time and a time interval of the biological waveform. Computer
A derivation unit that creates an autocorrelation waveform by performing autocorrelation processing on a biological waveform detected from a living body;
A part of the autocorrelation waveform is cut out to obtain a cutout waveform, a basic wavelet waveform is generated based on the cutout waveform, and further, functions as a conversion unit that wavelet transforms the biological waveform based on the basic wavelet waveform Data analysis program

Images