KR101048763B1 - Apparauts and method for detecting signal - Google Patents

Abstract

PURPOSE: A signal detecting device and method are provided to detect a specific signal where a noise is eliminated by using a frequency domain and a time domain in a bio signal. CONSTITUTION: A signal detecting device includes a frequency domain processor(201) and a time domain processor(203). The frequency domain processor transforms inputted bio signals into frequency domain signals, and detects the frequency domain in which the special signal exists in the transformed frequency domain. The time domain processor transforms the detected frequency domain signals into the time domain signals, filters the time domain signals according to a noise frequency and detects a specific signal.

Description

Signal detection device and method {APPARAUTS AND METHOD FOR DETECTING SIGNAL}

The present invention relates to an apparatus and method for processing a biosignal, and more particularly, to an apparatus and method for detecting a specific signal in a biosignal.

The electrocardiogram (ECG) signal of the biosignal is a diagnostic signal representing the movement of the heart muscle by an electrode located in the body of the patient, and may be represented as shown in FIG. 1. In FIG. 1, an ECG of one period is composed of a P waveform and a T waveform including a relatively low frequency component, and a Q waveform, an R waveform, and an S waveform, which are referred to as a QRS composite wave including a high frequency component. The QRS complex wave occurs when the heart ventricle is depolarized, and the P and T waves occur in each case where the depolarization diffuses throughout the atrium and ventricular repolarization occurs.

In the analysis of the ECG signal, since further detailed investigations such as the ST segment, the PR segment, and the RR (from the R waveform to the R waveform) interval are based on the detection result of the QRS composite wave, the accurate detection of the QRS composite wave in the ECG signal is It is an essential step. However, ECG signals include a variety of noises, including physical movement, changes in electrode contact location, respiration, power line interference, and muscle contraction. Moreover, the spectrum of noise is in the wide frequency range (0-10 kHz). For example, baseline instability due to respiration includes the low frequency region (0-2 Hz), but power line interference and muscle contraction include the high frequency region (45-180 Hz). Accordingly, the QRS composite wave from which the noise has been efficiently removed from the ECG signal should be detected. Therefore, there is a need for a method of efficiently detecting a specific signal in a biological signal.

The present invention provides an apparatus and method for detecting a specific signal from which noise is effectively removed from a biological signal.

The present invention also provides an apparatus and method for detecting a specific signal from which noise is removed using a frequency domain and a time domain in a biological signal.

An apparatus for detecting a specific signal from an input biosignal according to the present invention includes: a frequency domain processor for converting the input biosignal into a frequency domain and detecting a frequency domain in which the specific signal exists in the converted frequency domain; And a time domain processor for converting the detected frequency domain into a time domain, and filtering the time domain according to a frequency characteristic of noise in the converted time domain to detect the specific signal.

In addition, the method for detecting a specific signal from the input bio-signal according to the present invention, converting the input bio-signal into a frequency domain, detecting a frequency domain in which the specific signal is present in the converted frequency domain, Converting the detected frequency domain to a time domain, and filtering the time domain according to the frequency characteristic of the noise in the converted time domain to detect the specific signal.

The present invention removes a frequency domain except a frequency domain in which a specific signal exists in a frequency domain of an input body signal, and then removes noises having low values in the time domain in consideration of the characteristics of the frequency, thereby reducing the sensitivity of the detected specific signal. It can greatly improve forecastability.

On the other hand various other effects will be disclosed directly or implicitly in the detailed description of the embodiments of the present invention to be described later.

1 is a view showing a general ECG signal,
2 is a diagram illustrating an apparatus for detecting a QRS composite wave according to an embodiment of the present invention;
3 is a diagram illustrating a frequency domain processing process according to an embodiment of the present invention;
4 is a view showing a visual area processing process according to an embodiment of the present invention;
5 illustrates a method for detecting a QRS composite wave according to an embodiment of the present invention.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings. It should be noted that in the following description, only parts necessary for understanding the operation according to the present invention will be described, and descriptions of other parts will be omitted so as not to distract from the gist of the present invention.

The main aspect of the present invention is to convert the biological signal into the frequency domain and to detect only the frequency domain in which the specific signal desired to be detected in the converted frequency domain, and then convert the detected frequency domain to the time domain and the converted In this case, only the specific signal is detected by filtering the time domain in which the noise exists in the time domain.

To this end, an apparatus and method for detecting a specific signal in a biological signal according to an embodiment of the present invention will be described in detail.

Hereinafter, a method of detecting a QRS complex wave in an ECG signal will be described as an example, but the present invention is applicable to a method for detecting a specific signal in a biosignal other than an ECG signal.

2 shows an apparatus for detecting a QRS composite wave according to an embodiment of the present invention.

Referring to FIG. 2, the QRS composite wave detection apparatus includes a frequency domain processor 201 and a time domain processor 203.

The frequency domain processor 201 receives an ECG signal from a user, and then uses a fast fourier transform (FFT) or a biorthogonal spline wavelet transform (BSWT) to display a frequency at which the QRS composite wave exists in the frequency domain of the input ECG signal. Only area is detected.

Here, the BSWT is effective for filtering the remaining frequency domain except for the frequency domain where the QRS complex wave exists in the frequency domain of the ECG signal. In this embodiment, the frequency domain processor 201 preserves the frequency domain in which the QRS composite wave exists in the ECG signal input using the BSWT and filters the remaining frequency domain except the frequency domain in which the QRS composite wave exists. This will be described.

The frequency domain processor 201 converts an input ECG signal into a frequency domain using a BSWT. In the frequency domain of the transformed ECG signal, P waves, T waves, and QRS complex waves exist as shown in FIG. 1. Here, the P-wave, T-wave and QRS complex wave are present in the frequency range of 2 Hz to 45 Hz, the noise due to respiration in the ECG signal is present in the frequency range of 0 Hz to 2 Hz, and the noise due to power interference and muscle contraction is 45 Hz. In the frequency range of 180 Hz. Accordingly, the frequency domain processor 201 filters the frequency domain except for the frequency range of 2 Hz to 45 Hz in which the QRS composite wave is present using the BSWT as shown in Equation 1 below.

)는 아래 <수학식 1>를 이용하여 특정 레벨에서의 근사화 계수 및 특정 레벨 (

)까지의 세부 계수들(QRS 복합파의 경우

)의 합산으로서 분해된다.When the BSWT is used in the frequency domain processor 201, an ECG signal (

) Is an approximation coefficient at a specific level and a specific level (

Coefficients up to

Is decomposed as the sum of

는 레벨 6에서의 신호 근사화이며,

는 도 3에서와 같이 스케일링 함수와 웨이블릿(Wavelet) 함수 각각으로부터 획득된 신호이다. here,

Is a signal approximation at level 6,

3 is a signal obtained from each of the scaling function and the wavelet function.

)에서 0~2 Hz에서의 저주파 성분인

, 45~180 Hz에서의 고주파 성분인

의 합(즉,

+

)을 감산하여, ECG 신호의 주파수 영역 신호에서 QRS 복합파가 존재하는 2Hz 내지 45Hz의 주파수 영역만을 검출한다. 여기서 상기 검출된 2Hz 내지 45Hz의 주파수 영역은 QRS 복합파, P파, T파 및 움직임 인공 잡음을 모두 포함한다. 이에 주파수 영역 처리기(201)는 상기 검출된 2Hz 내지 45Hz의 주파수 영역에서 P파, T파 및 움직임 인공 잡음의 필터링을 위해, 상기 검출된 2Hz 내지 45Hz의 주파수 영역을 시간 영역 처리기(203)로 전달한다.Accordingly, the frequency domain processor 201 uses an ECG signal (Equation 1) and FIG.

) Is a low frequency component from 0 to 2 Hz

, High frequency component from 45 to 180 Hz

Sum of (i.e.

+

) Is detected to detect only the frequency range of 2 Hz to 45 Hz in which the QRS complex wave is present in the frequency domain signal of the ECG signal. Here, the detected frequency range of 2 Hz to 45 Hz includes all of the QRS composite wave, P wave, T wave, and motion artificial noise. Accordingly, the frequency domain processor 201 transfers the detected frequency range of 2Hz to 45Hz to the time domain processor 203 for filtering P wave, T wave, and motion artificial noise in the detected frequency range of 2Hz to 45Hz. do.

Then, the time domain processor 203 converts the frequency domain of 2Hz to 45Hz in which the QRS complex wave received from the frequency domain processor 201 is present into a time domain, and then converts the time domain as described below. Detects only QRS complex wave by filtering out noise. Where the noise in the transformed time domain is P wave, T wave and motion artificial noise.

The signal in the transformed time domain is shown in FIG. 4 (a). Since the noise in the histogram in the transformed time domain is generally lower than the amplitude of the QRS composite wave, as in FIG. 4 (a), the time domain processor 203 detects noises with low amplitude to detect only the QRS composite wave. The threshold value is set according to the frequency characteristic to filter signals existing in the time domain after the threshold value. At this time, since the boundary between the QRS compound wave and the noise is ambiguous as shown in FIG. 4 (a) and the frequency of the QRS compound wave is very low compared to the noise, if the interval interval is adjusted according to the amplitude of the histogram, FIG. The histogram modified as shown in (c) can be obtained. Here, the time interval in the increase function of FIG. 4 (b) may be expressed as Equation 2 below.

That is, through Equation 2, the time interval is set to an increasing function of i.

As such, the time domain processor 203 can raise the histogram in the high frequency region by allocating a signal in the time domain in which the QRS complex wave of FIG. 4 (a) exists at an interval proportional to the signal size of FIG. 4 (b). have. In this case, the histogram satisfies Equation 3 below.

The time domain processor 203 transforms the histogram into a gentle curve as shown in FIG. 4 (d) by applying a low pass filter, and then sets a time corresponding to the lowest frequency in the modified curve as a threshold value. Only the QRS complex wave is detected by filtering the time domain in which the noise located in the time domain above the set threshold is present.

In addition, a peak detector (not shown) for detecting a peak of the R wave may be configured after the QRS complex wave detection apparatus according to the exemplary embodiment of the present invention. The QRS composite wave detected according to the embodiment of the present invention is then used to detect the peak of the R wave. However, in the QRS complex wave detection apparatus according to an embodiment of the present invention, when a similar QRS complex wave generated by the cardiac tissue does not respond to any excitation during the period immediately after the normal QRS complex wave is generated in addition to the normal QRS complex wave, the peak detector may be Error peaks can be detected as normal peaks.

)을 설정한 후, 상기 검출된 QRS 복합파에서 R파의 피크가 상기 설정된

을 만족하면 정상 피크로 분류한다. 반면, 상기 피크 검출기는 설정된

의 반 구간 내에서 새로운 피크가 검출되면 이를 오류 피크로 분류하여 부정확한 피크 선택을 방지한다. 여기서, 상기

는 정상 피크가 발생될 때마다 갱신된다.Accordingly, in order to prevent peaks from being detected in the pseudo QRS complex wave, the peak detector in the system using the QRS complex wave detection apparatus according to the present invention is configured to generate an adaptive RR interval in consideration of the change in the previous RR interval in the user's heart rate.

), The peak of the R wave in the detected QRS complex wave is set to

If is satisfied, it is classified as a normal peak. On the other hand, the peak detector is set

If a new peak is detected within half of, it is classified as an error peak to prevent incorrect peak selection. Where

Is updated each time a normal peak occurs.

In addition, the peak detector uses an adaptive threshold value as shown in Equation 4 to solve the case where the amplitude of the R wave continuously fluctuates due to cardiac arrhythmias and the residual high amplitude noise peak is confused with the peak of the detected QRS complex wave. Set as above to detect the peak above the set adaptive threshold.

를 가정하면,

는 적응형 임계값이고, QRS 복합파의 피크의 IIR(Infinite Impulse Response) 필터링된 크기이며

은 잡음 피크 후보물(candidates)이다. 그리고 상기

는 새로운 피크가 발생될 때마다 갱신된다. 상기

는 ECG 신호의 변동성을 추적하며 고-진폭 잡음 피크와 오류 R 피크를 제거한다.In Equation 4 above

Assuming

Is an adaptive threshold, the Infinite Impulse Response (IIR) filtered magnitude of the peak of the QRS complex wave,

Are noise peak candidates. And said

Is updated each time a new peak occurs. remind

Tracks the variability of the ECG signal and removes high-amplitude noise peaks and error R peaks.

와

를 이용하여 정상 피크를 검출함으로써, R파의 피크 검출에서 발생되는 오류 확률을 감소시킨다.Thus, the peak detector can be used to detect the

Wow

By detecting the normal peak using, it reduces the probability of error occurring in the peak detection of the R wave.

5 shows a method for detecting a QRS composite wave according to an embodiment of the present invention.

Referring to FIG. 5, in operation 501, the frequency domain processor 201 receives an ECG signal from a user and then converts the input ECG signal into a frequency domain in operation 503. In step 505, the frequency domain processor 201 filters the remaining frequency domain except for the frequency domain in which the QRS composite wave exists using the BSWT in the frequency domain of the transformed ECG signal. Detect the area.

In addition, since the PRS, the T wave, and the motion artificial noise other than the QRS compound wave are included in the frequency domain in which the QRS compound wave detected in step 507 exists, the process proceeds to steps 509 and 511 to detect only the QRS compound wave. do.

In step 509, the time domain processor 203 converts the frequency domain in which the detected QRS composite wave exists into the time domain to filter noise in the frequency domain in which the detected QRS composite wave exists. In step 511, the time domain processor 203 converts the converted time domain into a histogram using an increment function having a time interval as shown in Equation 2, and uses the changed histogram to generate a noise distribution and a QRS complex wave. The time corresponding to the lowest frequency between distributions is set as a threshold. The time domain processor 203 filters the time domain in which the noise located in the time domain above the set threshold is present. In step 513, the time domain processor 203 detects only the QRS complex wave in the transformed time domain.

와

를 만족하는 경우 R파의 피크를 검출하고, 만족하지 않는 경우 피크를 오류 피크로 분류하여, R파의 피크 검출에서 발생되는 오류 확률을 감소시킨다.Although not shown in FIG. 5, the detected QRS complex wave is preset in the peak detector after step 513.

Wow

If is satisfied, the peak of the R wave is detected, and if not satisfied, the peak is classified as an error peak, thereby reducing the probability of error occurring in the peak detection of the R wave.

When the embodiment of the present invention is applied to the actual ECG signal, the following results can be obtained. The actual ECG signal uses a 48 half-hour fragment of two-channel hospital ECG records (“MLII” and “V5”) in the MIT-BIH arrhythmia database. And to compare the present invention and the prior art as shown in Table 1 below, as an example of the prior art, a method using a diyadic wavelet, the classification technique by the k-th nearest neighbor rule (Classification by K) -th nearest neighbors rule.

In Table 1 below, the performance measurement values comparing the present invention and the prior art are calculated by Equations 5 and 6 below.

는 민감도(sensitivity)이고,

는 예보성(positive predictivity)이다. 그리고

는 포지티브(positive)로 검출된 위치의 개수이고,

와

는 각각 오류 포지티브(false positive) 및 오류 네거티브(false negative) 검출된 위치들을 나타낸다. In <Equation 5> and <Equation 6>

Is the sensitivity,

Is positive predictivity. And

Is the number of positions detected as positive,

Wow

Denotes false positive and false negative detected locations, respectively.

,

<0.5%).As shown in Table 1 below, the present invention is over-performing the previous approach and shows excellent error detection performance (

,

<0.5%).

Tested algorithm

Proposed method 0.9937 0.9947 Dyadic wavelet 0.8417 0.9309 Classification by K-th nearest neighbors rule 0.9670 0.9609

Therefore, the present invention can greatly improve the sensitivity and the predictability by removing the out-of-band in the frequency domain and the low magnitude noise in the time domain.

Meanwhile, in the detailed description of the present invention, specific embodiments have been described, but various modifications are possible without departing from the scope of the present invention. Therefore, the scope of the present invention should not be limited to the described embodiments, but should be defined not only by the appended claims, but also by the equivalents of the claims.

Claims (12)

In the apparatus for detecting a specific signal from the input bio-signal,
A frequency domain processor for converting the input biosignal into a frequency domain and detecting a frequency domain in which the specific signal exists in the converted frequency domain;
And a time domain processor for converting the detected frequency domain into a time domain and filtering the time domain according to a frequency characteristic of noise in the converted time domain to detect the specific signal.
The method of claim 1, wherein the frequency domain processor,
A specific signal detection apparatus using one of the Fast Fourier Transform (FFT) and the Biorthogonal Spline Wavelet Transform (BSWT).
The method of claim 2, wherein the frequency domain processor,
And detecting the frequency domain in which the specific signal exists by filtering the remaining frequency domain except the frequency domain in which the specific signal exists using the BSWT in the converted frequency domain.
The method of claim 1, wherein the time domain processor,
Generate a histogram representing the frequency change of the magnitude of the signal amplitude over time in the transformed time domain using an increment function having a predetermined time interval, and threshold the time corresponding to the lowest frequency in the generated histogram. And setting the value to detect the specific signal by filtering the time domain after the set threshold value.
The method of claim 1,
And a peak detector for detecting a peak in the detected specific signal based on an interval between peaks in the biosignal and an amplitude threshold value of the peak.
6. The apparatus of claim 5, wherein each of the interval between peaks and the amplitude threshold of the peak is adaptively changed according to the interval between previous peaks and the amplitude threshold of the previous peak.
In the method for detecting a specific signal from the input bio-signal,
Converting the input biosignal into a frequency domain and detecting a frequency domain in which the specific signal exists in the converted frequency domain;
And converting the detected frequency domain into a time domain, and filtering the time domain according to a frequency characteristic of noise in the converted time domain to detect the specific signal.
The method of claim 7, wherein the detecting of the frequency domain comprises:
A specific signal detection method using one of the Fast Fourier Transform (FFT) and the Biorthogonal Spline Wavelet Transform (BSWT).
The method of claim 8, wherein the detecting of the frequency domain comprises:
And detecting the frequency domain in which the specific signal exists by filtering the remaining frequency domain except for the frequency domain in which the specific signal exists in the converted frequency domain by using the BSWT.
The method of claim 7, wherein the detecting of the specific signal comprises:
Converting the detected frequency domain into a time domain;
Generating a histogram representing a frequency change of a signal amplitude magnitude over time in the transformed time domain using an increment function having a predetermined time interval;
Setting a time corresponding to the lowest frequency in the generated histogram as a threshold;
And detecting the specific signal by filtering the time domain after the set threshold value.
The method of claim 7, wherein
Detecting a peak in the detected specific signal based on an interval between peaks in the biosignal and an amplitude threshold value of the peak.
12. The method of claim 11, wherein each of the interval between peaks and the amplitude threshold of the peaks is adaptively changed according to the interval between previous peaks and the amplitude threshold of the previous peak.

Images