TWI660579B - A method and a system for reducing interfering frequency of power line - Google Patents

Abstract

The invention discloses a method and a system for reducing the power line interference frequency. The method includes: A: acquiring an electrocardiogram signal. B: Convert the ECG signal into a digital signal. C: The digital signal is divided in units of samples per second. D: Analyze the signal segments per unit length, and obtain at least one power line interference frequency according to the analysis result of each of the signal segments. E: According to the at least one power line interference frequency, the digital signal is digitally filtered to obtain a digital signal processed by reducing the power line interference frequency. F: Convert the digital signal processed by reducing the power line interference frequency into an electrocardiogram signal processed by reducing the power line interference frequency. The system includes: an acquisition device, a conversion device, a processing device, and a filtering device.

Description

Method and system for reducing power line interference frequency

The present invention relates to a method and system for detecting and reducing power line interference frequency, and more particularly, to a method capable of effectively controlling the frequency resolution of the captured ECG and reducing the power line interference frequency of the ECG in the telemedicine mode. And system.

Electrocardiography (ECG / EKG) is a technique used to record the electrophysiological activity of cardiac tissue.It records the depolarization and repolarization of cardiac tissue through pairs of electrodes on the limb. (Repolarization) process and rhythm. Through this technology, abnormal heart rhythms and abnormal activities caused by tissue damage or blood stasis are detected.

In the past, when people needed to detect the state of heart health, they were treated and measured by professionals in hospitals. Nowadays, with the advancement of science and technology, in order to meet the needs and immediacy of health monitoring, many emerging hardware and services have begun to rise, and at the same time, the development of telemedicine has been promoted. In this case, because of the convenience and cost reduction, the demand for long-term physiological monitoring will also grow.

This trend also represents a significant increase in the amount of data. In the past, ECGs had to be interpreted by physicians or medical examiners. However, long-term monitoring data without the assistance of software would have to be interpreted by humans only when the amount of data increased significantly. A huge burden. Although there are some skills Surgery has gradually developed in the automatic interpretation of ECG features, but not necessarily applicable to the ECG data collected through telemedicine today. Because in the past, the composition of the database was relatively simple, usually only provided by the same hospital or medical unit, and there were medical examiners to control the quality and classification, such as the database under Physiot's Physiobank. In the telemedicine mode, users on the other side usually lack the assistance of professionals and measurement sites without other signal interference, so they are susceptible to many external interferences in data acquisition, and many methods of feature detection are also available. Easy to fail.

Therefore, in the past when processing disturbed data, due to the short monitoring time, the severity of the interference and the length of the impact time can be used to decide to discard the captured data and re-measure, or to recapture the captured data. After splicing, some detection methods are used to find the required characteristics of the captured signal. However, for long-term monitoring data, because of the large amount of data, the disturbed signal fragments may be scattered at many points in time. At this time, it is not only difficult to decide whether to discard and re-measure. The splicing out of the used fragments is also a heavy one. Heavy burden.

In the processing of ECG, common interferences include baseline wandering, power line interference, motion artifact, muscle tremor, and electrode contact noise. . For the above-mentioned several interferences, many related signal processing technologies have been developed so far.

Power line interference refers to the current interference caused by electromagnetic induction on the circuit or line of the measuring device during the signal acquisition process, which is affected by the surrounding AC power supply lines. Power line interference is typically characterized by a very strong fundamental frequency, which varies from country to country. Power line interference is an important issue in Biomedical signal processing. In the modern urban living mode, many measuring devices are electrified, and the test subjects are usually in the power line layout space with many devices. Signals are usually disturbed by power lines.

In the part dealing with power line interference, most of the conventional techniques analyze the captured signals with known interference frequencies. However, under the aforementioned telemedicine model, the data is bound to come from all over the world. Of course, the data from different parts of the world will be affected by different frequencies due to regional differences. Therefore, unless relying on additional data notes , Otherwise many detection methods are unavailable or ineffective. In addition, in actual measurement, there may still be other external factors that cause the power line interference to occupy a significant portion of the captured signal after being processed by a conventional signal capture device.

Therefore, before the raw data of the biomedical signal is analyzed, digital filtering is usually performed to minimize the interference caused by the power line to facilitate subsequent processing. The frequency of interference is related to the frequency of generator sets in various countries or regions. To determine the interference frequency of the signal, it is common practice to use Fast Fourier Transform (FFT) for frequency analysis. According to the Nyquist-Shannon sampling theorem, the effective frequency term analyzed by FFT is from 0 Hz to f _s ^{/ 2} Hz. However, under the condition that the frequency of interest is only 50Hz and 60Hz, it is extremely poor to use FFT to distinguish the frequency of power line interference.

On the other hand, because the power line is usually a single frequency interference, a notch filter is used for many circuit designs or signal processing. However, the disadvantage of the aforementioned notch filter is that it will also have a great impact on those signals whose frequency is close to the center frequency to be filtered out, and it will also have a severe ringing effect on higher frequency signals.

Therefore, adaptive analysis methods have been proposed in the conventional technology to deal with power line interference. However, the adaptive algorithm requires good initial conditions, and the filtering effect is also affected by its convergence speed. In order to avoid the aforementioned ringing effect on the signal and Taking into account the processing speed of the signal, some people have proposed to determine the phase of power line interference through a linear criterion, and then use a subtraction procedure to eliminate the power line interference.

In short, the above methods are based on the known power line interference frequency. If the power line interference frequency cannot be known beforehand, the relevant processing parameters cannot be determined.

Therefore, in order to overcome the aforementioned problems, the present invention has emerged.

In order to solve the foregoing problem and achieve the power line interference frequency can be effectively reduced even if the power line interference frequency is unknown, the present invention provides a method for reducing the power line interference frequency. The method includes: A: acquiring an ECG signal; B: converting the ECG signal Into a digital signal; C: divide the digital signal by the number of samples per second as a unit; D: analyze the signal fragments per unit length, and based on the analysis results of each of these signal fragments to obtain at least A power line interference frequency; E: digitally filtering the digital signal according to the at least one power line interference frequency to obtain a digital signal processed by reducing the power line interference frequency; F: converting the digital signal processed by reducing the power line interference frequency into ECG signal processed after reducing power line interference frequency.

In one embodiment, in step C, if the signal segment is less than one unit in the end, the signal pair is made up to one unit or discarded.

In an embodiment, in step C, the division is performed by using a sampling frequency of 1000 Hz as a unit.

In one embodiment, in step D, a Goertzel algorithm is performed.

In one embodiment, in step D, the target frequencies analyzed are 50HZ and 60HZ.

The present invention further provides a system for reducing the power line interference frequency. The system includes: an acquisition device for acquiring an electrocardiogram signal; a conversion device for converting an electrocardiogram signal to a digital signal; and a processing device, which After the digital signal is divided by the number of samples per second as a unit, the signal segment of each unit length is analyzed, and at least one power line interference frequency is obtained according to the analysis result; a filtering device is provided according to The at least one power line interference frequency is subjected to digital filtering of the digital signal to obtain a digital signal processed by reducing the power line interference frequency.

In an embodiment, if the last signal segment is less than one unit, the processing device complements the signal segment to a unit or discards the signal segment.

In one embodiment, the processing device divides the sampling frequency in units of 1000 Hz.

In one embodiment, the processing device executes a Goertzel algorithm.

In one embodiment, the target frequencies analyzed by the processing device are 50HZ and 60HZ.

In order to have a deeper understanding of the features and functions of the present invention, the embodiments are described in detail below with reference to the drawings.

1‧‧‧ capture device

2‧‧‧ Conversion device

3‧‧‧Processing device

4‧‧‧ Filtering device

A, B, C, D, E, F‧‧‧ steps

FIG. 1 is a schematic block diagram of a system for reducing a power line interference frequency according to the present invention.

FIG. 2 is a flowchart of a method for reducing the power line interference frequency according to the present invention.

Figure 3 shows an example of power line interference.

Fig. 4 is a diagram illustrating an example in which the power line frequency changes with the power supply frequency.

Figure 5a shows the distribution of power line interference on the frequency spectrum under a large-scale observation (viewing range: 0 ~ 100Hz).

Figure 5b shows the distribution of power line interference on the frequency spectrum under a small-scale observation (viewing range: 56 ~ 640Hz).

Figures 6a, 6b, and 7a, 7b are performance analysis charts of different algorithms.

The invention discloses a system for reducing the power line interference frequency. The system includes an acquisition device 1, a conversion device 2, a processing device 3, and a filtering device 4, wherein the acquisition device 1 is for acquiring an electrocardiogram signal; The conversion device 2 is used to convert ECG signals and digital signals; the processing device 3 is used to divide the digital signal based on the number of samples per second as a unit, analyze the signal segment of each unit length, and according to the analysis As a result, at least one power line interference frequency is obtained; the filtering device 4 is configured to perform digital filtering on the digital signal according to the at least one power line interference frequency to obtain a digital signal processed by reducing the power line interference frequency.

The present invention also discloses a method for reducing the power line interference frequency. Please refer to FIG. 2, which includes: A: capturing an electrocardiogram signal; B: converting the electrocardiogram signal into a digital signal by the conversion device 2; C: converting the digital signal Divide by the number of samples per second as the unit; D: The processing device 3 analyzes the signal segments per unit length, and The analysis result of each to obtain at least one power line interference frequency; E: according to the at least one power line interference frequency, digitally filtering the digital signal with the filtering device 4 to obtain a digital signal processed by reducing the power line interference frequency ; F: The digital signal processed by reducing the power line interference frequency is converted by the conversion device 2 into an electrocardiogram signal processed by reducing the power line interference frequency.

The method and system of the present invention will be described in detail below. In step A, an ECG signal is first captured by the capture device 1. The capture device may be a wearable ECG device, a 12-lead automatic ECG device, Palm-type ECG equipment, Bluetooth transmission ECG acquisition equipment, wireless ECG detection equipment, etc.

Then, in step B, the ECG signal is converted into a digital signal by the conversion device 2. The conversion method of the present invention is described as follows: First, in this step, the present invention first uses a Discrete Fourier Transform (DFT) The ECG signal is converted into projections on a variety of different frequency bases. If the length of the ECG signal x [n] is N, the DFT result X [k] is expressed by mathematical formula:

Then, rewrite Equation 1 into the form of Convolution:

Furthermore, as can be seen from the above method 2, DFT is the signal x [n] and a linear time-invariant (LTI) filter Convolution. If this LTI filter is rewritten as a z-transform, it is

According to Equation 3, the transfer function of this filter is

Therefore, the relationship between the input signal x [n] and the output signal yk [n] can be expressed as

Using Equation 5, we can calculate the k-th frequency component of a signal x [n] of length N

In formula 6, the formula on the right side of the equal sign is the same as the form of DFT (see formula 1, but the difference is that the total upper bound of formula 6 is N). This means that when using this form of algorithm, the signal must have N + 1 points, but in fact the signal has only N points. Therefore, the extra data needed must be supplemented by itself. However, the data cannot be changed arbitrarily, and the points that are randomly added are likely to cause analysis errors because they differ too much from the actual signal boundary values. Therefore, in the following, the present invention rewrites Equation 4 with Euler's equation as follows:

After that, the chain rule is used to introduce a temporary variable S (z) to re-express the relationship between input and output.

From Equation 8, we know

From formula 9-2 can know that, for a length N of the signal x [n], required to obtain y _k [N], then the operator needs to s _k [N] so far. 9-1 seen from the formula, the desire to obtain s _k [N], the need to count x [N]. Comparing the conclusion of the above two formulas with that of Formula 6, no additional supplementary points are needed for calculation. It should also be noted that although x [ n -1] and x [ n -2] are undefined when n = 0, we can set it to 0 to make the filter response to zero at this time. Response (Zero state response).

In step C, the digital signal is divided by using the number of samples per second as a unit. When the signal is actually processed, if the conventional analysis method is used to distinguish the frequency of power line interference, the following problems will be encountered: First, it is the problem of frequency drift. Because the frequency of the power line is not fixed, it usually occurs. With the imbalance between supply and demand between the power supply of the power plant and the external power consumption changes. When the supply of electricity is greater than the demand, the frequency of power supply becomes higher; otherwise it becomes lower. The range of drift varies according to the regulations of various countries and the efficiency of the generator set, usually within ± 0.5Hz. Second, the problem is that the frequency resolution is too high. Both the conventional analysis method and the FFT frequency resolution are affected by the signal length. As shown in Equation 1, if the signal length N is larger, the frequency base represented therein is divided into more items, and the frequency resolution is higher. Due to the problem of frequency drift, conventional analysis methods (such as the Goertzel algorithm) can only calculate one DFT term at a time. If you specify 50Hz and 60Hz as the frequency to be analyzed by the analysis method, the calculated value will most likely not Actual power line interference components. As shown in Figure 5a, under a large-scale observation, the interference frequency of the power line is about 60Hz; but when further inspection is performed, as shown in Figure 5b, it is found that the distribution of the strongest interference frequency of the power line is actually not accurate At 60Hz.

In addition, the number of DFT terms to be obtained will affect the length of the calculation time. Within a fixed range, the smaller the distance between DFT items, the more DFT terms are required, and the distance between DFT items is the frequency resolution of DFT; and because most of today ’s physiological signal acquisition equipment In terms of sampling frequency, the range of frequency drift is usually not large. According to the recommendations of the American Heart Association (AHA), ECG signal measurement equipment with monitoring as its main purpose should have a sampling frequency above 150 Hz. Therefore, In terms of a frequency drift range of ± 0.5 Hz, the error caused by the sampling frequency at 150 Hz is about 0.3%. Therefore, the key of the present invention in this step is to reduce the frequency resolution first and then perform subsequent analysis steps. In other words, as long as the frequency resolution can be reduced, the time spent in subsequent analysis steps can be reduced. However, directly downsampling the signal is easily affected by aliasing. Therefore, in order to overcome the above problems, the present invention further adjusts the length of the analyzed signal. In step C, the processing unit 3 divides the digital signal by the number of samples per second as a unit, and the processing unit 3 divides by a sampling frequency of 1000 Hz. If the signal segment is less than one unit in the end, Then the signal pair is made up to a unit or discarded. Furthermore, since the processing unit 3 is divided by a sampling frequency of 1000 Hz, the frequency resolution of each signal spectrum is fixed at 1 Hz, and because this resolution completely covers the aforementioned frequency drift range (± 0.5 Hz) Therefore, the results of subsequent analysis are not affected by frequency drift.

Further, in step D, the processing device 3 uses a Goertzel algorithm to analyze signal segments per unit length, and obtains at least one power line interference frequency according to the analysis result of each of the signal segments. As mentioned above, since the processing unit 3 divides by using a sampling frequency of 1000 Hz, the frequency resolution of each signal spectrum will be fixed at 1 Hz, and because this resolution completely covers the aforementioned frequency drift range ( ± 0.5Hz), so this step The analysis result of step D is not affected by the aforementioned frequency drift, and the Goertzel algorithm only needs to analyze the target frequency (50 Hz and 60 Hz in this embodiment). In another embodiment, in step D, the processing device 3 may also analyze multiple points at the target frequency of the Goertzel algorithm (in this embodiment, around 50Hz and 60Hz), and then use the maximum value of the individual area as For analysis results.

The following is the Goertzel algorithm virtual code in step D

(Pseudocode):

1 // --- definition ---

2 sig: signal series

3 n: signal length

4 fs: sampling frequency

5 ft: target frequency

6

7 k = (int) (0.5 + n * ft / fs)

8 omega = 2 * pi * k / N

9 cosine = cos (omega)

10 sine = sin (omega)

11 coeff = 2 * cosine

12 s0, s1, s2 = 0 // temporary space for saving previous data

13

14 for i from 0 to n

15 s0 = coeff * s1-s2 + sig [i]

16 s2 = s1

17 s1 = s0

18

19 real = (s1-s2 * cosine)

20 imag = s2 * sine

21 mag = sqrt (real * real + imag * imag) // result.

In terms of computational complexity, given the K term DFT term to be obtained, for a signal of length N, according to Equation 9, the complexity of the Goertzel algorithm is O ( KNM ), where M represents Equation 9 -1 The amount of calculations per iteration. The FFT must be determined by the algorithm used. If a 2-base-fast Fourier transform (radix-2 FFT) is used, the computational complexity of the FFT is O ( MN log ₂ N ).

The calculation complexity of the Goertzel algorithm and the radix-2 FFT is compared without considering the M term described in the previous paragraph. If the Goertzel algorithm is more efficient than the FFT, the following conditions must be met

Therefore, according to Equation 10, in this step D, as long as the number of DFT terms to be obtained when using the Goertzel algorithm is less than the logarithm of the signal length, the calculation efficiency will be better than using FFT directly.

Then, in step E, the digital signal is digitally filtered by a filtering device such as a notch filter according to the at least one power line interference frequency, so as to obtain a digital signal processed by reducing the power line interference frequency. Finally, in step F, the digital signal processed by reducing the power line interference frequency is converted into an electrocardiogram signal processed by reducing the power line interference frequency by the conversion device 2 and inversely calculated by the formula 8 and the formula 9.

In addition, due to the differences in the specifications of other power plants or other factors, the range of frequency drift may still exceed the ± 0.5Hz mentioned above. Therefore, in one embodiment, the present invention makes corresponding adjustments to this exceptional situation as follows: the resolution in step C is maintained at 1 Hz, but the frequency to be analyzed is expanded outward with the target frequencies (50 Hz and 60 Hz) as the center Then, a plurality of points are selected, and finally the selected plurality of points are compared with each other, or, in another embodiment, the resolution in step C is maintained at 1 Hz, and the target frequency to be analyzed is set as an intensity threshold of power line interference. After the step C is completed, if one of the two target frequencies is greater than the intensity threshold, the detection is deemed to be successful. If the foregoing conditions are not met, the frequency resolution is reduced again and the analysis is performed again.

Effectiveness analysis of the method and system of the present invention

The following are the results of the performance analysis of the present invention in step C and step D by the processing unit 3 using a short-term Goertzel algorithm (the abbreviation of step C and step D). Know the Goertzel algorithm, fast Fourier transform (FFT), short-time version of fast Fourier transform (FFT-st / STFT for time-frequency analysis).

The number of data points used in the performance analysis is 1233600, which is divided into 100 equal parts, and the analysis frequency has two values of 50Hz and 60Hz. For each additional amount of data, five calculations are performed repeatedly to calculate the average time-consuming. The data were tested in integer and floating-point numbers, and the analysis results were shown in Figure 6a, Figure 6b, Figure 7a, and Figure 7b. Among them, stft, goertzel-st and goertzel-st-m represent short-term versions of the algorithm, while goertzel-m and goertzel-st-m represent versions of the algorithm that can input multiple target frequencies at one time.

For the processing of the two data types, the short-term version of the FFT processing speed is indeed about 5 times faster than the original FFT; and the four Goerztel algorithms are all nearly 2 times faster than the short-term version of the FFT. From the perspective of analyzing the power line interference at 50Hz or 60Hz, when the short-term Goertzel algorithm of the present invention adjusts the frequency resolution to 1, as long as the two points of 50Hz and 60Hz are analyzed, the conventional Goertzel algorithm requires The points analyzed will increase as the data length increases. Therefore, for detecting the power line interference frequency, using the short-term Goertzel algorithm of the present invention still has better performance.

Therefore, the present invention has the following advantages:

1. The method and system of the present invention can also accurately detect the frequency of power line interference when using a filter for filtering.

2. In the treatment of power line interference, in conjunction with the development of the telemedicine mode, the present invention effectively overcomes the uncertainty of interference frequencies from around the world. The advantage of the Goertzel algorithm that can control the frequency resolution is compared with the disadvantage of the conventional Goertzel algorithm being limited by the influence of the data length. The short-term Goertzel algorithm of the present invention can significantly improve the analysis performance.

3. The present invention effectively improves the conventional Goertzel algorithm and the short-time Fourier transform conventional technique, and provides a more efficient improvement of the Goertzel algorithm, that is, the short-term Goertzel algorithm.

The above are the specific embodiments of the present invention and the technical means used. According to the disclosure or teaching of this article, many changes and modifications can be derived. If the equivalent changes made according to the concept of the present invention, the resulting When the effect does not exceed the substantial spirit covered by the description and drawings, it should be regarded as falling within the technical scope of the present invention.

According to the content disclosed above, the present invention can indeed achieve the intended purpose of the invention, and provide a method and system for reducing the frequency of power line interference. It has industrial and practical value, so it is necessary to file an invention patent application according to law.

Claims (8)

A method for reducing power line interference frequency, the method includes: A: capturing an electrocardiogram signal; B: converting the electrocardiogram signal into a digital signal; C: dividing the digital signal by using the number of samples per second as a unit; D: Analyze signal segments per unit length, and obtain at least one power line interference frequency according to the analysis result of each of the signal segments; E: digitally filter the digital signal according to the at least one power line interference frequency, A digital signal processed by reducing the power line interference frequency is obtained; F: converting the digital signal processed by reducing the power line interference frequency into an electrocardiogram signal processed by reducing the power line interference frequency; wherein in step C, if the last signal segment is less than one Unit, then make up the signal pair to a unit or discard. The method for reducing the power line interference frequency as described in item 1 of the scope of the patent application, wherein in step C, the division is performed by using a sampling frequency of 1000 Hz as a unit. The method for reducing the power line interference frequency as described in item 1 of the scope of the patent application, wherein in step D, a Goertzel algorithm is performed. The method for reducing the power line interference frequency as described in item 1 or 3 of the scope of patent application, wherein in step D, the target frequencies analyzed are 50HZ and 60HZ. A system for reducing power line interference frequency, the system includes: an acquisition device for acquiring an electrocardiogram signal; a conversion device for converting an electrocardiogram signal and a digital signal to each other; and a processing device for providing the After the digital signal is divided by the number of samples per second as a unit, the signal segment per unit length is analyzed, and at least one power line interference frequency is obtained according to the analysis result. If the last signal segment is less than one unit, the The processing device makes up the signal pair to a unit or discards it; a filtering device for digitally filtering the digital signal according to the at least one power line interference frequency to obtain a digital signal processed by reducing the power line interference frequency. The system according to item 5 of the patent application range, wherein the processing device is divided by a sampling frequency of 1000 Hz. According to the system described in claim 5 of the patent application scope, the processing device executes a Goertzel algorithm. According to the system described in claim 5 or 7, the target frequencies analyzed by the processing device are 50HZ and 60HZ.