Electrocardiosignal processing method and apparatus, storage medium, and program product

Abstract

A method of processing an electrocardiosignal capable of accurately obtaining heartbeat information, a device for processing an electrocardiosignal, a computer-readable storage medium, and a computer program product. The method of processing an electrocardiosignal includes: a detection signal acquisition step of acquiring a detection signal by detecting an electrocardiosignal of a living body by a sensor; a signal preprocessing step of dividing the detection signal into a plurality of segmented signals with a prescribed time length, calculating an average value of a plurality of signals included in each of the segmented signals, and obtaining a preprocessed signal by subtracting the average value of the segmented signal from each signal sample point included in the segmented signal; and a self-correlation calculation step of calculating a self-correlation function of the preprocessed signal.

Description

TECHNICAL FIELD

[0001] The present application relates to an electrocardiosignal processing method, an electrocardiosignal processing apparatus, a computer readable storage medium, and a computer program product. BACKGROUND

[0002] In the past, there has been a method of acquiring an electrocardiosignal of a living body by a plurality of sensors provided on some wearable devices or devices (for example, seats of a car or the like) that are in indirect contact with a human body to acquire an electrocardiosignal of a living body through clothes.

[0003] However, since the electrodes of the sensors (for example, capacitive non-contact sensors) are not in direct contact with the skin of the living body, they are easily disturbed by various noises in use. For example, in the case where a plurality of capacitive non-contact sensors are provided on the seats of a car, since the sitting posture of a person differs, sometimes some of the plurality of sensors are not in contact with the body or are in a state of poor contact, and sometimes the mutual friction among the seat, clothes, and the living body also causes electrostatic disturbance. For the sensors that are not in contact with the body or are in a state of poor contact, the detection signal obtained by the electrodes can only contain noise (not contain any electrocardiosignal), or can contain different kinds and degrees of noise in the electrocardiosignal.

[0004] In view of the above, it is preferable to judge the noise degree of the signal before analyzing and processing the detection signal, and to perform subsequent processing only when the signal-to-noise ratio (S / N or SNR) is high, so that a more correct electrocardio analysis result can be obtained. Since there are cases where the sensor is working while performing analysis and processing of the detection signal, in this case, the S / N of the signal needs to be judged in real time. Generally, the S / N of the signal is estimated by judging the periodicity of the signal, so how to detect the periodicity of the signal is very important. As a method of detecting the periodicity of the electrocardiosignal, there is a method of calculating an autocorrelation function. In the case where there is a significant peak in the calculation result of the autocorrelation function, it is considered that the signal is periodic, and the position of the peak coincides with the signal period.

[0005] However, it is found in the calculation process of the autocorrelation function that since some detection signals can only contain noise (not contain any electrocardiosignal) or contain different kinds and degrees of noise in the electrocardiosignal, there is a problem that there is no significant peak in the calculation result, or although there is a significant peak, the position of the peak does not coincide with the signal period.

[0006] Further, the signal data stream from the sensor is generally processed once every predetermined time (for example, every 1 second), and the signal length for one processing needs to be limited to within several seconds (for example, the signal within the last three seconds is processed each time). The signal length is limited so that the calculation program can also run on a low-performance embedded processor, thereby increasing the range of application of the processing method. If the signal length for one processing is longer, the processing time also becomes longer, and the memory requirement also becomes larger. Further, in a case where the electrodes of the sensor are not fixed to the body, the occurrence rate of noise is high. The frequency of a person's heartbeat is generally between 0.5 Hz and 3 Hz, and the possibility that the frequency of noise is higher than the heartbeat frequency is also high. In fact, there are often cases where a signal with a short signal time length and noise with a frequency higher than the signal frequency is processed.

[0007] In processing an electrocardiogram signal with a short signal time length and noise with a frequency higher than the signal frequency, the position of the significant peak in the calculation result of the autocorrelation function deviates from the period of the electrocardiogram signal, and even coincides with the period of the noise, so that the periodicity cannot be accurately determined. Thus, there is a problem that the signal with a large noise cannot be accurately filtered based on the periodicity of the signal, and accurate heartbeat information cannot be obtained. SUMMARY

[0008] In view of the above problems, the present application provides an electrocardiogram signal processing method, an electrocardiogram signal processing device, a computer readable storage medium, and a computer program product, which can accurately obtain heartbeat information.

[0009] An electrocardiogram signal processing method, characterized by comprising: a detection signal acquisition step of acquiring a detection signal by detecting an electrocardiogram signal of a living body by a sensor; a signal preprocessing step of dividing the detection signal into a plurality of segmented signals with a predetermined time length, calculating an average value of a plurality of signals included in each of the segmented signals, and obtaining a preprocessed signal by subtracting the average value of the segmented signal from each signal sample point included in the segmented signal; and an autocorrelation calculation step of calculating an autocorrelation function of the preprocessed signal.

[0010] According to the electrocardiogram signal processing method, by performing the average value subtraction processing for the segmented signal before calculating the autocorrelation function, the tendency of the calculation result of the autocorrelation function to gradually decay is reduced, and the wave peak of the electrocardiogram signal is highlighted, so that the periodicity of the electrocardiogram signal can be accurately determined. Further, accurate heartbeat information can be obtained by screening out the signal with a large noise that does not have periodicity.

[0011] In the above electrocardiogram signal processing method, when the detection signal is a continuous signal x(t), the segmented signal is y(t), the average value is the preprocessed signal is and the autocorrelation function is Rss The formula for calculating (τ) is as follows:

[0012] R ss (τ)=∫0 s(t)s(t+τ)dt

[0013] Where t is the integration variable, τ is the delay, and T is the time length of the preprocessed signal s(t).

[0014] Therefore, this ECG signal processing method can be applied to the calculation of autocorrelation of continuous signals.

[0015] In the above electrocardiogram signal processing method, the detected signal is a discrete signal x[n] collected at a fixed frequency fs, the segmented signal is y[n], and the average value is... At that time, the preprocessed signal is The autocorrelation function is R ss The formula for calculating [m] is as follows:

[0016]

[0017] Where n is the summation variable, m is the translation amount, L is the signal length of the preprocessed signal s[n], and the value range of n in s[n] is 0≤n≤L-1.

[0018] Therefore, this ECG signal processing method can be applied to the calculation of autocorrelation of discrete signals. Furthermore, L in the above formula conforms to the following relationship: L = fs·T, where T is the time length of the segmented signal y[n].

[0019] The above-mentioned ECG signal processing method further includes: a signal periodicity judgment step, which, after the autocorrelation calculation step, searches for the highest peak in the autocorrelation coefficient waveform corresponding to each segmented signal, and calculates the difference between the highest peak and other adjacent peaks. The difference represents the intensity comparison between the signal and noise. When the difference is less than a first predetermined value, the segmented signal is identified as noise and discarded. The autocorrelation coefficient is the value obtained after normalizing the autocorrelation function.

[0020] The above-mentioned ECG signal processing method further includes: a signal periodicity judgment step, which, after the autocorrelation calculation step, searches for the highest peak in the autocorrelation coefficient waveform corresponding to each segmented signal, and calculates the difference between the highest peak and other adjacent peaks. The difference represents the intensity comparison between the signal and noise. When the difference is greater than or equal to a first predetermined value, the segmented signal is considered valid. The autocorrelation coefficient is the value obtained after normalizing the autocorrelation function.

[0021] According to the electrocardiosignal processing method, the highest peak is compared with the rest of the peaks for each segmented signal corresponding autocorrelation coefficient waveform, and the difference between the peak value of the noise and the peak value of the electrocardiosignal is used to determine the reliability of each segmented signal. The greater the difference, the smaller the noise, and the smaller the difference, the greater the noise. By discarding the segmented signal with large noise, the segmented signal can be screened. Further, by using the segmented signal with high reliability to perform subsequent heartbeat information extraction and other processing, the accuracy of the processing result of the electrocardiosignal is improved.

[0022] In the above electrocardiosignal processing method, the first predetermined value is set to 0.1. The first predetermined value is set in advance based on the difference between the peak value of the noise and the peak value of the electrocardiosignal.

[0023] In the above electrocardiosignal processing method, in the signal periodicity determination step, the period Ts of the preprocessed signal corresponding to each autocorrelation coefficient waveform is calculated according to the position of the highest peak in the waveform. The period Ts of the preprocessed signal is the same as the period Ty of the segmented signal.

[0024] Therefore, the period Ts of the preprocessed signal and the period Ty of the segmented signal can be determined by the position of the highest peak in the autocorrelation coefficient waveform.

[0025] In the above electrocardiosignal processing method, it is determined whether the period Ty of the segmented signal calculated in the signal periodicity determination step satisfies the condition 0.3s

[0026] According to the electrocardiosignal processing method, the possible period range 0.3-1.5s of the heartbeat signal can be used to further exclude noise, thereby further improving the reliability of the segmented signal.

[0027] In the above electrocardiosignal processing method, in the signal periodicity determination step, the sample variance S of the periods Ty of a plurality of consecutive segmented signals is calculated, and when S<0.005 2 , the segmented signal is identified as noise and discarded.

[0028] According to the electrocardiosignal processing method, the characteristics of the system noise with fixed frequency can be used to further exclude noise, thereby further improving the reliability of the segmented signal.

[0029] In the above electrocardiosignal processing method, the specified time length T is set according to the possible period range of human heartbeat, and the specified time length T is at least the length of two or more heartbeat periods T R .

[0030] According to the electrocardiosignal processing method, the heartbeat signal with at least two periods in each segmented signal is set, so that the highest peak value in the autocorrelation coefficient waveform obtained after calculation is consistent with the period of the electrocardiosignal.

[0031] In the electrocardiosignal processing method, the specified time length T is an arbitrary value satisfying 3s≤T≤5s.

[0032] According to the electrocardiosignal processing method, in general cases, considering that the heartbeat period of a person is less than 1.5s, the specified time length T is set to 3s. Even if the very extreme case is considered, the heartbeat period of a person will not exceed 2.5s, so the specified time length T is set to 5s at most.

[0033] In the electrocardiosignal processing method, in the autocorrelation calculation step, the wave peaks contained in the preprocessed signal s[n] are extracted, and the interval between the adjacent two wave peaks in the extracted wave peaks is greater than or equal to a specified time interval L s , thereby obtaining a signal S d ; converting the signal S d into a square wave , calculating the autocorrelation function of the square wave , and obtaining the autocorrelation coefficient waveform of the square wave .

[0034] According to the electrocardiosignal processing method, compared with the case without converting the square wave, the calculation amount, calculation time and memory consumption of the autocorrelation calculation step are greatly reduced, and the characteristics related to periodicity of the autocorrelation function R ss [m] are almost completely retained in the autocorrelation coefficient waveform of the square wave , so that the periodicity of the electrocardiosignal can be more quickly judged without reducing the accuracy.

[0035] In the electrocardiosignal processing method, the width L of the square wave p satisfies L ε <L p <L s / 2, where L ε is a predetermined value (empirical value) greater than 0 and greater than the maximum period difference ΔT R , and the maximum period difference ΔT R is the maximum difference that can occur between the heartbeat periods T R within the specified time length T. In addition, the specified time interval L s is the minimum heartbeat period within the specified time length T.

[0036] If the width L p is less than or equal to the predetermined value L ε, then it is possible that no matter how the square wave is shifted, all the peaks in the square wave cannot simultaneously satisfy the requirement of coinciding with each other within the specified time length T, which results in a decrease in the approximation between the autocorrelation function of the square wave and the autocorrelation function R ss [m] of the preprocessed signal s[n]. On the other hand, if the width L p is greater than or equal to L s / 2, then one square wave in the square wave after shifting can simultaneously overlap with more than two square waves in the square wave , which results in a complicated square wave search and an increase in the computational cost.

[0037] According to the electrocardiosignal processing method, since the upper limit and the lower limit of the width L R are set based on the value related to the heartbeat period T p , the decrease in the approximation between the autocorrelation function of the square wave and the autocorrelation function R ss [m] of the preprocessed signal s[n] due to the too narrow width L p , and the complicated square wave search and the increase in the computational cost due to the width L p being greater than or equal to L s / 2 can be avoided.

[0038] In the electrocardiosignal processing method, the width L p is any value in the range of 0.06s p <0.1s.

[0039] According to experience, the predetermined value L ε is generally 0.05s, and the minimum heartbeat period (the specified time interval L s ) of a human being is generally 0.3s. If the specified time interval L s is greater than 0.3s, then it is possible that some peaks are missed in the process of peak extraction. Therefore, 0.06s p <0.1s is a recommended value based on experience and with a certain margin.

[0040] In the electrocardiosignal processing method, the calculation of the autocorrelation function includes the following steps: (1) shifting the square wave to obtain a square wave after shifting; and (2) searching for the square wave and the square wave (3) calculating the product sum of the overlapping portions R1 to Rr using the following equation;

[0041]

[0042] where R is the number of the overlapping portions R1 to Rr, v ir v jr is the height of each of the overlapping portions R1 to Rr, and t irjr is the width of each of the overlapping portions R1 to Rr.

[0043] An electrocardiosignal processing device characterized by comprising: a detection signal acquisition unit that acquires a detection signal by detecting an electrocardiosignal of a living body with a sensor; a signal preprocessing unit that divides the detection signal into a plurality of segment signals in a predetermined time length, calculates an average value of a plurality of signals included in each of the segment signals, and subtracts the average value of each segment signal from each signal sample point included in the segment signal to obtain a preprocessed signal; and an autocorrelation calculation unit that calculates an autocorrelation function of the preprocessed signal.

[0044] A computer-readable storage medium having stored thereon a computer program, wherein the computer program, when executed by a processor, implements the steps of the electrocardiosignal processing method.

[0045] A computer program product comprising a computer program, wherein the computer program, when executed by a processor, implements the steps of the electrocardiosignal processing method. BRIEF DESCRIPTION OF DRAWINGS

[0046] Figure 1 is a diagram showing the configuration of the electrocardiosignal processing device of the present application.

[0047] Figure 2 is a flowchart showing the electrocardiosignal processing method of the first embodiment.

[0048] Figure 3 is an example of a segment signal y[n] and a preprocessed signal s[n] in the first embodiment.

[0049] Figure 4 is Figure 3 an example of the autocorrelation coefficient waveforms respectively corresponding to the segment signal y[n] and the preprocessed signal s[n] in

[0050] Figure 5 is a diagram showing the detailed steps of the signal periodicity judgment step in the first embodiment.

[0051] Figure 6is a view showing detailed steps of the signal periodicity judging step of the modified example 2 of the first embodiment.

[0052] Figure 7 is a view showing detailed steps of the autocorrelation calculating step in the second embodiment.

[0053] Figure 8 is a view showing a signal Sd obtained by extracting a peak included in a preprocessed signal s[n].

[0054] Figure 9 is a view showing a signal Sd obtained by extracting a peak included in a preprocessed signal s[n].

[0055] Figure 10 is a view showing v ir , v jr , t irjr corresponding to an overlapping portion of the square wave.

[0056] Figure 11 is a view showing a waveform of an autocorrelation coefficient p ss [m] in the first embodiment and a waveform of an autocorrelation coefficient in the second embodiment. DETAILED DESCRIPTION

[0057] (First Embodiment)

[0058] The electrocardiosignal processing device 1 in the first embodiment is included in a vehicle-mounted system not shown, for example. Such a vehicle-mounted system has at least a function of obtaining an electrocardiosignal of a driver or a passenger by a plurality of capacitive non-contact sensors provided on seats of a vehicle and processing the electrocardiosignal.

[0059] Figure 1 is a view showing a configuration of the electrocardiosignal processing device 1 of the present application. The electrocardiosignal processing device 1 includes a detection signal obtaining unit 10, a signal preprocessing unit 20, an autocorrelation calculating unit 30, a signal periodicity judging unit 40, and a heartbeat information extracting unit 50. In addition, in the present embodiment, the electrocardiosignal processing device 1 further has a heartbeat information output unit 60 for displaying heartbeat information.

[0060] The detection signal obtaining unit 10 obtains a detection signal by detecting an electrocardiosignal of a living body by a plurality of sensors provided on seats of a vehicle. It is generally desired to know heartbeat information and the like of a driver or a passenger in real time, and therefore each sensor constantly obtains an electrocardiosignal as a detection signal at a prescribed interval (for example, at an interval of 0.1 seconds) while detecting the presence of a driver or a passenger.

[0061] ​The signal preprocessing unit 20 preprocesses the detection signal before performing the autocorrelation function calculation. The signal preprocessing unit 20 divides the acquired detection signal into a plurality of segment signals in a prescribed time length, calculates the average value of the plurality of signals included in each segment signal, and subtracts the average value of each segment signal from each signal sample point included in the segment signal to obtain a preprocessed signal.

[0062] The autocorrelation calculation unit 30 calculates the autocorrelation function of the preprocessed signal obtained after the processing by the signal preprocessing unit 20. After calculating the autocorrelation function, the autocorrelation function is normalized, and the autocorrelation coefficient waveform corresponding to each segment signal is obtained.

[0063] The signal periodicity judgment unit 40 searches for the highest peak in the autocorrelation coefficient waveform corresponding to each segment signal, calculates the difference between the highest peak and the other peaks adjacent to the highest peak, and judges the segment signal as noise and discards the segment signal when the difference is less than a first predetermined value, and judges the segment signal as valid when the difference is greater than or equal to the first predetermined value.

[0064] Further, the signal periodicity judgment unit 40 can calculate the period of the preprocessed signal corresponding to each autocorrelation coefficient waveform from the position of the highest peak in the autocorrelation coefficient waveform. The period of each preprocessed signal is the same as the period of the corresponding segment signal. Thus, it is possible to judge whether the segment signal is valid or not based on the period of the preprocessed signal.

[0065] The heartbeat information extraction unit 50 extracts heartbeat information using the segment signal judged as valid by the signal periodicity judgment unit 40. The heartbeat information includes, for example, the number of heartbeats, the heartbeat period, the heartbeat timing, and the like. The method of extracting the heartbeat information is not limited, and for example, the findpeak function can be used to search for a peak in the segment signal, the time point of the peak corresponds to the R-wave peak timing in the heartbeat signal, and the time interval between adjacent peaks corresponds to the heartbeat period.

[0066] The heartbeat information output unit 60 is, for example, a liquid crystal display, a head-up display (HUD), or the like provided in a vehicle, and displays the heartbeat information extracted by the heartbeat information extraction unit 50.

[0067] Figure 2 is a flowchart showing the electrocardiosignal processing method of the first embodiment.

[0068] In step S1 (detection signal acquisition step), the detection signal acquisition unit 10 acquires a detection signal by detecting an electrocardiosignal of a living body by a sensor. The detection signal in the present embodiment is a discrete signal x[n] acquired at a fixed frequency fs. Further, the detection signal can be a continuous signal x(t) other than the discrete signal.

[0069] In step S2 (signal preprocessing step), the signal preprocessing unit 20 divides the detected signal into multiple segment signals over a predetermined time length T, calculates the average value of the multiple signals contained in each segment signal, and subtracts the average value of the segment signal from each signal sample point in the segment signal to obtain the preprocessed signal. The predetermined time length T is set based on the possible cycle range of the human heartbeat, and the predetermined time length T includes at least two heartbeat cycles T. R The length. This is because if a segmented signal contains a heartbeat period T... R If the number of segments is less than 2, then the segmented signal itself is not periodic. The specified time length T is preferably any value satisfying 3s ≤ T ≤ 5s.

[0070] Given a segmented signal y[n], the average value is... At that time, the preprocessed signal is Figure 3 This is an example representing a segmented signal y[n] and a preprocessed signal s[n] in the first embodiment. Figure 3 The vertical axis represents the signal value, and the horizontal axis represents the signal acquisition time. Here, the duration of the segmented signal is 3 seconds. For example... Figure 3 As shown, after preprocessing, the waveform of the segmented signal y[n] shifts downwards as a whole. The amount of this quantity, thus a portion of the values ​​in the preprocessed signal s[n] become negative values ​​less than 0. Furthermore, Figure 3 The segmented signal shown contains noise; specifically, it includes three peak P values ​​caused by the electrocardiogram signal. R1 ~P R3 And 8 peak values ​​P caused by noise N1 ~P N8 Here, the frequency of the noise is higher than the frequency of the electrocardiogram signal.

[0071] Next, in step S3 (autocorrelation calculation step), the autocorrelation calculation unit 30 calculates the autocorrelation function of the preprocessed signal.

[0072] The autocorrelation function is R. ss The formula for calculating [m] is as follows:

[0073]

[0074] Where n is the summation variable, m is the shift amount, and L is the signal length of the preprocessed signal s[n], where the value of n in s[n] is 0 ≤ n ≤ L-1. Furthermore, L in the above formula conforms to the following relationship: L = fs·T, where T is the time length of the segmented signal y[n]. For example, with a fixed frequency fs of 560Hz and T of 3 seconds, the signal length L = 560 × 3 = 1680. That is, each preprocessed signal s[n] contains 1680 discrete points.

[0075] In calculating the autocorrelation function R ss After [m], the autocorrelation coefficient waveform corresponding to each segment of the signal can be obtained.

[0076] Autocorrelation coefficient ρ ss The above autocorrelation function R ss [m] is obtained after normalization. Here, the normalization process is applied to R. ss [m] divided by R ss Treatment of [0]. Autocorrelation coefficient ρ ss The range of values ​​for is -1 ≤ ρ ss ≤1.

[0077]

[0078] Figure 4 yes Figure 3 Examples of autocorrelation coefficient waveforms corresponding to the segmented signal y[n] and the preprocessed signal s[n] are given. Figure 4 In the diagram, the vertical axis represents the value of the autocorrelation coefficient, and the horizontal axis represents the translation amount m. The autocorrelation coefficient ρ corresponding to the piecewise signal y[n] is... yy The waveform of [m] is located at Figure 4 The autocorrelation coefficient ρ corresponding to the preprocessed signal s[n] is located at the top of the image. ss The waveform of [m] is located at Figure 4 It is located in the lower part of the middle.

[0079] When calculating segmented signals with short signal lengths and noise frequencies higher than the ECG signal frequency, the autocorrelation coefficient waveform exhibits a decreasing trend. Therefore, the higher the signal frequency, the further forward the corresponding peak in the autocorrelation coefficient waveform is located (i.e.,...). Figure 4 (On the left side of the graph), the lower the frequency of the signal, the further back the peak position of the autocorrelation coefficient waveform corresponding to that signal is (i.e., on the left). Figure 4 (The right side of the middle).

[0080] The decreasing trend in the autocorrelation coefficient waveform is related to the properties of the autocorrelation function itself and the length of the target signal. Since the autocorrelation function is the cross-correlation of a signal with itself at different time points, the function value is the integral of the overlap between the signal and its shifted (delayed) signal. As the signal delay increases, the overlap between the signal and its shifted (delayed) signal gradually decreases, and its integral value also gradually decreases. Therefore, when processing signals with short lengths, the autocorrelation coefficient waveform will inevitably exhibit a decreasing trend. Figure 4 As shown in the diagram, the downward trend is more pronounced, especially for shorter signals.

[0081] for Figure 3 Regarding the segmented signal y[n] shown in the figure, in Figure 4 The autocorrelation coefficient ρ at the top of the table yy In the waveform of [m], since the frequency of noise is higher than the frequency of the ECG signal, it is related to P. N1 P N2 The corresponding peak P RyN1 P RyN2 Located on the front side of this waveform, with P R1 The corresponding peak P Ry1 Located in the waveform, compared to P RyN1 P RyN2 Further down the spectrum. In other words, due to the decreasing trend of the autocorrelation coefficient and the influence of signal frequency, it is related to P... N1 P N2 The corresponding peak P RyN1 P RyN2 Become more than P R1 The corresponding peak P Ry1 High. This situation will negatively impact the subsequent judgment of signal periodicity.

[0082] On the other hand, for Figure 3 Regarding the preprocessed signal s[n] shown in the figure, in Figure 4 The lower part of the autocorrelation coefficient ρ ss In the waveform of [m], since the frequency of noise is higher than the frequency of the ECG signal, it is related to P. N1 P N2 The corresponding peak P RsN1 P RsN2 Located on the front side of this waveform, with P R1 The corresponding peak P Rs1 Located in the waveform, compared to P RsN1 P RsN2 Further down the spectrum. But with respect to the autocorrelation coefficient ρ ss Unlike the waveform of [m], even under the decreasing trend of the autocorrelation coefficient and the influence of signal frequency, it is similar to P. N1 PN2 The corresponding peak P RsN1 P RsN2 There is no comparison with P R1 The corresponding peak P Rs1 high.

[0083] Next, in step S4 (signal periodicity judgment step), the signal periodicity judgment unit 40 judges whether each segment signal has periodicity, and determines whether the segment signal is valid or invalid based on the judgment result.

[0084] Figure 5 The detailed steps for determining the periodicity of a signal are shown in the figure.

[0085] In step S41, the highest peak is searched in the autocorrelation coefficient waveform of each segmented signal, and the difference between the highest peak and other adjacent peaks is calculated. Here, "the difference between the highest peak and other adjacent peaks" can be the minimum difference between the highest peak and one or more adjacent peaks, or the minimum difference between the highest peak and one or more adjacent peaks within a certain interval. Furthermore, the difference between the highest peak and other adjacent peaks represents the intensity comparison between the signal and noise; a larger difference indicates less noise, and a smaller difference indicates more noise, or even that only noise signals are present.

[0086] In step S42, it is determined whether the difference is less than a first predetermined value. In order to remove noisy segmented signals, it is preferable to set the first predetermined value to 0.1.

[0087] When the difference is greater than or equal to the first predetermined value (S42, Yes), proceed to step S43, determine that the segmented signal has periodicity, and recognize the segmented signal as valid.

[0088] If the difference is less than the first predetermined value (S42, No), proceed to step S44, where it is determined that the segmented signal does not have periodicity, and the segmented signal is identified as noise and discarded. Research indicates that after calculation for signals consisting only of noise or signals with high noise levels, the difference will be less than the first predetermined value.

[0089] Below, with Figure 4 The autocorrelation coefficient ρ shown in the figure ss The steps for determining the periodicity of a signal are illustrated using the waveform of [m] as an example.

[0090] In step S41, the signal periodicity determination unit 40 determines the autocorrelation coefficient ρ. ss Peak P was found in the waveform of [m]. Rs1This is the highest peak in the waveform, and the minimum difference Ws between this highest peak and other peaks within a certain interval (e.g., a 1-second interval centered on the highest peak) is calculated. Figure 4 In the middle, with peak P Rs1 The 1-second interval centered on the wave contains the peak P. RsN1 Peak P RsN2 Peak P RsN3 Peak P RsN4 The values ​​corresponding to these four other peaks are: Peak P Rs1 =0.50, peak P RsN1 =0.20, peak P RsN2 =0.19, peak P RsN3 =0.17, peak P RsN4 =0.15. Calculations show that the minimum difference Ws between the highest peak and other peaks within a certain adjacent interval (i.e., Figure 4 Medium wave peak P Rs1 With P RsN1 The difference between the values ​​is 0.3. Since this difference Ws is greater than or equal to the first predetermined value (0.3 > 0.1 in step S42), the segmented signal is deemed valid (step S43).

[0091] Return to Figure 2 Next, in step S5 (heartbeat information extraction step), the heartbeat information extraction unit 50 uses the segmented signal determined as valid by the signal periodicity judgment unit 40 to extract heartbeat information. Heartbeat information includes, for example, the number of heartbeats, the heartbeat period, and the peak time of the R-wave in the heartbeat signal. The method for extracting heartbeat information is not limited; for example, the findpeak function can be used to search for peaks in the segmented signal, where the peak time corresponds to the peak time of the R-wave in the heartbeat signal, and the time interval between adjacent peaks corresponds to the heartbeat period T. R .

[0092] Next, in step S6 (heartbeat information output step), the heartbeat information output unit 60 displays the extracted heartbeat information through a liquid crystal display, head-up display or the like installed in the vehicle.

[0093] Furthermore, the heartbeat information output step may not be required for ECG signal processing methods. The heartbeat information extracted by the heartbeat information extraction step can also be used in conjunction with other functional modules. For example, the heartbeat information can be used as a monitoring indicator for an abnormal condition monitoring module, or as one of the input data for dynamic health assessment.

[0094] According to the electrocardiogram signal processing apparatus and method of the first embodiment, the average value is subtracted from the segmented signal y[n] before calculating the autocorrelation function. the processing of the detection signal (i.e., the preprocessed signal ), reduces the tendency of the calculation result of the autocorrelation function to gradually decay, highlights the peaks of the electrocardiogram signal, and thus enables accurate determination of the periodicity of the electrocardiogram signal. Furthermore, accurate heartbeat information can be obtained by eliminating signals with large noise that do not have periodicity.

[0095] In addition, the highest peak is compared with the remaining peak values for each segmented waveform, and the reliability of each segmented signal is determined using the difference between the peak values of the noise and the peak values of the electrocardiogram signal. The greater the difference, the smaller the noise, and the smaller the difference, the greater the noise. By discarding segmented signals with large noise, the segmented signals can be screened. Furthermore, by using segmented signals with high reliability for subsequent processing such as extraction of heartbeat information, the accuracy of the processing result of the electrocardiogram signal is improved.

[0096] (Variation 1 of the first embodiment)

[0097] Variation 1 of the first embodiment is a variation of the first embodiment, and the only difference between the first embodiment and Variation 1 is the processing of step S2 in Figure 2 , and the other steps are basically the same as those of the first embodiment. Only the difference will be described below.

[0098] In the electrocardiogram signal processing method of the first embodiment, the detection signal is a discrete signal, and in Variation 1 of the first embodiment, the detection signal is a continuous signal.

[0099] In step S2 of Figure 2 , when the detection signal is a continuous signal x(t), the segmented signal is y(t), and the average value of the segmented signal y(t) is , the preprocessed signal is At this time, the autocorrelation function R ss (τ) is calculated as follows,

[0100] R ss (τ) = ∫0 T s(t)s(t+τ)dt

[0101] where t is the integral variable, τ is the delay, and T is the time length of the preprocessed signal s(t).

[0102] (Variation 2 of the first embodiment)

[0103] Variation 2 of the first embodiment is a variation of the first embodiment, and the only difference between the first embodiment and Variation 2 is the processing of step S4 in Figure 2 , and the other steps are basically the same as those of the first embodiment. Only the difference will be described below.

[0104] Figure 6The detailed steps of the signal periodicity judging step (step S4) of the first embodiment of the modified example 2 are shown in FIG. 4. As shown in FIG. 4, the signal periodicity judging step of the modified example 2 includes steps S41-S44 shown in FIG. 3, and further includes steps S45-S48 for judging whether the signal is valid. Figure 6 Figure 5 The detailed steps of the signal periodicity judging step (step S4) of the first embodiment of the modified example 2 are shown in FIG. 4. As shown in FIG. 4, the signal periodicity judging step of the modified example 2 includes steps S41-S44 shown in FIG. 3, and further includes steps S45-S48 for judging whether the signal is valid.

[0105] In step S45, the signal periodicity judging unit 40 calculates the period Ts of the preprocessed signal corresponding to each autocorrelation coefficient waveform based on the position of the highest peak in the waveform. The period Ts of the preprocessed signal is the same as the period Ty of the segmented signal corresponding to the preprocessed signal. In more detail, the position of the highest peak in each autocorrelation coefficient waveform represents the period Ts of the preprocessed signal and the period Ty of the segmented signal corresponding to the preprocessed signal.

[0106] Based on this step, the period Ts of the preprocessed signal and the period Ty of the segmented signal can be determined based on the position of the highest peak in the autocorrelation coefficient waveform.

[0107] Next, in step S46, it is judged whether the period Ty of the segmented signal calculated in step S45 satisfies the condition 0.3s < Ty < 1.5s. The segmented signal satisfying the condition is recognized as valid, and the segmented signal not satisfying the condition is recognized as noise and discarded.

[0108] Based on this step, noise can be further excluded using the possible period 0.3-1.5s of the heartbeat signal, and thus the reliability of the segmented signal can be further improved.

[0109] Next, in step S47, the sample variance S of the periods Ty of a plurality of continuous segmented signals is calculated.

[0110] Next, in step S48, the segmented signal is recognized as valid if S < 0.005 2 , and the segmented signal is recognized as noise and discarded if S > 0.005 2 .

[0111] Generally, the sample variance of the period of the fixed frequency system noise is smaller than the sample variance of the heartbeat period. If the sample variance of the period is smaller than 0.005 2 , it does not conform to the characteristics of human heartbeat, and the sample variance of the heartbeat period must be greater than or equal to 0.005 2 . Thus, this characteristic of the fixed frequency system noise can be used to further exclude noise, and thus the reliability of the segmented signal can be further improved.

[0112] (Second Embodiment)

[0113] ​The second embodiment is a variation of the first embodiment and its modifications, and the only difference between it and the first embodiment and its modifications 1 and 2 is that... Figure 2 The processing of step S3 (autocorrelation calculation step) is described below. Figure 7 Explain the differences.

[0114] Figure 7 This is a diagram showing the detailed steps of the autocorrelation calculation in the second embodiment.

[0115] In step S31, the autocorrelation calculation unit 30 extracts the peaks contained in the preprocessed signal s[n], and ensures that the interval between two adjacent peaks in the extracted peaks is greater than or equal to a predetermined time interval L. s Thus, signal S is obtained. d . Figure 8 This indicates that the signal S is obtained after extracting the peaks contained in the preprocessed signal s[n]. d A schematic diagram. Figure 8 The vertical axis represents the signal value, and the horizontal axis represents the signal acquisition time. The solid dots in the figure represent the extracted peaks, and the individual solid dots constitute the signal S. d .

[0116] Next, in step S32, the autocorrelation calculation unit 30 calculates the above signal S d Convert to square wave Figure 9 This means that the above signal S d Convert to square wave A schematic diagram. Figure 9 The vertical axis represents the signal value, and the horizontal axis represents the time when the signal was acquired.

[0117] The square wave width L p Satisfy L ε <L p <L s / 2, where L ε It is greater than 0 and greater than the maximum period difference ΔT R The predetermined value, the maximum period difference ΔT R For the multiple heartbeat cycles T contained within the specified time length T R The maximum possible difference between them. For example, a segmented signal preprocessing signal s[n] with a specified time length T of 3 seconds contains 3 heartbeat cycles T. R1 ~T R3 T R1 = 0.98 seconds, T R2 = 0.96 seconds, T R3 = 1.00 seconds, then the maximum period difference ΔT RThe time is 0.04 seconds. Based on experience, the predetermined value L... ε It is typically 0.05 seconds.

[0118] In addition, the specified time interval L s It is the minimum heartbeat cycle within a specified time period T. Based on experience, the minimum human heartbeat cycle is generally 0.3 seconds. If the specified time interval L... s If the time exceeds 0.3 seconds, some peaks may be missed during peak extraction. Therefore, the amplitude L... p The preferred value is 0.06s. <L p Any value within 0.1s. That is, 0.06s. <L p <0.1s is a recommended value based on experience and taking into account a certain margin.

[0119] Next, in steps S33 to S35, the above square wave is calculated. autocorrelation function

[0120] In step S33, the autocorrelation calculation unit 30 calculates the square wave After translation, the square wave is obtained.

[0121] In step S34, the square wave is searched at the same time. With the translated back wave The overlapping portion of the square wave is R1~Rr.

[0122] In step S35, the sum of the products of the overlapping portions R1 to Rr is calculated using the following formula.

[0123]

[0124] Where R is the number of overlapping square wave portions R1 to Rr, v ir v jr It is the height of the overlapping portion R1~Rr of each wave, t irjr R1 to Rr is the width of the overlapping portion of each wave.

[0125] Figure 10 It represents the v corresponding to the overlapping part of the square wave. ir v jr , t irjr A schematic diagram.

[0126] Assumption It consists of N square waves p i Composed of [n], it can be written as like Figure 10 As shown, p i [n] is The i-th square wave in the equation, square wave pj [n+m] is the jth square wave in i [n] and square wave p j [n+m] is also a square wave p i [n]p j [n+m] is a square wave p i [n]p j [n+m] has a width of t irjr , i.e., square wave p i [n] and square wave p j [n+m] overlap, and its height is v ir v jr , i.e., square wave p i [n] has a height of v ir , and square wave p j [n+m] has a height of v jr .

[0127] After calculating the autocorrelation function , the autocorrelation coefficient waveform corresponding to each segmented signal can be obtained. Figure 11 is the waveform of autocorrelation coefficient p ss [m] in the first embodiment and the waveform of autocorrelation coefficient in the second embodiment. Figure 11 The ordinate in

[0128] Although the waveform of autocorrelation coefficient is not a continuous curve but is composed of multiple points (solid dots in Figure 11 ), it is found that the waveform of autocorrelation coefficient p ss [m] in the first embodiment and the waveform of autocorrelation coefficient in the second embodiment are similar, especially the main features (i.e., the position and size of the highest peak) of the two are almost identical. Therefore, in the subsequent step S4 (signal periodicity judgment step), the periodicity of the signal can be accurately judged.

[0129] In fact, in Figure 11 , the waveform of autocorrelation coefficient p ss [m] is also composed of multiple points, only the density of the points is relatively large, and in the figure, m = 0, 1, 2, …, L-1, so each point of autocorrelation coefficient p ss [m] appears to be continuous. In this embodiment, in order to further improve the operation efficiency, we reduce the autocorrelation coefficient the translation granularity of the square wave. For example, in the case where the translation granularity of the square wave is set to 0.05 seconds, an autocorrelation coefficient can be calculated every 28 points of the value, i.e., only the autocorrelation coefficients for m = 0, 28, 56,... of the value. Thus Figure 11 the autocorrelation coefficients at each point in the square wave ss [m] appear more sparse than the autocorrelation coefficients p

[0130] According to the electrocardiosignal processing apparatus and method of the second embodiment, the amount of calculation, the calculation time, and the memory overhead of the autocorrelation calculation step are greatly reduced compared to the case where the square wave is not translated in the first embodiment, and the periodicity-related characteristics of the autocorrelation function R of the square wave are almost completely retained in the waveform of the autocorrelation coefficients ss [m], so that the periodicity of the electrocardiosignal can be more rapidly determined without reducing accuracy.

[0131] In addition, if the amplitude width L p is less than or equal to a predetermined value L ε , it can occur that no matter how the square wave is translated, all of the peak values within the prescribed time length T cannot be simultaneously overlapped with each other, which can reduce the approximation between the autocorrelation function of the square wave and the autocorrelation function R ss [m] of the preprocessed signal s[n]. On the other hand, if the amplitude width L p is greater than or equal to L s / 2, one square wave in the translated square wave may be simultaneously overlapped with two or more square waves in the square wave , which can cause the square wave search to become complicated, thereby increasing the calculation overhead.

[0132] Since the upper and lower limits of the peak interval L R and the square wave amplitude width L s are set based on values related to the heartbeat period T p , it is possible to avoid a reduction in the approximation between the autocorrelation function of the square wave and the autocorrelation function R ss [m] of the preprocessed signal s[n] due to the amplitude width L p being too narrow, and to avoid a situation where the square wave search becomes complicated and the calculation overhead increases due to the amplitude width L s being greater than or equal to L / 2.

[0133] The present application has been described with reference to several embodiments, but these embodiments are presented by way of example only and are not intended to limit the scope of the application. These new embodiments can be implemented in other various ways, and various omissions, substitutions, and changes can be made without departing from the spirit of the application. These embodiments and variations thereof are included in the scope, spirit of the application, and in the scope of the application as recited in the claims and equivalents thereof.

[0134] For example, the second embodiment of the present application, like the first embodiment, includes a signal preprocessing step in which the average value of each of the segmented signals is subtracted from each signal sample point included in the segmented signal to obtain a preprocessed signal. However, even if this signal preprocessing step is not performed, the periodicity of the electrocardiographic signal can be determined at least more quickly and with less overhead than in the case where the square wave conversion is not performed.

Claims

1. A method for processing electrocardiogram (ECG) signals, characterized in that, include: The detection signal acquisition step involves obtaining a detection signal by detecting the electrocardiogram signal of the organism using a sensor. The signal preprocessing step involves dividing the detected signal into multiple segment signals for a specified time length, calculating the average value of the multiple signals contained in each segment signal, and subtracting the average value of the segment signal from each signal sample point contained in the segment signal to obtain the preprocessed signal. The autocorrelation calculation step calculates the autocorrelation function of the preprocessed signal; The signal periodicity determination step involves searching for the highest peak in the autocorrelation coefficient waveform corresponding to each segmented signal after the autocorrelation calculation step, and calculating the difference between the highest peak and other adjacent peaks. The difference represents the intensity comparison between the signal and noise. When the difference is less than a first predetermined value, the segmented signal is identified as noise and discarded. The autocorrelation coefficient is the value obtained after normalizing the autocorrelation function.

2. The electrocardiogram signal processing method according to claim 1, characterized in that: The detection signal is a continuous signal The segmented signal is The average value is hour, The preprocessed signal is The autocorrelation function is The calculation formula is as follows: Where t is the integration variable, T is the delay, and T is the preprocessed signal. The length of time.

3. The electrocardiogram signal processing method according to claim 1, characterized in that: The detection signal is at a fixed frequency Collected discrete signals The segmented signal is The average value is At that time, the preprocessed signal is = The autocorrelation function is The calculation formula is as follows: Where n is the summation variable, m is the shift amount, L is the signal length of the preprocessed signal s[n], and the range of n in s[n] is... .

4. The electrocardiogram signal processing method according to claim 1, characterized in that, Also includes: The signal periodicity determination step involves searching for the highest peak in the autocorrelation coefficient waveform corresponding to each segmented signal after the autocorrelation calculation step, and calculating the difference between the highest peak and other adjacent peaks. The difference represents the intensity comparison between the signal and noise. When the difference is greater than or equal to a first predetermined value, the segmented signal is considered valid. The autocorrelation coefficient is the value obtained after normalizing the autocorrelation function.

5. The electrocardiogram signal processing method according to claim 4, characterized in that: The first predetermined value is set to 0.

1.

6. The electrocardiogram signal processing method according to claim 1, characterized in that: The specified time length T is set according to the possible range of human heartbeat cycles, and the specified time length T includes at least two heartbeat cycles T. R The length.

7. The electrocardiogram signal processing method according to claim 6, characterized in that: The specified time length T is any value that satisfies 3s≤T≤5s.

8. The electrocardiogram signal processing method according to claim 3, characterized in that: In the autocorrelation calculation step, For the preprocessed signal The peaks contained in the sample are extracted, and the interval between any two adjacent peaks in the extracted sample is greater than or equal to a predetermined time interval. Thus, signal S is obtained. d ; The above signal S d Convert to square wave ; Calculate the above square wave autocorrelation function .

9. The electrocardiogram signal processing method according to claim 8, characterized in that: The square wave width satisfy , in, It is greater than 0 and greater than the maximum period difference ΔT R The predetermined value, the maximum period difference ΔT R For the specified time length T, the internal beat cycle T R The maximum possible difference between them, the specified time interval It is the minimum heartbeat cycle within a specified time length T.

10. The electrocardiogram signal processing method according to claim 9, characterized in that: The width for Any value in the range.

11. The electrocardiogram signal processing method according to claim 8, characterized in that: The autocorrelation function The calculation involves the following steps: (1) The square wave After translation, the square wave is obtained. ; (2) Search for the square wave at the same time With the translated back wave The overlapping portion of the square waves is R1~Rr; (3) Calculate the sum of the products of the overlapping portions R1 to Rr using the following formula; Where R is the number of overlapping square wave portions R1~Rr. It is the height of the overlapping portion R1~Rr of each wave. R1~Rr represents the width of the overlapping portion of each wave.

12. A method for processing electrocardiogram (ECG) signals, characterized in that, include: The detection signal acquisition step involves obtaining a detection signal by detecting the electrocardiogram signal of the organism using a sensor. The signal preprocessing step involves dividing the detected signal into multiple segment signals for a specified time length, calculating the average value of the multiple signals contained in each segment signal, and subtracting the average value of the segment signal from each signal sample point contained in the segment signal to obtain the preprocessed signal. The autocorrelation calculation step calculates the autocorrelation function of the preprocessed signal; The signal periodicity determination step involves searching for the highest peak in the autocorrelation coefficient waveform corresponding to each segmented signal after the autocorrelation calculation step, and calculating the difference between the highest peak and other adjacent peaks. The difference represents the intensity comparison between the signal and noise. When the difference is greater than or equal to a first predetermined value, the segmented signal is considered valid. The autocorrelation coefficient is the value obtained after normalizing the autocorrelation function.

13. The electrocardiogram signal processing method according to claim 12, characterized in that: The detection signal is a continuous signal The segmented signal is The average value is hour, The preprocessed signal is The autocorrelation function is The calculation formula is as follows: Where t is the integration variable, T is the delay, and T is the preprocessed signal. The length of time.

14. The electrocardiogram signal processing method according to claim 12, characterized in that: The detection signal is at a fixed frequency Collected discrete signals The segmented signal is The average value is At that time, the preprocessed signal is = The autocorrelation function is The calculation formula is as follows: Where n is the summation variable, m is the shift amount, L is the signal length of the preprocessed signal s[n], and the range of n in s[n] is... .

15. The electrocardiogram signal processing method according to claim 12, characterized in that: The first predetermined value is set to 0.

1.

16. The electrocardiogram signal processing method according to claim 12, characterized in that: In the signal periodicity determination step, the period Ts of the preprocessed signal corresponding to each autocorrelation coefficient waveform is calculated based on the position of the highest peak in the waveform. The period Ts of the preprocessed signal is the same as the period Ty of the segmented signal.

17. The electrocardiogram signal processing method according to claim 16, characterized in that: If the period Ty of the segmented signal calculated in the signal periodicity determination step is satisfied with the condition 0.3s < Ty < 1.5s, the segmented signal that does not satisfy the condition is identified as noise and discarded.

18. The electrocardiogram signal processing method according to claim 16 or 17, characterized in that: In the signal periodicity determination step, the sample variance S of the period Ty of multiple consecutive segmented signals is calculated. If this occurs, the segmented signal is identified as noise and discarded.

19. The electrocardiogram signal processing method according to claim 12, characterized in that: The specified time length T is set according to the possible range of human heartbeat cycles, and the specified time length T includes at least two heartbeat cycles T. R The length.

20. The electrocardiogram signal processing method according to claim 19, characterized in that: The specified time length T is any value that satisfies 3s≤T≤5s.

21. The electrocardiogram signal processing method according to claim 14, characterized in that: In the autocorrelation calculation step, For the preprocessed signal The peaks contained in the sample are extracted, and the interval between any two adjacent peaks in the extracted sample is greater than or equal to a predetermined time interval. Thus, signal S is obtained. d ; The above signal S d Convert to square wave ; Calculate the above square wave autocorrelation function .

22. The electrocardiogram signal processing method according to claim 21, characterized in that: The square wave width satisfy , in, It is greater than 0 and greater than the maximum period difference ΔT R The predetermined value, the maximum period difference ΔT R For the specified time length T, the internal beat cycle T R The maximum possible difference between them, the specified time interval It is the minimum heartbeat cycle within a specified time length T.

23. The electrocardiogram signal processing method according to claim 22, characterized in that: The width for Any value in the range.

24. The electrocardiogram signal processing method according to claim 21, characterized in that: The autocorrelation function The calculation involves the following steps: (1) The square wave After translation, the square wave is obtained. ; (2) Search for the square wave at the same time With the translated back wave The overlapping portion of the square waves is R1~Rr; (3) Calculate the sum of the products of the overlapping portions R1 to Rr using the following formula; Where R is the number of overlapping square wave portions R1~Rr. It is the height of the overlapping portion R1~Rr of each wave. R1~Rr represents the width of the overlapping portion of each wave.

25. An electrocardiogram (ECG) signal processing device, characterized in that, include: The detection signal acquisition unit acquires the detection signal by detecting the electrocardiogram signal of the organism using a sensor; The signal preprocessing unit divides the detected signal into multiple segment signals for a specified time length, calculates the average value of the multiple signals contained in each segment signal, and subtracts the average value of the segment signal from each signal sample point contained in the segment signal to obtain the preprocessed signal. The autocorrelation calculation unit calculates the autocorrelation function of the preprocessed signal; as well as The signal periodicity judgment unit searches for the highest peak in the autocorrelation coefficient waveform corresponding to each segmented signal, and calculates the difference between the highest peak and other adjacent peaks. The difference represents the intensity comparison between the signal and noise. When the difference is less than a first predetermined value, the segmented signal is identified as noise and discarded. The autocorrelation coefficient is the value obtained after normalizing the autocorrelation function.

26. An electrocardiogram (ECG) signal processing device, characterized in that, include: The detection signal acquisition unit acquires the detection signal by detecting the electrocardiogram signal of the organism using a sensor; The signal preprocessing unit divides the detected signal into multiple segment signals for a specified time length, calculates the average value of the multiple signals contained in each segment signal, and subtracts the average value of the segment signal from each signal sample point contained in the segment signal to obtain the preprocessed signal. The autocorrelation calculation unit calculates the autocorrelation function of the preprocessed signal; as well as The signal periodicity judgment unit searches for the highest peak in the autocorrelation coefficient waveform corresponding to each segmented signal, and calculates the difference between the highest peak and other adjacent peaks. The difference represents the intensity comparison between the signal and noise. When the difference is greater than or equal to a first predetermined value, the segmented signal is considered valid. The autocorrelation coefficient is the value obtained after normalizing the autocorrelation function.

27. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the steps of the electrocardiogram signal processing method as described in claim 1 or claim 12.

28. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the steps of the electrocardiogram signal processing method as described in claim 1 or claim 12.

Images