CN104993832B - A kind of 3 correlation waveform smoothing methods based on high-speed sample data - Google Patents

Abstract

The present invention proposes a three-point correlation waveform smoothing method based on high-speed sampling data, starting from the consideration of the correlation between the three sampling points of continuous sampling, using the theoretical principle of the predictability of the change trend between continuous sampling data, Perform high-speed sampling data error discrimination and error correction processing, so that the sampling error falls back to the normal and ideal trend range, thereby improving the acquisition effect and the quality of signal processing. The method of the invention performs processing from software analysis, reduces the requirement of the high-speed sampling system on the hardware environment, and greatly saves hardware debugging, personnel expenses, resource consumption and the like. Since the focus is on the data processing after sampling, the path of the interference source is not concerned, and the dependence on the hardware environment is also reduced, which makes the applicability of the method stronger, the scope of application wider, and the universality stronger. At the same time, the method of debugging through software parameters also makes the means of digital signal processing convenient and flexible, saving time and effort.

Description

A three-point correlation waveform smoothing method based on high-speed sampling data

technical field

The invention relates to the field of digital signal processing, in particular to a method for smoothing sampling waveforms based on data correlation of three consecutive sampling points of high-speed sampling data.

Background technique

Under ideal working conditions, the periodic sine wave analog signal should also be an ideal sine wave waveform after sampling, but in the actual working environment, it is interfered by factors such as power supply, other board signals, space electromagnetic radiation, etc. In the case of sampling, the collected signal waveform is often not exactly the same as the theoretical sine wave, and sometimes the difference is very large, which makes the signal quality of the analog signal degrade after sampling processing, and the signal analysis effect is degraded.

After debugging and improving the circuit board hardware and optimizing the FPGA signal processing program, the waveform burrs, ripples, and noise of the sampled digital signal will be improved, but they still cannot be completely eliminated. When the hardware environment and FPGA environment are debugged to a certain extent and cannot be solved, sometimes it can be considered to add a certain waveform optimization algorithm to the upper computer CPU to improve the acquisition signal error defect, make up for the excessive sampling error in signal acquisition, improve the acquisition quality, improve Signal analysis effect.

Figure 1 shows the difference between ideal sampling and actual sampling. As shown in Figure 1, due to the existence of noise, interference, etc., the actual sampling point swings within a certain range of the ideal sampling curve, and the amplitude and deviation of the swing depend on the specific noise. Interference situations vary. When the noise interference is relatively small, the sampling point is basically consistent with the ideal situation. When the noise interference is relatively large, the sampling point will deviate from the ideal point within a certain range. Since the noise and interference cannot be completely eliminated in actual working conditions, the actual sampling time , it is normal and reasonable that the actual sampling point deviates from the ideal sampling point within a certain range.

When individual actual sampling points deviate greatly from the ideal situation, exceed the deviation range affected by normal noise, and form abnormal waveform burrs or distortions, it can be determined that the sampling point is incorrectly sampled or processed incorrectly, and its subsequent digital signal Certain adverse effects will be formed during the processing, so some pre-processing can be properly carried out to make the distortion point or burr point fall back to the influence of the normal noise range, thereby improving the effect of signal processing.

Usually, in the case of unsatisfactory sampling, the test and analysis of the hardware environment will be carried out first, and efforts will be made to find the interference paths and sources that affect signal sampling, and some measures will be implemented to reduce the impact of interference. However, it is difficult to do it through hardware debugging. to completely eliminate interference. After the hardware environment reaches a certain level of debugging, it will not be able to further improve. At the same time, the improvement of the hardware environment needs to increase the cost of manpower, time, and hardware resources to a certain extent, and in many cases, due to the different environments, the debugging results at one time do not meet the general requirements in multiple environments.

Contents of the invention

In order to solve the deficiencies in the above-mentioned prior art, the present invention proposes a three-point correlation waveform smoothing method based on high-speed sampling data. Considering the correlation between the three sampling points of continuous sampling, using the continuous sampling data Based on the theoretical principle of the predictability of the inter-time change trend, high-speed sampling data error discrimination and error correction processing are carried out, so that the sampling error falls back to the normal and ideal trend range, thereby improving the acquisition effect, improving the quality of signal processing, and improving the working performance of the acquisition system.

Technical scheme of the present invention is realized like this:

A three-point correlation waveform smoothing method based on high-speed sampling data, comprising the following steps:

Step (1): For the N-point high-speed sampling data values of continuous sampling, first determine its maximum value MAX and minimum value MIX by comparison method, as well as its positions t1 and t2 in the sampling sequence;

Step (2): After obtaining the extreme point, start from the extreme point t1, count the number of sampling points that decrease monotonically in one direction between the MAX point and the adjacent theoretical middle point MID, and count the number of sampling points from the MIX point to the adjacent theoretical middle point from t2 The number of sampling points that rise monotonously in one direction between the theoretical intermediate points MID; the maximum value of the four statistical values is counted as X;

Step (3): Judging the value of X, when X is greater than or equal to 5, enter step (4); otherwise return to step (1), and proceed to the next round of data discrimination processing;

Step (4): Set the starting point Y of the three-point correlation judgment to start from the second point, and assign Y to be equal to 2;

Step (5): Determine whether Y is smaller than N, and if so, take a total of three consecutive sampling values V _Y-1 , V _Y , and V _Y+1 at the Y position and its adjacent front and rear positions, and enter step (6) ; Otherwise, the whole processing flow ends, and returns to step (1) for the next round of data discrimination processing;

Step (6): Discriminate and process the three consecutively sampled data. If it belongs to the situation within the trend, proceed to step (7) to discriminate the deviation from the ideal range; Misjudgment;

Step (7): In the case of deviating from the ideal range, define the three points of continuous sampling in the above step (5) as p1, p2, and p3, and the data of the three points are Vp1, Vp2, Vp3, and p1 and The midpoint pp2 of p3, the data of pp2 point is Vpp2, and the judgment process of whether p2 deviates from the ideal range is carried out. If p2 is within the set ideal range, then Y is added by 1, and the next point is judged by returning to step (5). Processing; Otherwise, enter step (8);

Step (8): replace p2 data Vp2 with pp2 data Vpp2, Y adds 1, returns to step (5), and carries out the discrimination processing of next point;

Step (9): In the case of error discrimination processing, define the three points of continuous sampling in the above step (5) as p4, p5, p6, and the data of the three points are Vp4, Vp5, Vp6, and the values of p4 and p6 Midpoint pp5, the data of pp5 point is Vpp5, carry out the discriminating process to p5 whether obviously wrong in rising or falling, if so, then enter step (10); If not, then enter step (11);

Step (10): replace p5 data Vp5 with pp5 data Vpp5, Y adds 1, returns to step (5), and carries out the discrimination processing of next point;

Step (11): Carry out the discrimination processing of the extreme value data range on p5, if it is within the extreme value range, it is judged as the situation within the reasonable swing of the ideal range, Y is added by 1, and the step (5) is returned to proceed to the next point of discrimination processing; If not, replace p5 data Vp5 with pp5 data Vpp5, add 1 to Y, return to step (5), and proceed to the next point of discrimination processing.

Optionally, the comparison method for finding extreme points in the step (1) is:

Assign the first data of consecutive N point sampling data to MAX and MIX, and assign its position to t1 and t2, and then start from the second sampling point, compare each sampling point with MAX, if it is greater than or equal to MAX , then assign its value to MAX, assign its position to t1, otherwise skip the comparison of the next data point; and compare each sampling point with MIX, if it is less than or equal to MIX, assign its value to MIX, and set Assign its position to t2, otherwise skip the comparison of the next data point; proceed in turn until the last data point is compared, then determine the values of the extreme points MAX, MIX and their positions t1, t2.

Optionally, the statistical method to X in the step (2) is:

The statistical value X ₁ is reset to zero, starting from the t1 position, and the value is taken forward sequentially. When the value is less than or equal to MAX and greater than or equal to MID, the statistical value X ₁ is increased by 1; otherwise, it jumps out;

The statistical value X ₂ is reset to zero, starting from the position of t1, and the value is sequentially taken backwards. When the value is less than or equal to MAX and greater than or equal to MID, the statistical value X ₂ is increased by 1; otherwise, it jumps out;

The statistical value X ₃ is reset to zero, starting from the position of t2, the value is taken forward sequentially, when the value is greater than or equal to MIX, and less than or equal to MID, the statistical value X ₃ is increased by 1; otherwise, it jumps out;

The statistical value X ₄ is reset to zero, starting from the t2 position, and the value is sequentially taken backwards. When the value is greater than or equal to MIX and less than or equal to MID, the statistical value X ₄ is increased by 1; otherwise, it will jump out;

Assign the maximum value among the statistical values X ₁ , X ₂ , X ₃ , and X ₄ to X.

Optionally, the method for judging the situation within the trend in the step (6) should satisfy: V _Y-1 ≤ V _Y ≤ V _Y+1 or V _Y-1 ≥ V _Y ≥ V _Y+1 .

Optionally, the out-of-trend condition in the step (6) should satisfy: V _Y-1 <V _Y and V _Y+1 <V _Y , or V _Y-1 >V _Y and V _Y+1 >V _Y .

Optionally, in the step (7), the determination method of the data value Vpp2 of the midpoint pp2 of p1 and p3 is:

Optionally, in the step (7), the method of judging that the p2 point deviates from the ideal range is:

Among them, Δ2 is defined as the vertical distance between p1 and p3 points, for the equal interval sampling process, Δ2=|Vp1-Vp3|;

Δ1 is defined as the vertical distance between p2 and pp2 points, for the equal interval sampling process, Δ1=|Vp2-Vpp2|; if Δ1 is equal to 0, then force Δ1 to be equal to 0.001;

β1 is defined as the ratio of Δ2 to Δ1, which is used as the judgment that point p2 deviates from the theoretical range.

Optionally, in the step (9), the determination method of the data value Vpp5 of the midpoint pp5 of p4 and p6 is:

Optionally, the method for judging obvious errors in the rise or fall in the step (9) is: when Δ3≥Δ4, When Δ3<Δ4, in,

Δ3 is defined as the vertical distance between points p4 and p5, for equal interval sampling process, Δ3=|Vp4-Vp5|;

Δ4 is defined as the vertical distance between p6 and p5 points, for the equal interval sampling process, Δ4=|Vp6-Vp5|;

β2 is defined as the ratio of the large value to the small value in Δ3 and Δ4, which is used as the judgment of the deviation range of point p5 from p4 and p6.

Optionally, in the step (11), the basis for judging the extreme value data range is: Vp5>(MAX×(1-δ)) or Vp5<(MIX×(1+δ)); wherein,

δ is defined as the extreme value range factor, and the selection range of δ satisfies 0≤δ≤0.5.

Optionally, the value of the theoretical middle point MID in step (2) is determined according to the sampling and quantization bit width of the analog-to-digital converter device used specifically, and it is equal to the middle value determined by the maximum and minimum quantization values.

The beneficial effects of the present invention are:

(1) Using the theoretical principle of the predictability of the change trend between continuous sampling data, the high-speed sampling data error discrimination and error correction processing are carried out, so that the sampling error falls back to the normal and ideal trend range, thereby improving the acquisition effect and improving the signal processing. Quality, improve the performance of the acquisition system;

(2) Processing from software analysis reduces the requirements of the high-speed sampling system on the hardware environment, and greatly saves hardware debugging, personnel expenses, and resource consumption;

(3) Since the focus is on the data processing after sampling, the way of the interference source is not concerned, and the dependence on the hardware environment is also reduced, so that the applicability of this method is strengthened, the scope of application becomes wider, and the versatility is strengthened;

(4) The way of debugging through software parameters also makes the means of digital signal processing convenient and flexible, saving time and effort.

Description of drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings that need to be used in the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only These are some embodiments of the present invention. Those skilled in the art can also obtain other drawings based on these drawings without creative work.

Fig. 1 is the schematic diagram of ideal sampling and actual sampling;

Fig. 2 is a schematic diagram of a method for determining the relationship between A/D sampling frequency and operating frequency in the method of the present invention;

Fig. 3 is a schematic diagram of the situation where A/D sampling deviates from the ideal curve;

Fig. 4 is a schematic diagram of an obvious error point in A/D sampling;

FIG. 5 is a schematic diagram of the situation after A/D sampling correction.

Detailed ways

The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some, not all, embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without making creative efforts belong to the protection scope of the present invention.

The basic principle of the method of the present invention is to use the correlation between the three sampling data points of continuous sampling to discriminate the sampling data error, and perform error correction processing, so as to realize the high-speed sampling data due to interference factors, electromagnetic compatibility, timing stability, etc. The correction of individual sampling errors caused by it makes the high-speed sampling data waveform continuous and smooth, thereby improving the data quality of signal acquisition and improving the effect of signal processing under high-speed sampling.

During the actual application of the method of the present invention, the following aspects need to be considered: the relationship between the sampling frequency and the working frequency, the situation that the sampling point deviates from the ideal range, and the processing situation of the sampling error point.

(1) The relationship between sampling frequency and operating frequency

The method of the invention is applicable to the working state of low frequency and high sampling, and the discrimination processing of multiple continuous state signal points during the signal rising process or falling process.

Taking 2GSps sampling as an example, for a signal with a working frequency of 100MHz, 20 points can be sampled in one sampling period, and there are 10 points in a rising or falling process. During a monotonous rising or falling process, the trend of points Both state and state have certain rules, which is relatively easy for subsequent identification and processing. For a signal with a working frequency of 500MHz, only 4 points can be sampled in one sampling cycle, and the number of points in the rising or falling process is too small to reflect the regularity of the sampling points in the entire cycle. The application of the algorithm has certain limitations. Therefore, the method of the present invention only considers the situation that there are multiple sampling points in the rising or falling process of a working signal, that is, the relationship that the sampling frequency is at least 10 to 20 times the working frequency is the best application situation. Taking a 2GSps sampling situation as an example, the low-band frequency when the operating frequency is lower than 200MHz is the best state for application.

The relationship between the working frequency and the sampling frequency needs to be analyzed from the specific sampling data information. In the whole processing process, it is first necessary to determine the positions of the extreme points, maximum points, and minimum points of the data in the sampling data. The maximum value MAX at the time point t1 and the minimum value MIX at the time point t2 are marked in FIG. 2 .

After obtaining the extreme point, count the number of sampling points that decrease monotonically in one direction between the MAX point and the adjacent theoretical middle point MID, and count the number of sampling points that rise monotonously in one direction between the MIX point and the adjacent theoretical middle point MID, and count four times The maximum value of the value is counted as X. The relationship between the current rough working frequency and the sampling rate is judged based on the maximum value of the four statistical values. The method of taking the maximum value for four statistics is to avoid inaccurate statistical analysis when the extreme point MAX or MIX is located near the beginning and end of the data in a single statistics.

The above statistical methods for X are specifically as follows:

The statistical value X ₁ is reset to zero, starting from the t1 position, and the value is taken forward sequentially. When the value is less than or equal to MAX and greater than or equal to MID, the statistical value X ₁ is increased by 1; otherwise, it jumps out;

The statistical value X ₂ is reset to zero, starting from the position of t1, and the value is sequentially taken backwards. When the value is less than or equal to MAX and greater than or equal to MID, the statistical value X ₂ is increased by 1; otherwise, it jumps out;

The statistical value X ₃ is reset to zero, starting from the position of t2, the value is taken forward sequentially, when the value is greater than or equal to MIX, and less than or equal to MID, the statistical value X ₃ is increased by 1; otherwise, it jumps out;

The statistical value X ₄ is reset to zero, starting from the t2 position, and the value is sequentially taken backwards. When the value is greater than or equal to MIX and less than or equal to MID, the statistical value X ₄ is increased by 1; otherwise, it will jump out;

Assign the maximum value among the statistical values X ₁ , X ₂ , X ₃ , and X ₄ to X.

For example, when the sampling rate is 2GSps, the number of sampling points obtained by statistics is about 5, and it can be judged that the current frequency is about 100MHz. The number obtained by this method is about a quarter of the number of sampling points in the entire sampling period of the working frequency. The MID used here is the middle nominal theoretical value of the full-scale sampling of the A/D sampler. The reason why the statistics of the sampling data between MAX or MIX unidirectional and monotonous to MID is because in the actual sampling process, the distance between MAX and MIX It may contain multiple sampling periods, and due to interference such as noise, MAX and MIX are often not the maximum and minimum values of the same sampling period. If the number of sampling points between MAX and MIX is used to count, it is often necessary to add more Adding restrictive conditions to carry out multi-period discrimination processing is not as accurate and simple as the statistical method with MID as a reference.

(2) When the sampling point deviates from the ideal range

In the normal working process, the sampling point is always attached to a certain range of the ideal curve. This range changes with the size of the noise influence, but the signal analysis needs should be met within the noise influence range. When the sampling system deviates slightly due to other interference problems, it will have a certain adverse effect on the signal analysis results.

The enlarged sampling point p2 on the left side of Figure 3 deviates too much from the theoretical curve. In theory, the three sampling points p1, p2, and p3 should be able to be connected into a curve that approximates a straight line. In theory, point p2 should be located near the midpoint pp2 of the connection between p1 and p3, but in the actual sampling process, point p2 may deviate from pp2 In the case of large points, it can be considered that within the scope of noise influence, the sampling error of point p2 is relatively large. This point will have an impact on data processing, and pre-processing is required.

The idea of the method of the present invention is to pull the point with a large deviation back to the vicinity of the theoretical midpoint. However, the basis for discrimination needs to consider many factors, for example, points p1 and p3 may also be points with large deviations. Therefore, the judgment is not based on the position of a single point, but the distance between points is used as the judgment criterion, which can reduce the one-sidedness of individual points.

In the invented method, due to equal interval sampling, the vertical distance Δ2 between p1 and p3 points and the vertical distance Δ1 between p2 and pp2 points are used for discrimination. In theory, when p2 is closer to the theoretical value At pp2, the larger the ratio of Δ2/Δ1, the more accurate the point p2 is. When the proportional multiple is smaller, it means that the p2 deviates from the pp2 point more, and the sampling error of the p2 point is larger.

In this case, the basis for discrimination is Vp1≤Vp2≤Vp3 (rising condition) or Vp1≥Vp2≥Vp3 (declining condition), and Δ2/Δ1<5, and p2 deviates from the theoretical point pp2 by more than 2/5 of the nominal distance Case. Of course, the criterion of Δ2/Δ1 can be adjusted according to specific noise conditions. When the above two conditions are met, it can be judged that the p2 sampling point deviates too much from the theoretical value point, the sampling error is too large, and artificial correction is required, and the theoretical value point pp2 is replaced by p2 point for subsequent data processing, thereby eliminating the error point p2 influences.

(3) The case of sampling obvious wrong points

In the normal sampling process, the monotonous rise or fall should maintain its consistent monotonicity, but in the actual sampling process, due to interference problems, there will be p5 in the points p4, p5, and p6 shown on the right side of Figure 4 The error situation, which no longer conforms to the theoretical monotonous rise or fall, has reversed during the normal trend process, which can be characterized as a sampling error point, which will affect subsequent data processing and needs to be corrected through pre-processing.

Point p5 is used as an error point, and the basis for discrimination is as follows: first, it cannot satisfy the situation that the error in the above (2) is too large, that is, it should be Vp5<Vp4 and Vp5<Vp6, or Vp5>Vp4 and Vp5>Vp6, also That is, the middle point is either smaller than the adjacent points, or larger than the adjacent points, that is, the middle point does not satisfy the trend of monotonically increasing or decreasing. Then, to judge the situation where the middle point deviates from the neighboring point, first define the vertical distances Δ3 and Δ4 from the middle point p5 to the two neighboring points p4 and p6, and use the larger one as the numerator and the smaller one as the denominator to calculate the ratio multiple, As shown in Δ4/Δ3 in Figure 4, the reason for this judgment is to determine that point p5 occurs during the rise or fall of the waveform, rather than at the bottom or top of the waveform. Here, Δ4/Δ3≥2 is given for judgment, and most of the vertices and bottom points can be eliminated. When the judgment condition is valid, use the theoretical midpoint pp5 of p4 and p6 to replace p5 to eliminate the error of p5.

When Δ4/Δ3<2, it can be judged that the distance between p5 and p4 and p6 is not very different, and it can be attributed to the fact that p5 occurs at the top or bottom of the waveform, because there is a difference in amplitude between adjacent points It is not large. Due to noise interference, this situation may occur, such as the top and bottom areas marked in Figure 4. Under noise interference, this area may produce a small wave-shaped sampling situation, which does not meet the small rise or fall. trend.

Identify the signal correction of small incremental changes. First, judge whether p5 is the maximum point MAX or the minimum point MIX found above. If it is at the extreme point, the adjacent points should be smaller or larger than it. Yes Correct situation; otherwise, point p5 may be a small wave-shaped noise interference error situation. Since the data collected at one time may contain multiple waveform periods, the maximum and minimum judgments are not based on one point, but on a small range of extreme points to eliminate other waveform periods in multiple waveform periods. The range of extreme points in the cycle. If it is considered that the maximum points of multiple sampling periods in a sampling process are all within the range of MAX～MAX×(1-δ), and the minimum values are all within the range of MIX×(1+δ)～MIX (here it is assumed that all sampled data are is a positive value distribution), the range of δ can be set according to the actual situation, such as δ positioning 0.05, then point p5 needs to meet MIX×1.05≤Vp5≤MAX×0.95, then it is determined that a top or bottom noise interference error has occurred at point p5 , at this time still use its theoretical midpoint pp5 to replace p5, and carry out a small correction operation.

After the above small swing error correction pre-processing, the sampling points are greatly corrected, making the actual sampling effect close to the predicted result of theoretical analysis, as shown in Figure 5. In the subsequent signal processing process, the small swing error or small glitch interference caused by part of the noise or other interference can be effectively avoided.

The values of X, β1, β2, and δ described in the method of the present invention can be changed according to the specific hardware working environment, software debugging conditions, etc., to adapt to the optimization effect of the method of the present invention under different conditions.

The method of the present invention starts from the consideration of the correlation between the three sampling points of continuous sampling, and uses the theoretical principle of the predictability of the change trend between continuous sampling data to perform high-speed sampling data error discrimination and error correction processing, so that sampling errors fall back To the normal and ideal trend range, thereby improving the acquisition effect, improving the quality of signal processing, and improving the performance of the acquisition system. The method of the invention performs processing from software analysis, reduces the requirement of the high-speed sampling system on the hardware environment, and greatly saves hardware debugging, personnel expenses, resource consumption and the like. Since the focus is on the data processing after sampling, the path of the interference source is not concerned, and the dependence on the hardware environment is also reduced, which makes the applicability of the method stronger, the scope of application wider, and the universality stronger. At the same time, the method of debugging through software parameters also makes the means of digital signal processing convenient and flexible, saving time and effort.

The above descriptions are only preferred embodiments of the present invention, and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included in the scope of the present invention. within the scope of protection.

Claims (11)

1. A three-point correlation waveform smoothing method based on high-speed sampling data, is characterized in that, comprises the following steps: Step (1): For the N-point high-speed sampling data values of continuous sampling, first determine its maximum value MAX and minimum value MIX by comparison method, as well as its positions t1 and t2 in the sampling sequence; Step (2): After obtaining the extreme point, start from the extreme point t1, count the number of sampling points that decrease monotonically in one direction between the MAX point and the adjacent theoretical middle point MID, and count the number of sampling points from the MIX point to the adjacent theoretical middle point from t2 The number of sampling points that rise monotonously in one direction between the theoretical intermediate points MID; the maximum value of the four statistical values is recorded as X; Step (3): Judging the value of X, when X is greater than or equal to 5, enter step (4); otherwise return to step (1), and proceed to the next round of data discrimination processing; Step (4): Set the starting point Y of the three-point correlation judgment to start from the second point, and assign Y to be equal to 2; Step (5): Determine whether Y is smaller than N, and if so, take a total of three consecutive sampling values V _Y-1 , V _Y , and V _Y+1 at the Y position and its adjacent front and rear positions, and enter step (6) ; Otherwise, the whole processing flow ends, and returns to step (1) for the next round of data discrimination processing; Step (6): Discriminate and process the three consecutively sampled data. If it belongs to the situation within the trend, proceed to step (7) to discriminate the deviation from the ideal range; Misjudgment; Step (7): In the case of deviating from the ideal range, define the three points of continuous sampling in the above step (5) as p1, p2, and p3, and the data of the three points are Vp1, Vp2, Vp3, and p1 and The midpoint pp2 of p3, the data of pp2 point is Vpp2, and the judgment process of whether p2 deviates from the ideal range is carried out. If p2 is within the set ideal range, then Y is added by 1, and the next point is judged by returning to step (5). Processing; Otherwise, enter step (8); Step (8): replace p2 data Vp2 with pp2 data Vpp2, Y adds 1, returns to step (5), and carries out the discrimination processing of next point; Step (9): In the case of error discrimination processing, define the three points of continuous sampling in the above step (5) as p4, p5, p6, and the data of the three points are Vp4, Vp5, Vp6, and the values of p4 and p6 Midpoint pp5, the data of pp5 point is Vpp5, carry out the discriminating process to p5 whether obviously wrong in rising or falling, if so, then enter step (10); If not, then enter step (11); Step (10): replace p5 data Vp5 with pp5 data Vpp5, Y adds 1, returns to step (5), and carries out the discrimination processing of next point; Step (11): Carry out the discrimination processing of the extreme value data range on p5, if it is within the extreme value range, it is judged as the situation within the reasonable swing of the ideal range, Y is added by 1, and the step (5) is returned to proceed to the next point of discrimination processing; If not, replace p5 data Vp5 with pp5 data Vpp5, add 1 to Y, return to step (5), and proceed to the next point of discrimination processing. 2. a kind of three-point correlation waveform smoothing method based on high-speed sampling data as claimed in claim 1, is characterized in that, the comparison method of finding extremum point in described step (1) is: Assign the first data of consecutive N point sampling data to MAX and MIX, and assign its position to t1 and t2, and then start from the second sampling point, compare each sampling point with MAX, if it is greater than or equal to MAX , then assign its value to MAX, assign its position to t1, otherwise skip the comparison of the next data point; and compare each sampling point with MIX, if it is less than or equal to MIX, assign its value to MIX, and set Assign its position to t2, otherwise skip the comparison of the next data point; proceed in turn until the last data point is compared, then determine the values of the extreme points MAX, MIX and their positions t1, t2. 3. a kind of three-point correlation waveform smoothing method based on high-speed sampling data as claimed in claim 1, is characterized in that, in the described step (2), the statistical method to X is: The statistical value X ₁ is reset to zero, starting from the t1 position, and the value is taken forward sequentially. When the value is less than or equal to MAX and greater than or equal to MID, the statistical value X ₁ is increased by 1; otherwise, it jumps out; The statistical value X ₂ is reset to zero, starting from the position of t1, and the value is sequentially taken backwards. When the value is less than or equal to MAX and greater than or equal to MID, the statistical value X ₂ is increased by 1; otherwise, it jumps out; The statistical value X ₃ is reset to zero, starting from the position of t2, the value is taken forward sequentially, when the value is greater than or equal to MIX, and less than or equal to MID, the statistical value X ₃ is increased by 1; otherwise, it jumps out; The statistical value X ₄ is reset to zero, starting from the t2 position, and the value is sequentially taken backwards. When the value is greater than or equal to MIX and less than or equal to MID, the statistical value X ₄ is increased by 1; otherwise, it jumps out;, Assign the maximum value among the statistical values X ₁ , X ₂ , X ₃ , and X ₄ to X. 4. a kind of three-point correlation waveform smoothing method based on high-speed sampling data as claimed in claim 1, is characterized in that, the discriminant method of situation in the trend in described step (6) should satisfy: V _Y- 1≤V _Y ≤ V _Y+1 or V _Y-1 ≥ V _Y ≥ V _Y+1 . 5. a kind of three-point correlation waveform smoothing method based on high-speed sampling data as claimed in claim 1, is characterized in that, in described step (6), the situation outside the trend should satisfy: V _Y-1 ＜V _Y and V _Y+1 <V _Y , or V _Y-1 >V _Y and V _Y+1 >V _Y . 6. a kind of three-point correlation waveform smoothing method based on high-speed sampling data as claimed in claim 1, is characterized in that, in described step (7), the data value Vpp2 determination method of the midpoint pp2 of p1 and p3 is : 7. a kind of three-point correlation waveform smoothing method based on high-speed sampling data as claimed in claim 1, it is characterized in that, in described step (7), carry out the discriminant method that p2 point deviates from ideal range: Among them, Δ2 is defined as the vertical distance between p1 and p3 points, for the equal interval sampling process, Δ2=|Vp1-Vp3|; Δ1 is defined as the vertical distance between p2 and pp2 points, for the equal interval sampling process, Δ1=|Vp2-Vpp2|; if Δ1 is equal to 0, then force Δ1 to be equal to 0.001; β1 is defined as the ratio of Δ2 to Δ1, which is used as the judgment that point p2 deviates from the theoretical range. 8. a kind of three-point correlation waveform smoothing method based on high-speed sampling data as claimed in claim 1, is characterized in that, in described step (9), the data value Vpp5 determination method of the midpoint pp5 of p4 and p6 is : 9. A kind of three-point correlation waveform smoothing method based on high-speed sampling data as claimed in claim 1, is characterized in that, in the described step (9), the discriminant method of obvious error in rising or falling is: when Δ3≥Δ4 hour, When Δ3<Δ4, in, Δ3 is defined as the vertical distance between points p4 and p5, for equal interval sampling process, Δ3=|Vp4-Vp5|; Δ4 is defined as the vertical distance between p6 and p5 points, for the equal interval sampling process, Δ4=|Vp6-Vp5|; β2 is defined as the ratio of the large value to the small value in Δ3 and Δ4, which is used as the judgment of the deviation range of point p5 from p4 and p6. 10. A kind of three-point correlation waveform smoothing method based on high-speed sampling data as claimed in claim 1, it is characterized in that, in described step (11), carry out the discriminant basis of extremum data range: Vp5＞(MAX ×(1-δ)) or Vp5<(MIX×(1+δ)); where, δ is defined as the extreme value range factor, and the selection range of δ satisfies 0≤δ≤0.5. 11. a kind of three-point correlation waveform smoothing method based on high-speed sampling data as claimed in claim 1, is characterized in that, in the described step (2), the value of theoretical middle point MID is according to the analog-to-digital converter of concrete use The device is determined by sampling and quantizing the bit width, which is equal to the middle value determined by the quantization maximum value and minimum value.