A non-contact periodic motion weighted signal reconstruction method

Abstract

The application belongs to the field of signal processing and relates to a non-contact periodic motion weighted signal reconstruction method, which comprises the following steps: 1) obtaining a digital quadrature baseband signal containing target motion information; 2) calibrating the direct current bias of the digital quadrature baseband signal to obtain a pretreated signal; 3) performing a differential cross-multiplication operation on the pretreated signal to obtain a preliminary demodulation waveform; 4) obtaining an upper envelope line, a lower envelope line and an average envelope line of the preliminary demodulation waveform; 5) segmenting the preliminary demodulation waveform according to parameters obtained from the preliminary demodulation waveform; and 6) reconstructing the preliminary demodulation waveform obtained from step 3) based on the results obtained from steps 4) and 5) to obtain a weighted reconstructed motion signal. The application provides a non-contact periodic motion weighted signal reconstruction method which can improve the robustness and accuracy of motion information demodulation.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of signal processing, and particularly relates to a weighted signal reconstruction method, and especially to a non-contact periodic motion weighted signal reconstruction method. BACKGROUND

[0002] Radar sensing technology has important application value in actual vital sign monitoring, especially non-contact heartbeat detection, due to its high precision and non-contact measurement capability. The core of this technology is to accurately capture the subtle motion caused by heart activity through radar signals, and the accurate extraction of physiological signals fundamentally depends on the demodulation of accurate phase information. Phase information can accurately reflect the millimeter-level or even sub-millimeter-level periodic displacement changes of the chest surface, and is an important basis for realizing heart monitoring and heart timing. The radar system transmits electromagnetic waves, and measures the displacement of the target according to the phase difference between the echo signal and the transmitted signal. After the echo signal is down-converted and digitized, the in-phase and quadrature baseband signals can be represented as:

[0003]

[0004] where x(t) represents the displacement of the target over time. DC I (t) and DC Q (t) represent the DC offset in the I / Q channel, A I and A Q represent the signal amplitude. For a well-designed radar system, A I = A Q = A. λ represents the carrier wavelength. θ = 4πd0 / λ + θ0, which is determined by the initial phase θ0 of the transmitted signal and the nominal distance d0 between the target and the radar.

[0005] In motion signal demodulation, periodic motion demodulation is an important step, especially in applications involving target micro-motion or periodic motion. Periodic motion demodulation is essential in radar signal processing, mainly in extracting micro-motion features, suppressing noise, and improving detection accuracy. Through demodulation, the performance of radar in gesture recognition, vital sign monitoring, mechanical vibration detection and other applications can be significantly improved.

[0006] However, the accurate extraction of phase information faces many challenges, one of which is that the quadrature I / Q signals received by the radar may have amplitude imbalance and phase offset problems. These problems will introduce measurement errors and reduce the accuracy of displacement measurement.

[0007] In actual measurement, only the phase change caused by target displacement x(t) is concerned Due to the periodicity of the phase variation, the phase difference based displacement measurement method obtains the relative distance x(t) of the target instead of the absolute distance R. After corresponding simplification, the quadrature baseband signal can be further represented as:

[0008]

[0009] The arctangent function is originally used to extract the phase information from S I / S Q The signal can be described as follows:

[0010]

[0011] However, when calculating the phase span exceeding (-π / 2, π / 2), the range limitation of the arctangent function may cause phase discontinuity or ambiguity. The introduction of the differential cross multiplication algorithm solves this problem, but the complex trigonometric calculation still brings high calculation cost. The process can be described as:

[0012]

[0013] Where, I(t)=S I (t)-DC I (t), Q(t)=S Q (t)-DC Q (t)

[0014] The above process of subtracting the direct current offset from the baseband signal is called direct current offset calibration. Although the existing algorithm can effectively restore the motion, there is still a certain gap between the accuracy of the restored motion information and the actual motion information. The accuracy of the existing demodulation scheme is highly dependent on the accurate estimation of the circle center, and inaccurate estimation will introduce errors in the calculation, which will cause serious distortion when the errors continue to accumulate. SUMMARY

[0015] In order to solve the above technical problems in the background art, the present application provides a non-contact periodic motion weighted signal reconstruction method which can improve the robustness and accuracy of motion information demodulation.

[0016] In order to achieve the above purpose, the present application adopts the following technical scheme:

[0017] A non-contact periodic motion weighted signal reconstruction method, characterized in that: the non-contact periodic motion weighted signal reconstruction method comprises the following steps:

[0018] 1) Obtain a digital quadrature baseband signal containing target motion information;

[0019] 2) Calibrate the direct current bias of the digital quadrature baseband signal obtained in step 1) to obtain a preprocessed signal;

[0020] 3) performing a differential cross multiplication operation on the preprocessed signal obtained in step 2) to obtain a preliminary demodulated waveform;

[0021] 4) obtaining an upper envelope, a lower envelope and an average envelope of the preliminary demodulated waveform obtained in step 3);

[0022] 5) segmenting the preliminary demodulated waveform according to parameters obtained from the preliminary demodulated waveform;

[0023] 6) reconstructing the preliminary demodulated waveform obtained in step 3) based on the results obtained in step 4) and the results obtained in step 5) to obtain a weighted reconstructed motion signal.

[0024] Preferably, each sampling point in the digital quadrature baseband signal in step 1) comprises an I sampling value and a Q sampling value; and the target motion information comprises heartbeat information of a target or information of weak motion of a periodic pendulum ball.

[0025] Preferably, the specific implementation of step 2) is:

[0026]

[0027] wherein:

[0028] I and Q are respectively the I sampling value and the Q sampling value of each sampling point in the digital quadrature baseband signal;

[0029] I0and Q0are respectively the I sampling value and the Q sampling value after mean value subtraction preprocessing;

[0030] N is the total number of sampling points in the digital quadrature baseband signal.

[0031] Preferably, the specific implementation of step 3) is:

[0032]

[0033] wherein:

[0034] I0and Q0are respectively the I sampling value and the Q sampling value after mean value subtraction preprocessing;

[0035] and are respectively the differential forms of I0and Q0;

[0036] 0-t is the sampling time of the I sampling value and the sampling time of the Q sampling value;

[0037] x(t) is the waveform obtained by preliminary demodulation.

[0038] Preferably, the specific implementation of step 4) is:

[0039] 4.1) finding the peaks of the preliminary demodulated waveform obtained in step 3) to obtain the peaks and the troughs of the waveform;

[0040] 4.2) calculating the upper envelope, the lower envelope and the average envelope of the preliminary demodulated waveform from the peaks and the troughs of the waveform obtained in step 4.1).

[0041] Preferably, the step 4.1) is implemented as follows:

[0042] [peak_up, locs_up] = findpeaks(x(t))

[0043] [peak_low, locs_low] = findpeaks(-x(t))

[0044] wherein:

[0045] x(t) is the preliminary demodulated waveform;

[0046] peak_up and peak_low are the amplitudes of the peaks and the troughs of the preliminary demodulated waveform, respectively;

[0047] locs_up and locs_low are the corresponding location indices of the peaks and the troughs of the preliminary demodulated waveform, respectively;

[0048] findpeaks() is a peak finding function;

[0049] The step 4.2) is implemented as follows:

[0050] [E upper , E lower ] = envelope(x(t))

[0051]

[0052] wherein:

[0053] E upper and E lower are the upper envelope and the lower envelope of the preliminary demodulated waveform, respectively;

[0054] E avge is the average envelope of the preliminary demodulated waveform.

[0055] Preferably, the step 5) is implemented as follows:

[0056] 5.1) setting the threshold for segmentation according to the parameters obtained from the preliminary demodulated waveform;

[0057] 5.2) Segmenting the preliminary demodulated waveform according to the segment threshold obtained in step 5.1).

[0058] Preferably, the specific implementation of the step 5.1) is:

[0059]

[0060] Wherein:

[0061] f s is the sampling frequency of each sampling point in the digital quadrature baseband signal;

[0062] f is the motion information frequency of the preliminary demodulated waveform;

[0063] Threshold is the segment threshold;

[0064] The specific implementation of the step 5.2) is:

[0065]

[0066] Wherein:

[0067] x i is the i-th segment segmented according to the segment threshold.

[0068] Preferably, the specific implementation of the step 6) is:

[0069] 6.1) Obtaining the average envelope line deviation of each period segment based on the result obtained in step 4), and taking the deviation as a weight parameter;

[0070] 6.2) Reconstructing the preliminary demodulated waveform obtained in step 3) with the result obtained in step 6.1) and the result obtained in step 5) to obtain a weighted reconstructed motion signal.

[0071] Preferably, the specific implementation of the step 6.1) is:

[0072]

[0073] Wherein:

[0074] w i is the weight parameter of the i-th segment;

[0075] K is the number of different period segments;

[0076] is the average envelope line of the i-th segment;

[0077] t is the time at which the i-th segment is located;

[0078] The specific implementation of the step 6.2) is:

[0079]

[0080] wherein:

[0081] x recon is the motion signal obtained after weighting and reconstruction;

[0082] w i is the weight parameter of the ith segment;

[0083] K is the number of different cycle segments;

[0084] x exteni is the ith segment obtained after cycle extension of x exteni The expression of x exteni is:

[0085]

[0086] wherein:

[0087] T is the cycle of x i ;

[0088] x i is the ith segment obtained after preliminary segmentation according to the segmentation threshold;

[0089] t is the time at which the ith segment is located;

[0090] j is the cycle number of extension.

[0091] The present application has the following advantages:

[0092] The application provides a non-contact periodic motion weighted signal reconstruction method, and the method comprises the following steps: obtaining a digital quadrature baseband signal containing target motion information through sampling; each sampling point in the digital quadrature baseband signal comprises an I-channel sampling value and a Q-channel sampling value; removing the average value of the two-channel digital quadrature baseband signals I and Q to calibrate the direct current bias; performing an improved differential cross multiplication operation on the I-channel sampling value and the Q-channel sampling value of each two adjacent sampling points in the calibrated digital quadrature baseband signal to obtain a preliminary demodulation waveform; performing peak value searching on the preliminary demodulation waveform to obtain an upper envelope line, a lower envelope line and an average envelope line of the preliminary demodulation waveform; setting a segmented threshold according to the parameters obtained from the preliminary demodulation waveform, and segmenting the preliminary demodulation waveform according to the segmented threshold; performing periodic continuation and smoothing processing on the different segments after processing, calculating the average envelope line deviation degree of each cycle segment, and evaluating the noise influence on the signal in different time periods; taking the deviation degree as a weight parameter, and assigning a corresponding weight to each cycle segment; and finally, superimposing each cycle segment according to the weight to obtain the target motion information. The application adjusts the weight coefficients of different cycle segments by analyzing the noise level and signal quality of each cycle segment, so as to weaken the segments with large noise and enhance the contribution of high-quality segments. The method provided by the application can better adapt to the dynamic changes of actual signals and improve the reconstruction accuracy. Based on the non-contact periodic motion weighted signal reconstruction method provided by the application, non-contact periodic motion detection in various fields such as biological medicine can be realized. BRIEF DESCRIPTION OF DRAWINGS

[0093] Figure 1 is a flowchart of the non-contact periodic motion weighted signal reconstruction method provided by the application;

[0094] Figure 2 is a schematic diagram of the heartbeat signal measurement of the periodic micro mechanical motion measurement and the human target according to the application;

[0095] Figure 3 is the time domain demodulation result of the method and the MDACM solution of the 0.6mm micro periodic mechanical motion provided by the application;

[0096] Figure 4 is the time domain demodulation result of the method and the MDACM solution of the 1.2mm micro periodic mechanical motion provided by the application;

[0097] Figure 5 is the time domain demodulation result of the method and the MDACM solution of the 2.8mm micro periodic mechanical motion provided by the application;

[0098] Figure 6 is the time domain demodulation result of the method and the MDACM solution of the 5mm micro periodic mechanical motion provided by the application;

[0099] Figure 7 are time domain results of the heartbeat signal of the human body measured by the instrument when the method is provided by the present application. DETAILED DESCRIPTION

[0100] Reference Figure 1 The present application provides a non-contact periodic motion weighted signal reconstruction method, which comprises the following steps:

[0101] S1, obtaining a digital quadrature baseband signal containing target motion information by sampling; each sampling point in the digital quadrature baseband signal includes an I channel sampling value and a Q channel sampling value.

[0102] Wherein, the target motion information can include target heartbeat information, pulse information or blood flow information, etc., and is not limited to biological motion information. For example, it can also be the information of weak motion of a periodic pendulum ball or near-periodic motion. In step (1), the I channel data and Q channel data obtained by sampling are represented by I and Q, respectively.

[0103] S2, subtracting the average value of the two-channel digital quadrature baseband signals I and Q to calibrate the DC bias and obtain a preprocessed signal.

[0104] The I channel sampling value and the Q channel sampling value are respectively subjected to mean value subtraction preprocessing to obtain a preprocessed signal. The processing of the I channel data and the Q channel data can be represented by the following formula:

[0105]

[0106] Wherein, I0 and Q0 are the I channel sampling value and the Q channel sampling value after mean value subtraction preprocessing, respectively.

[0107] S3, performing an improved differential cross multiplication operation on the I channel sampling value and the Q channel sampling value of every two adjacent sampling points in the calibrated digital quadrature baseband signal to obtain a preliminary demodulation waveform.

[0108] It can be represented as:

[0109]

[0110] Wherein: and are the differential forms of I0 and Q0. 0-t is the time of I channel data and Q channel data sampling.

[0111] S4, by peak finding on the waveform obtained by preliminary demodulation, the peak and the trough of the waveform can be obtained, and further the upper envelope line, the lower envelope line and the average envelope line of the preliminary demodulation waveform can be obtained.

[0112] Specifically, the peak finding and envelope line calculation process in step S4 is as follows:

[0113] S401, peak finding is performed on the waveform x(t) obtained by the preliminary demodulation using a peak finding function, which can be expressed by the following formula:

[0114] [peak_up, locs_up] = findpeaks(x(t))

[0115] [peak_low, locs_low] = findpeaks(-x(t))

[0116] wherein x(t) is the waveform obtained by the preliminary demodulation, peak_up and peak_low are the amplitudes of the peak and valley of the waveform obtained by the preliminary demodulation, locs_up and locs_low are the corresponding position indexes of the peak and valley of the waveform obtained by the preliminary demodulation, and findpeaks() is the peak finding function.

[0117] S402, the upper envelope, the lower envelope and the average envelope of the waveform obtained by the preliminary demodulation can be calculated as follows:

[0118] [E upper , E lower ] = envelope(x(t))

[0119]

[0120] wherein E upper and E lower are the upper envelope and the lower envelope, and E avge is the average envelope.

[0121] S5, the threshold for segmentation is set according to the parameters obtained by the preliminary demodulation, and the waveform obtained by the preliminary demodulation is segmented according to the threshold for segmentation;

[0122] Although there is a certain error between the time-domain waveform obtained by the preliminary demodulation and the actual motion signal, the motion frequency error between the signal motion frequency obtained by the preliminary demodulation and the actual motion information is very small. The motion information frequency obtained by the preliminary demodulation is f, and since the error between the frequencies is small, the threshold can be set according to f to segment the waveform obtained by the preliminary demodulation.

[0123] The waveform obtained by the preliminary demodulation is segmented appropriately. The waveform obtained by the preliminary demodulation is segmented according to the threshold set according to f. The threshold for segmentation can be expressed as:

[0124]

[0125] wherein f s is the sampling frequency. After the threshold is calculated, the waveform obtained by the preliminary demodulation can be segmented. The specific segmentation is as follows:

[0126] First, the preliminary demodulated waveform is segmented according to the specified peak interval. Then, the relationship between the length of the segment and the threshold multiple is determined. If the segment length and the threshold multiple differ by a data length or less, the next segment is determined. If the segment length and the threshold multiple differ greatly, the next segment is spliced with the segment to continue the determination until the threshold condition is met. It can be represented as

[0127]

[0128] Wherein: x i is the i-th segment segmented according to the peak.

[0129] S6, the processed different segments are periodically extended and smoothed, and the average envelope deviation of each cycle segment is calculated to evaluate the noise influence on the signal in different time periods;

[0130] By analyzing the noise level and signal quality of each cycle segment, the weighting coefficient of different cycle segments is adjusted to weaken the segments with large noise and enhance the contribution of high-quality segments. The process includes:

[0131]

[0132] Wherein: T is the period of x i , x exteni is the i-th segment obtained by periodic extension.

[0133] S601, the deviation degree is taken as a weight parameter, and each cycle segment is assigned a corresponding weight. The specific calculation process can be represented as:

[0134]

[0135] Where w i is the weight parameter of the i-th segment; K is the number of different cycle segments, is the average envelope of the i-th segment, t i <t<t i+1 is the time at which the i-th segment is located.

[0136] S602, each cycle segment is weighted and superimposed according to the weight to obtain the target motion information. It can be represented by the following formula:

[0137]

[0138] Wherein: x recon is the motion signal obtained after weighting reconstruction.

[0139] In order to verify the effectiveness of the non-contact periodic motion information demodulation method based on periodic continuation provided by the embodiment of the present application, a series of simulation comparison experiments are carried out. The experimental setup is shown in Figure 2 .

[0140] Figure 2 The experimental scene setting for recovering micro mechanical motion and human life signals is shown in FIG. 1. In the experimental scene setting, a radio frequency transceiver transmits a single-frequency electromagnetic wave signal to a moving target. The electromagnetic wave is received by the receiver of the radio frequency transceiver after being returned by the moving target. After being down-converted and sampled in the receiver, the signal is input into a signal processing unit for demodulation to recover the motion information of the moving target.

[0141] In the experimental setup of periodic mechanical motion, the motor is controlled to perform sinusoidal motion with an amplitude of 0.6 mm, 1.2 mm, 2.8 mm and 5 mm and a frequency of 1.2 Hz. The MDACM demodulation algorithm (Modified Differentiate and Cross-Multiply, MDACM) and the method provided by the present application are used for comparison, and the ideal motion information is used for comparison.

[0142] Table 1. Error of demodulation results of the method of the embodiment of the present application and other algorithms

[0143]

[0144] Figure 3 , Figure 4 , Figure 5 and Figure 6 respectively show the waveform comparison of different demodulation algorithms and ideal signals under different motion amplitudes. Table 1 shows the root mean square error of the normalized demodulation results (NRMSE) of the method provided by the present application and the MDACM demodulation method. See Figure 3 When the motion amplitude is 0.6 mm, the normalized root mean square error of the demodulation results of the MDACM algorithm and the ideal signal is 0.2663, and the NRMSE of the method provided by the present application is 0.1331. In Figure 3 , it can be seen that the MDACM is affected by the direct current bias or noise, and the demodulation deviates. See Figure 4 When the motion amplitude is 1.2 mm, the NRMSE obtained by the MDACM algorithm is 0.1468, and the NRMSE of the method provided by the present application is 0.1310. See Figure 5When the motion amplitude is 2.8mm, the NRMSE obtained by the MDACM algorithm is 0.1696, and the NRMSE of the method provided by the application is 0.1503. See Figure 6 When the motion amplitude is 5mm, the NRMSE obtained by the MDACM algorithm is 0.1365, and the NRMSE of the method provided by the application is 0.1176. Compared with the MDACM algorithm demodulation, the method provided by the application reduces the NRMSE by 50%, 10.7%, 11.4% and 13.8% respectively.

[0145] In the experiment of restoring the human heartbeat signal under normal conditions, the conventional pulse sensor and the radio frequency transceiver are used for measurement respectively. Figure 7 The method provided by the application and the human heartbeat signal measured by the PPG instrument are displayed. The experimental results show that the demodulated signal retains the inherent physical logic characteristics of PPG. In the RRI detection evaluation, the method provided by the application shows excellent consistency with the reference instrument, and realizes the maximum and minimum RRI errors of 3.5% and 2%. The method provided by the application improves the fidelity of time domain signal reconstruction, and can be comparable to the detection effect of the contact sensor, especially improves the accuracy of the vital sign parameter extraction.

Claims

1. A method for reconstructing weighted signals of non-contact periodic motion, characterized in that: The weighted signal reconstruction method for non-contact periodic motion includes the following steps: 1) Obtain a digital orthogonal baseband signal containing target motion information; 2) The digital quadrature baseband signal obtained in step 1) is calibrated with DC bias to obtain a preprocessed signal; 3) Perform a differential cross-multiplication operation on the preprocessed signal obtained in step 2) to obtain a preliminary demodulated waveform; 4) Obtain the upper envelope, lower envelope, and average envelope of the preliminary demodulated waveform obtained in step 3); 5) Segment the initially demodulated waveform based on the parameters obtained from the initial demodulated waveform; 6) Based on the results obtained in step 4) and step 5), the preliminary demodulated waveform obtained in step 3) is reconstructed to obtain the weighted reconstructed motion signal: 6.1) Based on the results obtained in step 4), obtain the average envelope deviation of each periodic segment, and use this deviation as a weighting parameter: in: It is the weight parameter of the i-th segment; K is the number of different periodic segments; It is the average envelope of the i-th segment; t is the time of the i-th segment; 6.2) Reconstruct the preliminary demodulated waveform obtained in step 3) using the results obtained in step 6.1) and step 5) to obtain the weighted reconstructed motion signal: in: It is the motion signal obtained after weighted reconstruction; It is the weight parameter of the i-th segment; K is the number of different periodic segments; It is the segment obtained by periodically extending the i-th segment. The expression is: in: T is The cycle; It is the i-th segment initially segmented based on the segmentation threshold; t is the time of the i-th segment; j is the period number of the extension.

2. The method for reconstructing weighted signals of non-contact periodic motion according to claim 1, characterized in that: Each sampling point in the digital quadrature baseband signal in step 1) includes I-channel sampling value and Q-channel sampling value; the target motion information includes the target's heartbeat information or the information of the weak motion of the periodic pendulum ball.

3. The method for reconstructing weighted signals of non-contact periodic motion according to claim 2, characterized in that: The specific implementation method of step 2) is as follows: in: I and Q are the I-channel sampled value and Q-channel sampled value of each sampling point in the digital quadrature baseband signal, respectively. and These are the I-channel and Q-channel sampled values ​​after mean reduction preprocessing, respectively. N is the total number of sampling points in the digital quadrature baseband signal.

4. The method for reconstructing weighted signals of non-contact periodic motion according to claim 3, characterized in that: The specific implementation method of step 3) is as follows: in: and These are the I-channel and Q-channel sampled values ​​after mean reduction preprocessing, respectively. and They are and The differential form; These are the sampling times for the I-channel and Q-channel samples. This is the waveform obtained from the initial demodulation.

5. The method for reconstructing weighted signals of non-contact periodic motion according to claim 4, characterized in that: The specific implementation method of step 4) is as follows: 4.1) Find the peak value of the preliminary demodulated waveform obtained in step 3), and obtain the peak and trough of the waveform; 4.2) Based on the peaks and troughs of the waveform obtained in step 4.1), calculate the upper envelope, lower envelope, and average envelope of the preliminary demodulated waveform.

6. The method for reconstructing weighted signals of non-contact periodic motion according to claim 5, characterized in that: The specific implementation method of step 4.1) is as follows: in: This is the waveform obtained from the initial demodulation; and These are the peak value and the amplitude of the peak and trough of the initial demodulated waveform, respectively. and These are the indexes of the peak and trough positions of the initial demodulated waveform, respectively. It is a peak-finding function; The specific implementation method of step 4.2) is as follows: in: and These are the upper and lower envelopes of the initial demodulated waveform, respectively. It is the average envelope of the initial demodulated waveform.

7. The method for reconstructing weighted signals of non-contact periodic motion according to claim 6, characterized in that: The specific implementation method of step 5) is as follows: 5.1) Set the segmentation threshold based on the parameters obtained from the initial demodulated waveform; 5.2) The waveform of the initial demodulation is segmented according to the segmentation threshold obtained in step 5.1).

8. The method for reconstructing weighted signals of non-contact periodic motion according to claim 7, characterized in that: The specific implementation method of step 5.1) is as follows: in: It is the sampling frequency of each sampling point in the digital quadrature baseband signal; It is the motion information frequency of the initial demodulated waveform; It is the segmentation threshold; The specific implementation method of step 5.2) is as follows: in: It is the i-th segment that is initially segmented based on the segmentation threshold.

Images