A millimeter wave radar vital sign signal clustering analysis method

Abstract

The application discloses a kind of millimeter wave radar vital sign signal clustering analysis methods, belong to millimeter wave radar signal processing technical field, this millimeter wave radar vital sign signal clustering analysis method is to extract the amplitude information data of human echo signal obtained by long time monitoring, amplitude information data is divided into time window;The amplitude information data in each time window window is preprocessed;The power spectrum information of echo amplitude of preprocessed amplitude information data is obtained;With typical human respiratory rate as reference frequency selection window, clustering is carried out once;Afterwards, the power spectrum sequence obtained by clustering, the similarity coefficient between each two sequences is analyzed, and a similarity coefficient matrix is constructed;Finally, according to the similarity coefficient matrix, the sequence is clustered twice, and the final clustering result is obtained;Using the application can realize the rapid clustering of millimeter wave radar long-time echo data, identify the data segment in which echo phase is seriously disturbed.

Description

Technical Field

[0001] This invention belongs to the field of millimeter-wave radar signal processing technology, specifically relating to a clustering analysis method for vital sign signals from millimeter-wave radar. Background Technology

[0002] With the increasing demand for health monitoring, vital sign monitoring technology has received growing attention. Traditional contact-based vital sign detection technologies mainly encompass various methods such as electrocardiogram (ECG), photoplethysmography (PPG), and pressure sensors. While these technologies can provide highly accurate results, they have several limitations in practical applications. Specifically, ECG testing requires attaching electrodes to the skin surface; prolonged wear can cause skin discomfort or allergic reactions and may adversely affect sleep quality. PPG technology relies on wearable devices such as wristbands or watches, and these devices may become unstable during exercise or sleep, leading to decreased accuracy. Furthermore, contact-based detection methods are difficult to implement efficiently in situations involving multiple people simultaneously or in complex environments, and the process of wearing and maintaining these devices increases costs and inconvenience.

[0003] Millimeter-wave radar technology, as a non-contact detection method, has shown great potential in the field of vital sign monitoring in recent years. Its basic principle is to emit electromagnetic waves in the millimeter-wave frequency band to illuminate the human chest cavity or body surface, and then extract respiratory and heartbeat information using the reflected echo signals. Compared with traditional vital sign monitoring technologies, millimeter-wave radar does not require direct contact with the human body, avoiding the discomfort and skin irritation associated with traditional contact methods. Furthermore, millimeter-wave radar can penetrate common materials such as clothing and thin blankets, making it suitable for various scenarios. Compared with optical methods, millimeter-wave radar is unaffected by lighting conditions and can operate in complete darkness. Millimeter-wave radar technology is not only suitable for medical scenarios such as hospitals and nursing homes, but also plays an important role in smart homes, automotive safety, and environmentally assisted living. For example, in smart homes, millimeter-wave radar can be used to monitor sleep quality, while in automobiles, it can be used to detect driver vital signs to prevent drowsy driving.

[0004] The core of millimeter-wave radar technology lies in the transmission and reception of high-frequency electromagnetic waves. Millimeter-wave electromagnetic waves (typically 24GHz, 60GHz, or 77GHz) have shorter wavelengths, enabling them to more sensitively detect minute human movements. Current methods for acquiring human vital signs using millimeter-wave radar typically utilize the phase information of the target's range cell in the radar echo. This phase is then unwrapped and differentially processed to obtain the overlapping time-domain waveforms of respiration and heartbeat, which are then separated using filtering and other methods. However, the target's phase in the echo is easily affected by noise and clutter, leading to unwrapping failure and the inability to obtain accurate vital sign information. Furthermore, in applications requiring long-term continuous monitoring, such as sleep monitoring and daily care, routine human movements like turning over or standing up can affect the continuity of the heartbeat and respiration echo phases, resulting in numerous data segments in the echo that deviate from the correct heartbeat and respiration measurements. Therefore, when using millimeter-wave radar to conduct long-term continuous monitoring of the human body, the echo data segments should be clustered to identify those segments whose echo phase is severely interfered with. These segments should be removed during subsequent phase extraction to avoid erroneous heart rate and respiration measurements affecting the assessment of sleep quality, health status, etc. Summary of the Invention

[0005] Based on the problems existing in the above-mentioned background technology, this invention proposes a clustering analysis method for vital sign signals of millimeter-wave radar, which solves the problem that the echo phase of millimeter-wave radar is easily interfered with during long-term monitoring in the prior art.

[0006] The embodiments of the present invention are implemented as follows:

[0007] This invention provides a method for clustering and analyzing vital sign signals using millimeter-wave radar, comprising the following steps:

[0008] S1. Extract the amplitude information data of the human echo signal obtained from long-term monitoring, and divide the amplitude information data into time windows;

[0009] S2. Preprocess the amplitude information data within each time window of step S1;

[0010] S3. Perform a Fourier transform on the amplitude information data after the preprocessing in step S2 and calculate its amplitude square to obtain the power spectrum information of the echo amplitude;

[0011] S4. Using the typical human breathing rate as a reference, set the frequency selection window, and calculate the percentage of energy in the power spectrum obtained in step S3 that is in the reference frequency selection window. Cluster the sequences that exceed the threshold into the same category and remove the sequences that are below the threshold.

[0012] S5. For the power spectrum sequences obtained by clustering in step S4, analyze the similarity coefficient between every two sequences to obtain the similarity coefficient matrix, which is a symmetric matrix;

[0013] S6. Calculate whether each element of the upper triangular part of the similarity coefficient matrix obtained in step S5 is greater than the set threshold. Perform secondary clustering on the two sequences corresponding to the elements that exceed the threshold to complete the final clustering. The results of the final clustering are used for subsequent analysis and evaluation of sleep stages, health status, etc.

[0014] Furthermore, in step S1, the extraction of amplitude information data of human echo signals obtained from long-term monitoring includes performing a maximum search on the echo of each fast time dimension after the distance-dimensional pulse compression processing has been completed. The peak value is the position corresponding to the human target, and the amplitude information of the data at that position is extracted.

[0015] Furthermore, in step S2, the preprocessing includes eliminating static background interference in the signal by mean subtraction, retaining dynamic amplitude changes caused by breathing and heartbeat, and normalizing the amplitude to eliminate the influence of hardware gain on the absolute value of the amplitude of each data segment.

[0016] Furthermore, in step S3, a one-dimensional fast Fourier transform is performed on the preprocessed amplitude information data, and then the square of the transform result is calculated to obtain the power spectrum information.

[0017] Furthermore, in step S5, the similarity coefficient between each pair of sequences is calculated by utilizing cosine similarity.

[0018] The beneficial effects of this invention are: by processing the amplitude of millimeter-wave radar echoes, this invention can eliminate data segments with unstable echo phases caused by normal human movement, thus avoiding their impact on the extraction of heartbeat and respiratory signals; in addition, by using amplitude information for rapid clustering, filters for extracting respiratory signals can be designed specifically in subsequent processing, thereby improving the accuracy of respiratory feature measurement. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly described below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. The above and other objects, features, and advantages of the present invention will become clearer through the accompanying drawings. The same reference numerals indicate the same parts in all the drawings. The drawings are not intentionally drawn to scale to actual dimensions; the focus is on illustrating the main points of the invention.

[0020] Figure 1 This is a flowchart of a clustering analysis method for vital sign signals from millimeter-wave radar.

[0021] Figure 2 This is a schematic diagram of the amplitude signal of a human target located in a millimeter-wave radar echo over a period of approximately 100 minutes.

[0022] Figure 3 This is a schematic diagram of the amplitude signal after preprocessing for a time window.

[0023] Figure 4 The power spectrum of amplitude information data after preprocessing for a time window.

[0024] Figure 5 A schematic diagram showing the percentage of energy in the power spectrum of each time window within the frequency range of 0.1-0.5 Hz relative to the total energy of its power spectrum.

[0025] Figure 6 This is a schematic diagram showing the similarity coefficients of the power spectrum sequences of data from every two time windows after one clustering. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0027] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0028] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0029] Furthermore, the terms "first," "second," etc., are used only to distinguish descriptions and should not be interpreted as indicating or implying relative importance.

[0030] Reference Figure 1 As shown, this embodiment provides a clustering analysis method for millimeter-wave radar vital sign signals, which includes the following steps:

[0031] S1. Extract amplitude information data of human echo signals obtained from long-term monitoring, and divide the amplitude information data into time windows; specifically, in this step, extracting amplitude information data of human echo signals obtained from long-term monitoring includes performing a maximum search on the echo of each fast time dimension after the distance-dimensional pulse compression processing has been completed, and the peak value is the position corresponding to the human target, and extracting the amplitude information of the data at that position. Figure 2 The data shown is the extracted peak amplitude information. The data duration is approximately 100 minutes. The data is divided into 100 time windows with a time window length of 60 seconds each.

[0032] S2. Preprocess the amplitude information data within each time window of step S1; In this step, the preprocessing in step S2 includes eliminating static background interference in the signal by mean subtraction, retaining the dynamic amplitude changes caused by breathing and heartbeat, and normalizing the amplitude to eliminate the influence of hardware gain on the absolute value of the amplitude of each data segment.

[0033] The specific calculation formula is as follows:

[0034] (1)

[0035] (2)

[0036] in, A i ( k ) is the first i Amplitude data within a time window k This indicates the sequence number of the amplitude data; the total number of amplitude data is... K ; The amplitude data after DC removal operation according to formula (1) is used to eliminate the influence of background echo and retain the dynamic amplitude changes caused by breathing and heartbeat. The amplitude data is normalized according to formula (2). The normalization process eliminates the influence of hardware gain on the absolute value of the amplitude of each data segment, so as to facilitate subsequent threshold detection processing. The amplitude information data within each time window is preprocessed to obtain the following: Figure 3 The results are shown.

[0037] S3. Perform a Fourier transform on the amplitude information data preprocessed in step S2 and calculate its amplitude square to obtain the power spectrum information of the echo amplitude; specifically, the formula for calculating the power spectrum information of the echo amplitude is:

[0038] (3)

[0039] in, P i( n The power spectrum information of the echo amplitude is shown. n Indicates the frequency sequence number. The power spectrum result of the amplitude information data after one time window preprocessing is as follows: Figure 4 As shown.

[0040] S4. Using the typical human breathing rate as a reference, set the frequency selection window, and calculate the percentage of energy in the power spectrum obtained in step S3 that is in the reference frequency selection window. Cluster the sequences that exceed the threshold into the same category and remove the sequences that are below the threshold.

[0041] Specifically, in this step, the human respiratory rate typically falls within the range of 0.1-0.5 Hz. The percentage of energy in the power spectrum within this frequency range for each time window is calculated, and the results are as follows: Figure 5 As shown, window data with energy ratios below a threshold are discarded and not processed further, while other window data exceeding the threshold are clustered into the same class, and their power spectrum sequences are used... P m ( n ) indicates that among them m = 0, 1, 2, …, M –1 represents the remaining window number after one clustering.

[0042] S5. For the power spectrum sequences obtained by clustering in step S4, analyze the similarity coefficient between every two sequences to obtain the similarity coefficient matrix, which is a symmetric matrix;

[0043] Specifically, the similarity coefficient between each pair of sequences is calculated using cosine similarity, and the formula is as follows:

[0044] (4)

[0045] in, P x ( n )and P y ( n ) represent the first clustering result. x The first window and the first y Power spectrum sequence of windows, The cosine similarity between the power spectrum sequences of these two windows. n = 0, 1, 2, …, N -1 indicates the frequency index; the similarity coefficient matrix is ​​as follows: Figure 6 As shown.

[0046] S6. Calculate whether each element of the upper triangular part of the similarity coefficient matrix obtained in step S5 is greater than the set threshold. The set threshold condition is that the similarity coefficient is 0.5. The two sequences corresponding to the elements that exceed the threshold are classified into the same class and secondary clustering is performed to complete the final clustering. The human echo data is finally clustered into 9 classes, each class corresponding to different respiratory characteristics, which can be used for subsequent analysis and evaluation of sleep stages, health status, etc.

[0047] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A clustering analysis method for vital sign signals from millimeter-wave radar, characterized in that, Includes the following steps: S1. Extract the amplitude information data of the human echo signal obtained from long-term monitoring, and divide the amplitude information data into time windows; S2. Preprocess the amplitude information data within each time window of step S1; S3. Perform a Fourier transform on the amplitude information data after the preprocessing in step S2 and calculate its amplitude square to obtain the power spectrum information of the echo amplitude; S4. Using the typical human breathing rate as a reference, set the frequency selection window, and calculate the percentage of energy in the power spectrum obtained in step S3 that is in the reference frequency selection window. Cluster the sequences that exceed the threshold into the same category and remove the sequences that are below the threshold. S5. For the power spectrum sequences obtained by clustering in step S4, analyze the similarity coefficient between every two sequences to obtain the similarity coefficient matrix, which is a symmetric matrix; S6. Calculate whether each element of the upper triangular part of the similarity coefficient matrix obtained in step S5 is greater than the set threshold. Perform secondary clustering on the two sequences corresponding to the elements that exceed the threshold to complete the final clustering. The results of the final clustering are used for subsequent analysis and evaluation of sleep stages and health status.

2. The millimeter-wave radar vital sign signal clustering analysis method according to claim 1, characterized in that, In step S1, the extraction of amplitude information data of human echo signals obtained from long-term monitoring includes performing a maximum search on the echo of each fast time dimension after the distance-dimensional pulse compression processing has been completed. The peak value is the position corresponding to the human target, and the amplitude information of the data at that position is extracted.

3. The millimeter-wave radar vital sign signal clustering analysis method according to claim 1, characterized in that, In step S2, the preprocessing includes eliminating static background interference in the signal by mean subtraction, retaining dynamic amplitude changes caused by breathing and heartbeat, and normalizing the amplitude to eliminate the influence of hardware gain on the absolute value of the amplitude of each data segment.

4. The millimeter-wave radar vital sign signal clustering analysis method according to claim 3, characterized in that, In step S3, a one-dimensional fast Fourier transform is performed on the preprocessed amplitude information data, and then the square of the transform result is calculated to obtain the power spectrum information.

5. The millimeter-wave radar vital sign signal clustering analysis method according to claim 3, characterized in that, In step S5, the similarity coefficient between each pair of sequences is calculated using cosine similarity.

Images