Multi-dimensional vital sign signal reconstruction method based on multi-antenna WiFi signals

Abstract

The invention relates to a multi-dimensional vital sign signal reconstruction method based on multi-antenna WiFi signals. The method comprises the following steps: acquiring x groups of original vital sign CSI signals; de-noising processing is carried out and the signals are decomposed into respiratory signals and heartbeat signals; segmenting into a plurality of motion segment signals; performing mathematical representation on feature data in each motion fragment signal; enhanced signal processing is carried out, and then a feature matrix C corresponding to each motion fragment signal is constructed and formed; differential calculation is carried out, and then a respiration sequence signal and a heartbeat sequence signal are obtained; clustering calculation is carried out, and the motion category corresponding to each motion fragment signal is obtained through reverse deduction based on the clustering index labels. According to the method, weak motion interference can be recognized with high precision, stable and reliable feature support is provided for fine evaluation of the breathing state and recognition of the abnormal state, and therefore the stability of breathing state analysis in different dynamic scenes is further improved.

Description

Technical Field

[0001] This invention relates to the field of vital sign sensing technology, and in particular to a method for reconstructing multidimensional vital sign signals based on multi-antenna WiFi signals. Background Technology

[0002] Vital signs such as respiration and heart rate are key indicators for assessing an individual's safety, health status, and sleep quality. IoT-based vital sign sensing technology facilitates medical diagnosis and provides timely warnings of potential risks, achieving universal sensing capabilities. Among various IoT devices, commercial WiFi, with its ease of deployment, offers a feasible home application solution for contactless vital sign sensing. WiFi signals capture the periodic influence of chest displacement caused by respiration and heartbeat on the signal, manifesting as sinusoidal waveform changes in amplitude or phase of Channel State Information (CSI). In an ideal sensing scenario, the human body is typically positioned between two transceiver devices, ensuring that chest displacement is along a line-of-sight (LoS) path. This scenario maximizes signal quality and establishes a clear correlation between phase changes and chest displacement. However, indoor LoS is often obstructed by obstacles such as furniture, walls, or doors. Figure 1 As shown, this results in the signal needing to undergo multipath reflection before reaching the receiver. Compared to respiratory signals under LoS, multipath reflection leads to significant signal energy attenuation, weak phase changes, and interference peaks. These factors collectively cause motion feature distortion, thus severely affecting the reliability of respiratory and heartbeat feature extraction.

[0003] By enhancing the transmitted signal energy to compensate for propagation loss caused by multipath reflection, the accuracy of recognizing coarse-grained human activities such as running, jumping, sitting, and walking can be significantly improved. However, the chest displacement induced by breathing is only between 0.6 mm and 12 mm, while the displacement related to heartbeat is even smaller. Vital movement only causes small phase changes, and simply increasing the source signal energy cannot effectively resist the phase distortion introduced by multipath reflection. Currently, methods to enhance the characteristics of vital movement signals mainly focus on modeling single temporal features of the signal, such as amplitude or phase information, to suppress environmental noise and compensate for signal attenuation. Among them, Fullbreathe constructs a conjugate signal by combining phase and amplitude information to enhance motion characteristics and improve area coverage; while Farsense uses the CSI ratio of two antennas to eliminate phase shift and suppress noise, thereby achieving perception at greater distances. In addition, some studies model the relationship between signals and chest movement through the geometric features of multipath reflection; another type of research uses learning-based methods to recover phase and amplitude information from severely attenuated signals. Single temporal feature modeling based on CSI can effectively recover damaged motion features in an ideal static perception environment, but it fails to fully consider the unavoidable weak motion interference in the actual deployment environment.

[0004] Two types of random, weak motion interference are prevalent in the environment: one is dynamic changes in the background environment, such as curtains swaying due to drafts or slight displacement of objects on a table; the other is unconscious human activity, such as minute head movements, shifts in the body's center of gravity, or voluntary fine-tuning of the limbs. It is noteworthy that these weak motion interferences are not accidental; even under strict static commands (such as sleep), they persist as endogenous noise sources throughout the entire data acquisition cycle. Weak motion interference is characterized by its small amplitude and potential periodicity, interacting with vital sign movements to affect signal phase changes, producing a series of alternating peaks and troughs that superimpose. After multiple reflections and attenuation, the characteristics of weak motion interference and vital sign movements become highly similar, blurring the vital sign features. Compared to the signal after multipath reflection, these weak motion interferences are hidden within insignificant signal fluctuations, making them difficult to effectively identify and eliminate using a single time-series feature model.

[0005] Meanwhile, due to the random, nonlinear, and time-varying characteristics of the impact of weak motion interference on wireless signals, respiratory-related signals are easily disturbed, distorted, or partially submerged in both the time and amplitude domains. This makes it difficult for existing signal characterization methods to stably depict the fine-grained dynamic changes during the respiratory process. Specifically, existing methods mostly focus on coarse-grained estimation of signal peak values, which is effective to some extent for the overall trend of respiratory signals. However, when weak motion interference is present, the extracted respiratory waveforms often exhibit problems such as peak distortion, amplitude drift, or phase discontinuity, thus limiting the ability to perform fine-grained analysis of the respiratory state.

[0006] Respiratory status encompasses not only the number of breaths per unit time but also key physiological characteristics such as the ratio of exhalation to inhalation, changes in respiratory depth, and the degree of rapidity or slowness of the respiratory rhythm. These characteristics are typically reflected in the peak and trough shapes of the respiratory waveform, the rise and fall slopes, the range of amplitude variations, and the non-stationarity differences between adjacent respiratory cycles. However, under the influence of weak motion disturbances, existing signal representation methods struggle to reliably separate and model these multidimensional features, leading to difficulties in distinguishing changes in respiratory depth, blurring the boundaries between exhalation and inhalation, and hindering continuous tracking of respiratory rhythm changes. Consequently, these limitations affect the refined assessment of respiratory status and the identification of anomalies. Summary of the Invention

[0007] The technical problem to be solved by the present invention is to provide a multi-dimensional vital sign signal reconstruction method based on multi-antenna WiFi signals, which can accurately identify weak motion interference, provide stable and reliable feature support for the refined assessment of respiratory status and the identification of abnormal status, thereby further improving the stability of respiratory status analysis in different dynamic scenarios.

[0008] The technical solution adopted by this invention to solve the above-mentioned technical problems is: a method for reconstructing multidimensional vital sign signals based on multi-antenna WiFi signals, including the following steps.

[0009] S1. Collect x sets of original vital signs CSI signals from wireless WiFi devices deployed in the application scenario, where x is a positive integer and x≥2;

[0010] S2. Denoise the original CSI signals of each group of vital signs and decompose them into respiratory signals and heartbeat signals.

[0011] S3. Divide the respiratory signal and heartbeat signal into multiple motion segment signals located between adjacent troughs;

[0012] S4. Mathematically represent the feature data in each motion segment signal. The feature data includes phase V, phase change rate S, CSI time sequence length N, phase peak P, and the time sequence length of the phase peak position from the start and end points of the motion segment.

[0013] S5. The mathematical representations of each motion segment signal are subjected to signal enhancement processing to construct the feature matrix C corresponding to each motion segment signal.

[0014] S6. Perform difference calculation on all feature matrices C corresponding to x groups of respiratory signals to obtain respiratory sequence signals that characterize the respiratory signal features;

[0015] Differential calculations are performed on all feature matrices C corresponding to x groups of heartbeat signals to obtain heartbeat sequence signals that characterize the heartbeat signal features;

[0016] S7. Perform clustering calculations on the respiratory sequence signal and heartbeat sequence signal, and inversely deduce the motion category corresponding to each motion segment signal based on the clustering index label.

[0017] As an improvement, in step S7, clustering calculations are performed using partitioned k-Means, density-based DBSCAN, or model-based Gaussian mixture models.

[0018] As an improvement, in step S1, a commercial router equipped with three antennas is used as the WiFi signal transmitter, and a Dell XPS laptop equipped with an Intel 5300 NIC is used as the receiver. The system runs a 64-bit Ubuntu 12.04LTS operating system and collects raw vital signs CSI data through the Linux 802.11n CSI tool.

[0019] As an improvement, step S2 includes step

[0020] S2.1. Use a Hampel filter to remove changes exceeding the specified intervals from the original vital signs CSI signal. The abnormal points are identified, and then a value determined by two adjacent points is inserted using linear interpolation at set intervals; where 𝛼 represents the center value of the signal, 𝛿 represents the degree of dispersion of the signal, and 𝛽 is a multiple that controls the sensitivity of the abnormality detection.

[0021] S2.2. The signal obtained in step S2.1 is subjected to CSI data dimensionality reduction and filtering processing by principal component analysis algorithm to obtain the filtered signal;

[0022] S2.3 For the filtered signal, select the subcarrier with the largest mean absolute error according to the set threshold;

[0023] S2.4. For subcarriers, the empirical mode decomposition algorithm is used to decompose the signal into respiratory and heartbeat signals based on the time-scale characteristics of the original vital signs CSI signal, thereby obtaining the preprocessed signal. ;

[0024] ,in, Represents the intrinsic modulus function. These are residuals, s=1, 2. Characterizing respiratory signals, It represents the heartbeat signal.

[0025] As an improvement, in step S5, the characteristic matrix C is calculated as follows:

[0026] ;

[0027] in, and Here are the weighting coefficients, and S represents the phase change rate matrix. V represents the phase matrix. E represents the end point of the motion segment signal, and P represents the peak point of the motion segment signal. and the end point Then they are located at coordinates and .

[0028] Compared with existing technologies, the advantages of this invention are as follows: The multi-dimensional vital sign signal reconstruction method based on multi-antenna WiFi signals in this invention enhances the uniqueness of vital sign motion in signal representation by performing multi-dimensional temporal and spatial feature amplification on the original vital sign signals, effectively improving the separability and stability of vital sign features under different motion modes. Based on the above feature enhancement mechanism, it can accurately identify weak motion interference, providing stable and reliable feature support for refined assessment of respiratory status and identification of abnormal states, thereby further improving the stability of respiratory status analysis under different dynamic scenarios. Attached Figure Description

[0029] Figure 1 A schematic diagram illustrating the propagation of vital signs signals through multiple reflections.

[0030] Figure 2 This is a flowchart of a method for reconstructing multidimensional vital sign signals based on multi-antenna WiFi signals in an embodiment of the present invention.

[0031] Figure 3 This is a schematic diagram of motion segment signals in an embodiment of the present invention.

[0032] Figure 4 This is a schematic diagram illustrating the misalignment differential calculation of multi-antenna signals in an embodiment of the present invention.

[0033] Figure 5 The image is a signal diagram that has not been processed by the multi-dimensional vital sign signal reconstruction method based on multi-antenna WiFi signals in the embodiments of the present invention.

[0034] Figure 6 This is a signal diagram processed by the multi-dimensional vital sign signal reconstruction method based on multi-antenna WiFi signals in this embodiment of the invention. Detailed Implementation

[0035] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments.

[0036] like Figure 2 As shown, the multidimensional vital sign signal reconstruction method based on multi-antenna WiFi signals in this embodiment includes the following steps S1 to S7.

[0037] S1. Collect x sets of raw vital signs CSI signals from wireless WiFi devices deployed in the application scenario, where x is a positive integer and x≥2.

[0038] The specific value of x depends on the number of antennas on the wireless WiFi device. In this implementation, a commercial router equipped with three antennas is used as the WiFi signal transmitter, and a Dell XPS laptop equipped with an Intel 5300 NIC is used as the receiver. The system runs a 64-bit Ubuntu 12.04 LTS operating system and collects raw vital signs CSI data through the Linux 802.11n CSI tool.

[0039] S2. Preprocess each group of original vital signs CSI signals to obtain preprocessed signals.

[0040] Specifically, in this embodiment, each set of original vital signs CSI signals is denoised and decomposed into respiratory signals and heartbeat signals.

[0041] S2.1. Use a Hampel filter to remove changes exceeding the specified intervals from the original vital signs CSI signal. The system identifies anomalies and then uses linear interpolation to insert a value determined by two adjacent points at set intervals. In this embodiment, the set interval is 0.05s, but it can be set according to specific circumstances. Here, 𝛼 represents the center value of the signal, 𝛿 represents the dispersion of the signal, and 𝛽 is a multiple controlling the sensitivity of anomaly detection.

[0042] S2.2. The signal obtained in step S2.1 is subjected to CSI data dimensionality reduction and filtering processing by principal component analysis algorithm to eliminate redundant dimensions, filter out the parts that are not related to the principal components, extract the important features of breathing and heartbeat in the CSI signal to remove the largest noise, and thus obtain the filtered signal.

[0043] S2.3 For the filtered signal, select the subcarrier with the largest mean absolute error according to the set threshold. The subcarrier has better sensitivity of CSI phase difference, which is convenient for subsequent vital sign detection.

[0044] S2.4. For subcarriers, the empirical mode decomposition algorithm is used to decompose the signal into respiratory and heartbeat signals based on the time-scale characteristics of the original vital signs CSI signal, thereby obtaining the preprocessed signal. The decomposed respiratory and heartbeat signals contain only a single frequency and retain the local features of the original signals at different time scales.

[0045] ,in, Represents the intrinsic modulus function. These are residuals, s=1, 2. Characterizing respiratory signals, It represents the heartbeat signal.

[0046] S3. The respiratory and heartbeat signals, after the aforementioned preprocessing, generate a sinusoidal phase waveform, reflecting the phase changes of vital signs. Its characteristic is that peaks and troughs alternate over time.

[0047] Respiratory and cardiac signals are each segmented into multiple motion segments located between adjacent troughs. Each motion segment moves from one trough to another and contains only one peak. Each motion segment includes various features, such as phase V, phase change rate S, CSI time series length N, phase peak value P, and the time series length of the phase peak position from the start and end points of the motion segment. These features collectively provide a more detailed description than a single phase feature, supporting fine-grained analysis of respiratory and cardiac states.

[0048] S4. To accurately describe the morphological characteristics of motion segment signals, a formal definition of motion segment signals is provided, and its mathematical expression and corresponding shape diagram are shown below. Figure 3 As shown, the characteristic data in each motion segment signal are mathematically represented, including phase V, phase change rate S, CSI time series length N, phase peak value P, and the time series length of the phase peak position from the start and end points of the motion segment. It is used to characterize the instantaneous phase amplitude characteristics of wave motion. The slope represents the instantaneous velocity characteristics of the wave motion at each moment of the motion segment signal. The CSI time series length N represents the duration of a complete wave motion. The amplitude of the phase peak P... The maximum amplitude of the fluctuation is characterized, while the temporal length of the phase peak position from the start and end points of the motion segment records the time required to reach that peak. This richer feature data can effectively compensate for attenuation in the signal propagation path and mitigate the destructive effects of multipath reflections. Therefore, by integrating these features to construct new signals, the varying characteristics within them can be amplified.

[0049] S5. The mathematical representations of each motion segment signal are subjected to signal enhancement processing to construct the feature matrix C corresponding to each motion segment signal. The motion segment signals are placed in a coordinate system with the starting point as the origin. The coordinates of the starting point O are set as follows: peak point and the end point The corresponding coordinates are respectively and Based on these coordinates, features such as the phase V, phase change rate S, CSI time series length N, phase peak P, and the time series length of the phase peak position from the start and end points of the motion segment signal, as mentioned earlier, can be mathematically characterized using parameters such as width, height, peak position, change amplitude, and velocity, thus collectively depicting its unique shape. To integrate this multidimensional information, a feature matrix is ​​constructed. It further integrates into the finish line With peak point The aim is to expand the representational capability of each motion segment signal along both the vertical and horizontal axes, thereby enhancing and amplifying the features.

[0050] Specifically, in this embodiment, the feature matrix C is calculated as follows:

[0051] ;

[0052] in, and Here are the weighting coefficients, and S represents the phase change rate matrix. V represents the phase matrix. E represents the end point of the motion segment signal, and P represents the peak point of the motion segment signal. and the end point Then they are located at coordinates and .

[0053] In the matrix In the mathematical expression, the first column consists of two parts: the product of the phase change rate (slope) at each moment and the CSI time sequence length N (motion segment signal length), and the phase value and peak point at that moment. The first column contains the product of the position coordinates. The second column contains the product of the phase change rate (slope) and the position coordinates of the endpoint E of the motion segment signal, as well as the product of the phase value at each point and the height of the motion segment signal phase amplitude. When the matrix... When mapped to wave motion, the first column reflects the relationship between velocity and duration, the amplitude of motion, and the time required to reach its peak; the second column depicts the correlation between the amplitude of the wave motion's progress and its end, as well as its overall amplitude and peak characteristics.

[0054] matrix By integrating the multidimensional temporal features and interrelationships of motion segment signals, the constructed new signal can comprehensively capture the rise amplitude, velocity, duration, and peak point of wave motion. This richer information can effectively compensate for signal attenuation during propagation and mitigate the destructive effects of multipath reflection, thus overcoming the limitations of traditional single-feature signals which are susceptible to path loss and signal attenuation. Finally, the feature matrix... The two-dimensional data was reconstructed, and the reconstructed signal enhanced the sharpness of the peaks, thereby improving the recognizability of motion features. This feature matrix... This study can initially reveal the differences between valuable chest movements and similar interfering movements, laying the foundation for improving their distinguishability. For example, the unique characterization of temporal features can depict the dynamic changes of respiratory signals at different time stages, thereby enabling fine-grained analysis of respiratory states, which at least include the duration ratio of inhalation to exhalation, changes in respiratory depth, and the speed characteristics of respiratory rhythm.

[0055] S6. Under weak motion interference, breathing, heartbeat, and other random movements with similar fluctuation patterns all affect the received signal. Although the construction of a new signal can initially amplify the features, it is still difficult to effectively distinguish them due to the highly concealed nature of the weak motion patterns.

[0056] The WiFi device is equipped with x antennas. Motion segments corresponding to the reconstructed signals are extracted from the received signals of each antenna. Within the same time period, the signal changes of all antennas are affected by the same motion. However, due to differences in antenna position and signal reception timing, the change characteristics of the received signals from each antenna are both similar and slightly different. These characteristics collectively characterize a specific motion pattern within that time period. Weak motion interference typically lacks continuity. Therefore, motion segments detected at adjacent moments after the interference ends will correspond to the breathing or heartbeat activity that is expected to be captured.

[0057] Based on this, in this embodiment, the difference calculation is performed on all feature matrices C corresponding to x groups of respiratory signals to obtain respiratory sequence signals that characterize the respiratory signal features; the difference calculation is performed on all feature matrices C corresponding to x groups of heartbeat signals to obtain heartbeat sequence signals that characterize the heartbeat signal features.

[0058] like Figure 4 As shown, differential calculation can further enhance the distinguishability between different movements.

[0059] The difference calculation algorithm involved in this step can be any existing difference algorithm.

[0060] S7. Perform clustering calculations on respiratory and heartbeat sequence signals, and inversely deduce the motion category corresponding to each motion segment signal based on the clustering index labels. This allows for the differentiation of motion categories such as stillness, arm movement, coughing, leg movement, and turning over.

[0061] Specifically, clustering calculations can be performed using partitioned k-Means, density-based DBSCAN, or model-based Gaussian mixture models. In practical applications, these clustering algorithms can be used separately for calculations, and based on the final motion category classification results, the clustering algorithm best suited to the application environment can be determined.

[0062] Figure 6 The image shows a signal image processed by the multi-dimensional vital sign signal reconstruction method based on multi-antenna WiFi signals in this embodiment of the invention. A comparison is made based on this signal image. Figure 5 The signal diagram shows that multi-dimensional temporal and spatial features of vital signs are amplified, enhancing the uniqueness of vital signs in signal representation.

[0063] This invention presents a multi-dimensional vital sign signal reconstruction method based on multi-antenna WiFi signals. By amplifying the original vital sign signals through multi-dimensional temporal and spatial feature amplification, the method enhances the uniqueness of vital sign motion in signal representation, effectively improving the separability and stability of vital sign features under different motion modes. Based on this feature enhancement mechanism, it can accurately identify weak motion interference, providing stable and reliable feature support for refined assessment of respiratory status and identification of abnormal states, thereby further improving the stability of respiratory status analysis under different dynamic scenarios.

[0064] This invention presents a multi-dimensional vital sign signal reconstruction method based on multi-antenna WiFi signals. By integrating key variation features such as rise rate, amplitude, width, height, and peak position, it constructs an enhanced new signal representation. This fine-grained representation significantly improves the uniqueness of respiratory and cardiac motion characteristics. These features complement each other, enabling the model to comprehensively understand signal change patterns from multiple perspectives. This enhances the system's sensitivity and adaptability to signal changes, especially when dealing with weak motion interference, allowing for better differentiation of different physiological states and further strengthening the system's robustness in complex environments.

[0065] This invention proposes a segmented processing method for signal features to capture subtle motion changes. It independently analyzes changes in each time segment at a higher temporal resolution, avoiding the problem of the overall signal being "diluted" by noise or motion interference. Through a multi-antenna reconstructed signal misalignment differential mechanism, it enhances the consistency of the same motion pattern and amplifies the differences between different motion patterns, achieving accurate identification and effective filtering of weak motion interference. Simultaneously, it precisely locates the specific position of anomalies in the signal, providing clear guidance for subsequent feature recovery and respiratory state analysis.

[0066] This invention analyzes the unique characteristics of reconstructed vital signs signals to perform fine-grained characterization of respiratory status. The respiratory status includes at least the ratio of the duration of the inhalation and exhalation phases, changes in respiratory depth, and the rapid or slow characteristics of the respiratory rhythm. This enables refined analysis of the respiratory process and identification of abnormal states, identifying potential health problems or predicting future health status.

Claims

1. A method for reconstructing multidimensional vital sign signals based on multi-antenna WiFi signals, characterized in that: Including steps S1. Collect x sets of original vital signs CSI signals from wireless WiFi devices deployed in the application scenario, where x is a positive integer and x≥2; S2. Denoise the original CSI signals of each group of vital signs and decompose them into respiratory signals and heartbeat signals. S3. Divide the respiratory signal and heartbeat signal into multiple motion segment signals located between adjacent troughs; S4. Mathematically represent the feature data in each motion segment signal. The feature data includes phase V, phase change rate S, CSI time sequence length N, phase peak P, and the time sequence length of the phase peak position from the start and end points of the motion segment. S5. The mathematical representations of each motion segment signal are subjected to signal enhancement processing to construct the feature matrix C corresponding to each motion segment signal. S6. Perform difference calculation on all feature matrices C corresponding to x groups of respiratory signals to obtain respiratory sequence signals that characterize the respiratory signal features; Differential calculations are performed on all feature matrices C corresponding to x groups of heartbeat signals to obtain heartbeat sequence signals that characterize the heartbeat signal features; S7. Perform clustering calculations on the respiratory sequence signal and heartbeat sequence signal, and inversely deduce the motion category corresponding to each motion segment signal based on the clustering index label.

2. The method for reconstructing multidimensional vital sign signals based on multi-antenna WiFi signals according to claim 1, characterized in that: In step S7, clustering calculations are performed using partitioned k-Means, density-based DBSCAN, or model-based Gaussian mixture models.

3. The method for reconstructing multidimensional vital sign signals based on multi-antenna WiFi signals according to claim 1, characterized in that: In step S1, a commercial router equipped with three antennas is used as the WiFi signal transmitter, and a Dell XPS laptop equipped with an Intel 5300 NIC is used as the receiver. The system runs a 64-bit Ubuntu 12.04 LTS operating system and collects raw vital signs CSI data through the Linux 802.11n CSI tool.

4. The method for reconstructing multidimensional vital sign signals based on multi-antenna WiFi signals according to any one of claims 1 to 3, characterized in that: Step S2 includes the following steps S2.

1. Use a Hampel filter to remove changes exceeding the specified intervals from the original vital signs CSI signal. The abnormal points are identified, and then a value determined by two adjacent points is inserted using linear interpolation at set intervals; where 𝛼 represents the center value of the signal, 𝛿 represents the degree of dispersion of the signal, and 𝛽 is a multiple that controls the sensitivity of the abnormality detection. S2.

2. The signal obtained in step S2.1 is subjected to CSI data dimensionality reduction and filtering processing by principal component analysis algorithm to obtain the filtered signal; S2.3 For the filtered signal, select the subcarrier with the largest mean absolute error according to the set threshold; S2.

4. For subcarriers, the empirical mode decomposition algorithm is used to decompose the signal into respiratory and heartbeat signals based on the time-scale characteristics of the original vital signs CSI signal, thereby obtaining the preprocessed signal. ; ,in, Represents the intrinsic modulus function. These are residuals, s=1, 2. Characterizing respiratory signals, It represents the heartbeat signal.

5. The method for reconstructing multidimensional vital sign signals based on multi-antenna WiFi signals according to any one of claims 1 to 3, characterized in that: In step S5, the characteristic matrix C is calculated as follows: ； in, and Here are the weighting coefficients, and S represents the phase change rate matrix. V represents the phase matrix. E represents the end point of the motion segment signal, and P represents the peak point of the motion segment signal. and the end point Then they are located at coordinates and .

Images