CN-122004809-A - Multidimensional 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, which comprises the steps of collecting x groups of original vital sign CSI signals, conducting denoising processing and decomposing the original vital sign CSI signals into respiratory signals and heartbeat signals, dividing the respiratory signals and the heartbeat signals into a plurality of motion segment signals, conducting mathematical representation on feature data in each motion segment signal, conducting enhancement signal processing to construct and form a feature matrix C corresponding to each motion segment signal, conducting differential calculation to obtain respiratory sequence signals and heartbeat sequence signals, conducting clustering calculation, and conducting reverse pushing based on clustering index labels to obtain motion categories corresponding to each motion segment signal. The method can identify weak motion interference with high precision, and provide stable and reliable feature support for fine evaluation of respiratory state and abnormal state identification, so that stability of respiratory state analysis in different dynamic scenes is further improved.

Inventors

• QIU TIE
• ZHAO YINGYING
• XIAO FU
• ZHOU XIAOBO

Assignees

• 天津大学

Dates

Publication Date

20260512

Application Date

20260330

Claims (5)

1. 1. A multi-dimensional vital sign signal reconstruction method based on multi-antenna WiFi signals is characterized by comprising the following steps of S1, respectively acquiring x groups of original vital sign CSI signals based on wireless WiFi equipment deployed in an application scene, wherein x is a positive integer and is more than or equal to 2; s2, denoising each group of original vital sign CSI signals and decomposing the same into respiratory signals and heartbeat signals; s3, dividing the respiration signal and the heartbeat signal into a plurality of motion segment signals between adjacent wave troughs respectively; s4, carrying out mathematical representation on characteristic data in each motion segment signal, wherein the characteristic data comprises a phase V, a phase change speed S, CSI time sequence length N, a phase peak value P and time sequence lengths of the position of the phase peak value from the starting point and the end point of the motion segment; s5, respectively carrying out enhancement signal processing on mathematical representations of the motion segment signals, and further constructing and forming a feature matrix C corresponding to the motion segment signals; S6, carrying out differential calculation on all feature matrixes C corresponding to the x group inhalation signals, and further obtaining respiratory sequence signals representing respiratory signal features; Performing differential calculation on all feature matrixes C corresponding to the x groups of heartbeat signals to further obtain heartbeat sequence signals representing the characteristics of the heartbeat signals; And S7, carrying out clustering calculation on the respiratory sequence signals and the heartbeat sequence signals, and reversely deducing based on the clustering index label to obtain the motion category corresponding to each motion segment signal.
2. 2. The method for reconstructing the multi-dimensional vital sign signals based on the multi-antenna WiFi signals of claim 1, wherein in step S7, clustering calculation is performed by using partitioned k-Means, density-based DBSCAN or model-based Gaussian mixture models.
3. 3. The method for reconstructing the multidimensional vital sign signals based on the multi-antenna WiFi signals according to claim 1, wherein in the step S1, a commercial router provided with three antennas is adopted as a WiFi signal transmitting end, a Dell XPS notebook computer carrying Intel 5300 NIC is adopted as a receiving end, the system runs a 64-bit Ubuntu 12.04 LTS operating system, and original vital sign CSI data are collected through a Linux 802.11n CSI tool.
4. 4. The method for reconstructing multi-dimensional vital sign signals based on the multi-antenna WiFi signal of any one of claims 1 to 3, wherein step S2 comprises the steps of S2.1, removing the change exceeding interval in the original vital sign CSI signal by utilizing Hampel filter Then, a value determined by two adjacent points is inserted by adopting linear interpolation every set interval time, wherein alpha represents the central value of the signal, delta represents the discrete degree of the signal, and beta is a multiple of the sensitivity of controlling abnormality judgment; S2.2, performing the dimension reduction filtering processing on the CSI data on the signal obtained in the step S2.1 through a principal component analysis algorithm to obtain a filtered signal; S2.3, selecting a subcarrier with the largest average absolute error according to a set threshold value aiming at the filtered signal; s2.4, aiming at the sub-carriers, using an empirical mode decomposition algorithm to decompose the signals into respiratory signals and heartbeat signals according to the time scale characteristics of the original vital sign CSI signals, thereby obtaining the preprocessing signals ; , wherein, The eigenmode function is represented by a model of the eigenmode, Is the residual, s=1, 2, The respiration signal is characterized by the fact that, The heartbeat signal is characterized.
5. 5. The multi-dimensional vital sign signal reconstruction method based on the multi-antenna WiFi signal according to any one of claims 1 to 3, wherein in step S5, the method for calculating the feature matrix C is as follows: ; Wherein, the And As the weight coefficient, S represents a phase change velocity matrix, V represents the phase matrix, and the phase matrix, E represents the end point of the motion segment signal, P represents the peak point of the motion segment signal, and the peak point And endpoint Respectively at the coordinates And (3) with 。

Description

Multidimensional vital sign signal reconstruction method based on multi-antenna WiFi signals Technical Field The invention relates to the technical field of vital sign sensing, in particular to a multi-dimensional vital sign signal reconstruction method based on multi-antenna WiFi signals. Background Vital signs such as respiration and heart rate are key indicators for assessing the life safety, health status and sleep quality of an individual. The vital sign sensing technology based on the Internet of things provides convenience for medical diagnosis, potential risks can be early warned in time, and universal sensing is realized. In various internet of things devices, commercial WiFi provides a feasible home application scheme for realizing non-contact vital sign sensing by virtue of the deployment convenience. The WiFi signal captures the periodic effects of respiratory and heartbeat-induced chest displacement on the signal, manifested as amplitude or phase sinusoidal-like waveform variations of Channel State Information (CSI). In an ideal perception scenario, the human body is typically located between two transceiving devices, with chest displacement motion on a linear-of-Sight (LoS) path. This scenario maximizes the received signal quality and establishes a clear correlation between the phase change and the chest displacement. However, the LoS in the room is often blocked by an obstacle such as furniture, a wall or a door, as shown in fig. 1, resulting in that the signal needs to reach the receiving end after multipath reflection. Compared with respiratory signals under LoS, multipath reflection can cause significant attenuation of signal energy, weak phase change and interference peak value, and the factors jointly cause motion characteristic distortion, so that reliability of respiratory and heartbeat characteristic extraction is seriously affected. The accuracy of coarse-grained human activity recognition such as running, jumping, sitting and walking can be remarkably improved by enhancing the energy of the transmission signal to compensate propagation loss caused by multipath reflection. However, respiratory induced chest displacements are only between 0.6 mm and 12 mm, while heartbeat-related displacements are more subtle. The vital sign motion only causes a slight phase change, and the phase distortion caused by multipath reflection cannot be effectively resisted by increasing the energy of the source end signal. Currently, methods of enhancing vital sign motion signal characteristics focus mainly on modeling a single timing characteristic of the signal, such as amplitude or phase information, to suppress ambient noise and compensate for signal attenuation. Wherein Fullbreathe constructs conjugate signals by combining phase and amplitude information to enhance motion characteristics and improve region coverage, and Farsense eliminates phase shift and suppresses noise by means of CSI ratio of two antennas, thereby realizing sensing at a longer distance. In addition, some studies model the relationship of the signal to chest motion by the geometric features of multipath reflections, and another class of studies use learning-based methods aimed at recovering phase and amplitude information from severely attenuated signals. The modeling of the single time sequence characteristic based on the CSI can effectively recover the damaged motion characteristic in an ideal static sensing environment, but weak motion interference unavoidable in an actual deployment environment cannot be fully considered. Two types of random weak motion interference exist in the environment generally, namely dynamic change of the background environment, such as curtain shaking or slight displacement of table articles caused by cross wind, and unintentional movement of the human body, such as micro rotation of the head, movement of the body center of gravity or autonomous fine adjustment of limbs. Notably, such weak motion disturbances are not occasional, and even under strict static instructions (such as sleep), they continue to exist as endogenous noise sources and continue throughout the data acquisition cycle. Weak motion interference has the characteristics of tiny amplitude and potential periodicity, and can act on signal phase change together with vital sign motion to generate a series of alternately-appearing wave crests and wave troughs which are mutually overlapped. After the signals are attenuated by multiple reflections, weak motion interference is highly similar to the change characteristics of vital sign motion, so that vital sign features are blurred. Compared with the signals after multipath reflection, the weak motion interference is hidden in the insignificant signal fluctuation, and is difficult to effectively identify and eliminate through a single time sequence characteristic model. Meanwhile, because the influence of weak motion interference on the wireless signal has randomness, nonlinearity and time-varyi