Research On Wireless Physical Sign Sensing For Health Monitoring

With the aging of the population and the change of living environment and eating patterns,chronic diseases have presented a trend of high incidence,and thus resulting in the growing health-care burden and the correspondingly increasing demand for health monitoring.Meanwhile,the three years of COVID-19 has further increased the demand for daily health monitoring,and people’s attention to health has increased to an unprecedented level.However,the traditional medical model is mainly focus on diagnosis and treatment of diseases,which cannot realize long-term continuous health monitoring.Therefore,there is an urgent need for the exploration of a new model of long-term,continuous and efficient health monitoring because of a higher incidence of chronic diseases,the surging demand for health monitoring and the limitations of traditional medical model.Wireless sensing technologies provide a new opportunity for health monitoring due to its non-contact and all-weather characteristics.Wi Fi-based contactless sensing has become a promising way to realize wireless sensing technologies by virtue of its advantages in universality,ease of use and cost.In this dissertation,we fulfill the theoretical and application research focusing on the key techniques of wireless physical sign sensing by Wi Fi-based sensing.The main contributions of this dissertation are summarized as follows:(1)A phase calibration method based on reference channel is proposed,which makes the phase information of channel state information(CSI)obtained from Wi Fi devices be applicable to sensing applications.The types and characteristics of phase offsets in CSI are analyzed systematically.Two different ways of the phase calibration method based on reference channel are proposed for different scenarios by combining theoretical analysis and experimental verification.The first way is to use the coaxial line as the reference channel,and the second way is to use a pair of antennas as the reference channel.Moreover,the sensing characteristics of CSI amplitude,phase and Doppler shift after calibration are also determined.First,there is a complementary relationship between the amplitude and phase information of CSI after calibration,and the complementarity between the two can improve the sensing accuracy of small-scale movements such as respiration.Second,the Doppler shift extracted from the calibrated CSI can be used to estimate the direction and speed of target motion.(2)A subcarrier combination method based on maximum ratio combining(MRC)is proposed,which make the most of the spatial diversity and the frequency diversity of Wi Fi signals to further improve the sensing accuracy.Specifically,the CSI obtained from different antennas and subcarriers at the same time are regarded as multiple copies from different branches but carrying the same information.The CSI information of all antennas and subcarriers are combined by MRC method to obtain the combined signal with the maximum signal to noise ratio(SNR).Based on this method,an exercise monitoring and assessment system(EMAS)based on Wi Fi for home-based respiratory rehabilitation exercises is implemented on a self-designed compact portable prototype.This system not only provides real-time visual feedback for rehabilitation exercise training,but also extracts the exercise duration,exercise intensity and breathing changes to evaluate the effect of rehabilitation exercise training.EMAS provides a practical scheme for daily home rehabilitation exercise monitoring.(3)A breathing pattern recognition method based on the combined model of Convolutional Neural Network and Long Short-Term Memory Network(CNN-LSTM)is proposed,which avoids the complexity and inefficiency of manual feature extraction and makes up for the inherent defects of a single network model.Using this model,the first Wi Fi-based contactless breathing pattern recognition system is realized on the self-designed Wi Fi transceiver integrated prototype.This system firstly uses a series of data processing methods to reliably extract different types of breathing patterns data from the raw CSI data,and then apply the proposed CNN-LSTM model to recognize automatically breathing patterns,whose recognition accuracy,precision,recall and F1-score were 97.8%,97.9%,97.8% and 97.8%,respectively.Moreover,the model was further evaluated on two new datasets with an accuracy of 97.2% and 100%,respectively.Experiments on different datasets manifest the accuracy and robustness of the system,which provides a potential tool for assisting the diagnosis of diseases related to breathing patterns.(4)A vital sign extraction method based on variational mode decomposition(VMD)is proposed,which can monitor continuously human respiration and heartbeat simultaneously by using Wi Fi signals.To solve the problem that heartbeat signals are easily drowned by respiratory signals and noise,the high directivity antennas are proposed to reduce interference and improve the SNR of received signals.To separate respiratory and heartbeat signals,a vital sign extraction method based on VMD is proposed.The respiratory signals and heartbeat signal are extracted from the received signals respectively using this method,and the respiratory rate and heart rate are acquired by fast Fourier transform(FFT)method.Based on the above methods,a contactless vital signs monitoring system based on Wi Fi devices is developed.Compared with the contact respiratory sensor and electrocardiogram(ECG)sensor,the overall estimation accuracy of the proposed system reached 99.37% and98.59%,respectively.And the maximum error in the estimation of respiratory rate and heart rate was 1 bpm and 5 bpm.The results show that the system can monitor accurately human respiratory rate and heart rate with performance comparable to that of contact respiratory sensor and ECG sensor,and provides a method for continuous,non-invasive,and accurate vital signs monitoring.In conclusion,this dissertation carries out research on wireless physical sign sensing method and its application for health monitoring based on Wi Fi sensing techniques,which makes positive contributions to the establishment of a new contactless and accurate health monitoring mode.