RF-ECG:Heart Rate Variability Assessment Based on COTS RFID Tag Array.85:11 Signal Signal Smoothing measurements Signal Interpolation Signal Preprocessing Activity Detection Signal Segmentation Chest Movement Recovery Moving Effect Respiration Cancellation Elimination DWT-based De-noising Tag Array PCA-based Signal Fusion Topology Template Estimation IBI Segmentation IBl Extraction Fig.8.System Architecture. the maximum reflection phase range is 8.4 radian of the tag at(2,1),which represents the displacement of about 22cm.The estimated displacement is very close to the groundtruth,which is 20cm according to the deployment Therefore,the experiment results validate the effectiveness of our reflection model of the tag array. 4 RF-ECG SYSTEM DESIGN 4.1 System Overview The system architecture of RF-ECG is shown in Figure 8.To perform the HRV assessment,a set of COTS RFID tags are attached in the chest area on the clothes of the human subject,which forms a tag array for comprehensive sensing.A COTS RFID reader is deployed to continuously interrogate these tags by issuing a continuous wave, and collect their backscattered RF-signals within the effective scanning range(1~3m).According to the received RF-signals from the tag array,we leverage three main components in RF-ECG to extract the HRV information, i.e.,Signal Preprocessing,Moving Effect Elimination and IBI Extraction.Signal Preprocessing first filters the received RF-signals from the tag array with smoothing and interpolation,and then detects the activities of human body to segment the signals for the following HRV estimation.Moving Effect Elimination removes the respiration influence by estimating the contour of chest movement and cancels the phase variation caused by the moving effect.IBI Extraction further extracts the Inter-Beat Interval(IBI)from the reflection signal captured by the tag array.Specifically,we first use Discrete Wavelet Transform(DWT)to further reduce the ambient noises by concentrating on the heartbeat frequency band.We then fuse the RF-signals from the 2-dimensional tag array into a signal sequence to generate a clear periodic pattern.We further coarsely segment the fused signal and use the Principal Component Analysis(PCA)to obtain a template to depict the principle features of a heartbeat period.Finally,we estimate a fine-grained IBI segmentation by maximizing the similarity between the signal segments and the template based on dynamic programming.The extracted IBI can be further used for HRV assessment. 4.2 Signal Preprocessing According to the empirical study in Section 3.2,for the RF-signals from the tag array,the heartbeat signal is almost buried by the respiration signal and the ambient noise.Hence,after we obtain the RF-signals from the tag array,it is essential to smooth the collected signal for further extraction of the heartbeat signal.According to [40]. the phase noises of RF-signals follow the Gaussian distribution.Therefore,we use Kalman filter [20]to process Proc.ACM Interact.Mob.Wearable Ubiquitous Technol,Vol.2,No.2,Article 85.Publication date:June 2018.RF-ECG: Heart Rate Variability Assessment Based on COTS RFID Tag Array • 85:11 Signal Smoothing Signal measurements Respiration Cancellation IBI Segmentation Signal Fusion Chest Movement Recovery Signal Interpolation Signal Preprocessing Moving Effect Elimination DWT-based De-noising IBI Extraction Tag Array Topology PCA-based Template Estimation Activity Detection & Signal Segmentation Fig. 8. System Architecture. the maximum reflection phase range is 8.4 radian of the tag at (2, 1), which represents the displacement of about 22cm. The estimated displacement is very close to the groundtruth, which is 20cm according to the deployment. Therefore, the experiment results validate the effectiveness of our reflection model of the tag array. 4 RF-ECG SYSTEM DESIGN 4.1 System Overview The system architecture of RF-ECG is shown in Figure 8. To perform the HRV assessment, a set of COTS RFID tags are attached in the chest area on the clothes of the human subject, which forms a tag array for comprehensive sensing. A COTS RFID reader is deployed to continuously interrogate these tags by issuing a continuous wave, and collect their backscattered RF-signals within the effective scanning range (1 ∼ 3m). According to the received RF-signals from the tag array, we leverage three main components in RF-ECG to extract the HRV information, i.e., Signal Preprocessing, Moving Effect Elimination and IBI Extraction. Signal Preprocessing first filters the received RF-signals from the tag array with smoothing and interpolation, and then detects the activities of human body to segment the signals for the following HRV estimation. Moving Effect Elimination removes the respiration influence by estimating the contour of chest movement and cancels the phase variation caused by the moving effect. IBI Extraction further extracts the Inter-Beat Interval (IBI) from the reflection signal captured by the tag array. Specifically, we first use Discrete Wavelet Transform (DWT) to further reduce the ambient noises by concentrating on the heartbeat frequency band. We then fuse the RF-signals from the 2-dimensional tag array into a signal sequence to generate a clear periodic pattern. We further coarsely segment the fused signal and use the Principal Component Analysis (PCA) to obtain a template to depict the principle features of a heartbeat period. Finally, we estimate a fine-grained IBI segmentation by maximizing the similarity between the signal segments and the template based on dynamic programming. The extracted IBI can be further used for HRV assessment. 4.2 Signal Preprocessing According to the empirical study in Section 3.2, for the RF-signals from the tag array, the heartbeat signal is almost buried by the respiration signal and the ambient noise. Hence, after we obtain the RF-signals from the tag array, it is essential to smooth the collected signal for further extraction of the heartbeat signal. According to [40], the phase noises of RF-signals follow the Gaussian distribution. Therefore, we use Kalman filter [20] to process Proc. ACM Interact. Mob. Wearable Ubiquitous Technol., Vol. 2, No. 2, Article 85. Publication date: June 2018