Evaluation of Heartbeat Signal Extraction Methods Using a 5.8 GHz Doppler Radar System in a Real Application Scenario

Abstract

Doppler-based radar systems have been seen as a promising tool to assess vital signs, since they are capable to monitor the respiratory and cardiac signal remotely, by measuring the chest-wall displacement. However, due to the spectral overlap of these signals, their proper separation is a challenging task. In this paper, we demonstrate the effectiveness of using Discrete Wavelet Transform in the cardiac signal extraction, by comparing this method with other approaches widely used in literature, namely a standalone Band-Pass Filtering, the Ensembled Empirical Mode Decomposition, the Continuous Wavelet Transform and the Wavelet Packet Decomposition. The comparison metrics were defined taking into consideration the heart rate computation accuracy, and also the peak detection consistency to further evaluate the Heart Rate Variability. The efficiency of those methods is also tested considering real application scenarios, characterized by non-controlled monitoring environment conditions and the ability to equally assess the vital signs of different subjects, regardless their physical stature.

Bio-Radar Cardiac Signal Model Used for HRV Assessment and Evaluation Using Adaptive Filtering

“…Thus, the signal at(n) contains both respiratory and cardiac signals, which are then separated in the Cardiac Extraction block. Herein, it is used a BPF combined with Discrete Wavelet Transform (DWT) [13]. We verified in preliminary tests that applying a BPF prior to DWT, attenuates the respiratory component and improves the resultant cardiac signal [13].…”

“…Herein, it is used a BPF combined with Discrete Wavelet Transform (DWT) [13]. We verified in preliminary tests that applying a BPF prior to DWT, attenuates the respiratory component and improves the resultant cardiac signal [13]. In this case, the BPF is a 100 th order FIR filter, with a pass band between 0.7 and 2 Hz.…”

“…The radar setup used to collect the subjects vital signs was composed by a software-defined radio as RF front-end, namely the USRP B210 board from Ettus Research TM , operating in CW mode with a 5.8 GHz carrier frequency. Transmission and reception were performed using two 2×2 antenna arrays [13]. These antennas are circularly and crossed polarized, to avoid the mutual coupling and to improve the path gain, according to [29], [30].…”

“…Several methods have been proposed in literature to separate both bio-signals, which encompass spectral manipulation [3], filtering [4] and signal decomposition [5], [6]. Multi-resolution analysis using wavelets is also an effective method as proved in [7]- [13].…”

“…In [23] this lack of signal resolution is also reported for a radar operating with the same frequency, but an autocorrelation-based algorithm is developed to rectify the missed peaks. In [13], we presented a study focused on the cardiac signal extraction on CW radars operating at 5.8 GHz. Six different methods were applied in radar signals and their effectiveness to extract the cardiac signal was compared.…”