Improving the signal processing path for online heartbeat estimation by using Ultra Wideband radar

Abstract

A number of companies, research establishments and universities worldwide are currently researching the use of Ultra Wideband radar technology to conduct non-contact and non-invasive measurements of heartbeat. Both cardiac and respiratory activity cause a visible and measurable motion to the chest wall. When aiming a radar to a subject's chest, this motion can be recorded and processed to obtain the rates of respiration and heartbeat. This thesis looks into the processing of chest wall displacement data to improve heart rate measurements. At the first stage of the signal processing, the low frequency, high amplitude respiration signal component must be separated from the much smaller heartbeat component. Traditionally, linear Finite Impulse Response (FIR) highpass filters have been used to remove the respiration component. FIR filters have a fixed corner frequency. If the respiration rate is higher than this fixed frequency, FIR filtering can not work. In addition, FIR filters must be quite long for suficient stopband attenuation, and thus they introduce a time delay to the system. Alternative signal processing techniques are based on cancellation of respiration harmonics, which eliminates the system delay and works at any respiration frequency. One such method from the literature was investigated, and this thesis presents a modified version that works better with actual radar data. After separation, the heart rate must be determined. Traditionally, frequency estimation of a sliding window has been used. A different method from the literature is based on adaptive filtering. It transforms the heartbeat signal into a train of unit pulses, each occurring at the start of a heartbeat. Instant heartbeat information is thus gathered, as opposed to the traditional method which yields the mean heart rate during the time window. Some system results were obtained for both constructed test signals and actual radar data. Generally FIR filtering performed better than harmonic cancellation, but the harmonic canceller outperformed FIR filtering when the respiration frequency tended towards the filter's corner frequency. Feasibility of a system that gives instant heart rate measurements in real-time was proven. The improvements on traditional systems are elimination of time delay, enabling separation at high respiration frequencies and giving the instant heart rate rather than an avereraged one.