Respiration Detection Algorithm Using UWB Impulse Radar

Non-contact through-wall life detection is helpful for searching survivors in the debris afterdisaster. Ultra-wideband (UWB) impulse radar endowed with high range resolution and highpenetrability offers an attractive technique to search survivors. When UWB impulse radar detectsa trapped person, the electromagnetic impulse is reflected by the chest surface of the inspectedperson. The round-trip travel time delay of the radar impulse changes periodically due to theperiodical motion of the chest surface caused by breathing motion. As a result, a slow-timeperiodical signal will be generated in the echo data. We call this periodical signal the slow-timerespiratory signal. If this respiratory signal is detected, the trapped person can be finally located.In real-life scenarios, strong interference signals always exist in the echo data. Due to thesmall chest surface motion, the respiratory signal normally is very weak. Therefore, thesignal-to-noise-and-clutter-ratio (SNCR) is always very low. The respiration detection usingUWB impulse radar under low SNCR is the research target of this paper.In this paper, the efforts in detecting respiration using UWB impulse radar involve twoaspects:1. An approach of slow-time band-pass filtering is applied to suppress the stronginterference signals existing in the echo data. Compared with those algorithmsproposed in previous works, this approach is proved to be more effective and has lowercomputational complexity.2. A new threshold-based decision algorithm for the respiratory signal is designed basedon the modified harmogram. This new algorithm detects the respiratory signal by itsharmonic structure and shows a better detection performance. By Monte Carlosimulation, the proposed algorithm shows about a2dB improvement of SNR. In orderto analyze the detection performance under more realistic conditions, a large number ofclutter samples are collected and used to assess the detection performance of theproposed algorithm against SNCR. The analysis result shows the proposed algorithmhas about a2dB improvement of SNCR.