Professor Abdelhak M. Zoubir and Ann-Kathrin Seifert, M.Sc. discuss their work applying radar sensing technologies to monitor human gait in order to detect changes in walking patterns.
Advanced radar systems play an important role in many areas of our daily life, such as air traffic control, speed enforcement systems, and advanced driver assistance systems. Recently, radar has also become of increased interest for indoor applications. In particular, human activity recognition systems are rapidly evolving with application to, for example, fall motion detection for elderly care, or human-computer interaction for smart devices.1,2
The Signal Processing Group at Technische Universität Darmstadt, Germany, headed by Professor Abdelhak M. Zoubir, works on radar-based human gait analysis with applications in diagnosis, rehabilitation, and assisted living.1,3 Their work is in close co-operation with Professor Dr Moeness G. Amin, Director of the Center for Advanced Communications at Villanova University, USA. The research aims at using radar to monitor human gait in order to detect changes in walking patterns.
Clinical gait analysis
Gait analysis plays an important role in many areas such as medical diagnosis, physiotherapy and rehabilitation. In fact, the human gait is directly affected by many neurological conditions (e.g. Parkinson’s disease), orthopaedic problems and medical conditions (e.g. heart failures). Thus, by monitoring the human gait, numerous medical conditions can be detected at an early stage and ancillary diagnostic procedures can be performed.
Similarly, gait analysis systems can be used to objectively assess the state of rehabilitation of a patient recovering from a physical injury. Further, detecting changes in gait patterns can also help to assess the risk of falling. That is why unobtrusive gait monitoring systems will become particularly important in assisted living facilities and future smart homes.
Radar-based human monitoring
Numerous benefits of using radar for sensing human motions exist. They include, most importantly, preservation of the privacy of individuals, i.e., given the recorded radar signal, one cannot deduce a person’s gender, body shape, or clothing. In addition, the person does not need to wear any sensor on their body, but the radar can observe their gait remotely from a distance. Finally, since radar systems are small, they can pave the way for ambulatory gait analysis, which would allow for monitoring a person’s gait over a longer duration of time.
Joint-variable radar data representation for signal analysis
When a person is walking in front of the radar, the electromagnetic wave is reflected from the entire body. Due to the Doppler effect, the received radar signal does not only contain the transmitted frequency, but also frequency shifted components. These Doppler shifts are directly proportional to the radial velocity of the moving body. Frequency shifts that arise from so-called micro-motions, such as swinging arms or legs, are referred to as micro-Doppler frequencies.
In order to extract the Doppler frequencies from the measured time domain radar signal, spectral analysis is utilised, which represents the received energy for different frequencies. However, since the velocities of the different body parts change over time, the received radar signals are highly non-stationary, i.e., the received power spectrum varies with time. In this case, spectral analysis fails to capture the time dependency of the (micro-)Doppler frequencies. For this reason, time-frequency representations are used to represent the radar returns of human motions. For micro-Doppler analysis, the spectrogram is typically applied, which is obtained as the squared magnitude of the short-time Fourier transform (STFT).
Micro-Doppler signatures of human gait
As an example, Fig. 1 shows a spectrogram of a human gait. Since the person was walking toward the radar system, we observe only positive Doppler shifts and increasing energy levels over time. From the characteristic sinusoidal-shaped micro-Doppler signatures in the spectrogram, we can observe that the person took four steps in the five second measurement. These signatures are due to the swinging motion of the legs, whereas the torso’s Doppler signature lies between 0-100 Hz. The swinging feet reveal the highest Doppler shifts with up to 350 Hz, which relates to a velocity of 2.2 m/s.
Detection of gait asymmetries
Whenever a person is limping or walking with an assistive walking device such as a cane, the micro-Doppler signatures reveal the asymmetry in the gait. Typically, a limping leg swings with a lower velocity which leads to a lower maximal Doppler shift of the stride signature (see the left spectrogram of Fig. 2). This can be explained by the fact that the person tries to minimise the stance time of the affected foot, hence the swing time is extended accordingly. In the case an assistive walking device is employed, its micro-Doppler signatures overlay with those of the walking motion as shown in the right spectrogram of Fig. 2. Here, every other micro-Doppler stride signature shows higher energy levels due to the overlaying Doppler signature of the cane.
Detecting these minor changes in the micro-Doppler signatures is the key to render radar-based gait analysis possible. For this, advanced signal processing techniques are developed by the Signal Processing Group in order to reliably detect changes in walking patterns. For more details on research in the field of radar-based gait analysis, the reader is referred to the references and the web address below.
1 M. G. Amin, Ed., ‘Radar for Indoor Monitoring: Detection, Classification, and Assessment’. CRC Press, 2017
2 M. G. Amin, Y. D. Zhang, F. Ahmad, K. C. D. Ho, ‘Radar signal processing for elderly fall detection: the future for in-home monitoring,’ IEEE Signal Processing Magazine, vol. 33, no. 2, pp. 71–80, 2016
3 A.-K. Seifert, M. G. Amin, A. M. Zoubir, ‘Toward unobtrusive in-home gait analysis based on radar micro-Doppler signatures’, pre-print available at: arxiv.org/abs/1809.06653