Gait Parameters Extraction Based on Torso Signal Using FMCW Radar

The previously proposed methods of using frequency-modulated continuous-wave (FMCW) radar to extract biomechanical gait parameters in Doppler spectrogram could result in loss of certain important gait parameters. We propose a combination of Velocity-Time (VT) spectrogram and Range-Time (RT) spectrogram to extract gait parameters based on the torso signal. The torso signal is separated by the maximum extraction algorithm to get the most stable signal with sufficient information including direction. Then the peak extraction algorithm is used to extract the peak velocity points to get time points when the foot touches the ground. And the number of steps, step time, stride time, number of turns, step length, stride length and torso velocity could be extracted. Besides, gait temporal symmetry and range symmetry also reflect the participants’ stable state during walking. Finally, we compare the number of turns and steps with the results recorded during the experiment. The accuracy of number of turns is achieved 95.0%, and the average error ratio in the number of steps is 2.8%.