60 x-Position (m) 50 40 30 20 10 -1 0 1 y-Position (m) 2 -1 0 1 y-Position (m) (a) 2 -1 (b) 0 1 y-Position (m) 2 (c) Figure 4. Real data results of a series-production automotive radar sensor for experiment 1 as described in Figure 3. Position estimates in Cartesian coordinates were obtained with (a) conventional radar processing, (b) high-resolution frequency estimation in the spatial domain, and (c) high-resolution frequency estimation in the Fourier domain with optimal selection of resolution dimension. 60 x-Position (m) 50 40 30 20 10 -1 0 1 y-Position (m) 2 -1 0 1 y-Position (m) (a) 2 (b) -1 0 1 y-Position (m) 2 (c) Figure 5. Real data results of a series-production automotive radar sensor for experiment 2 as described in Figure 3. Position estimates in Cartesian coordinates were obtained with (a) conventional radar processing, (b) high-resolution frequency estimation in the spatial domain, and (c) high-resolution frequency estimation in the Fourier domain with optimal selection of resolution dimension. has a significantly larger frequency separation than all others. The carrier is then moved straight toward the radar sensor at walking speed. 44 Figure 3(b) shows the corresponding frequency separations normalized to the respective resolution limits. The relative velocity separation is very low and thus omitted. Observe that, IEEE SIgnal ProcESSIng MagazInE | March 2017 |