In this paper, we propose a method of classifying the overhead target in automotive radar systems. The conventional automotive radar systems struggle to distinguish overhead road structures (e.g., bridges, overpasses, traffic signs) and ground-level targets (e.g., vehicles and pedestrians) due to the lack of elevation resolution. Consequently, the forward collision-avoidance assist (FCA) system or the autonomous emergency braking (AEB) system, may perceives detected overhead road structures as being on the road. Therefore, unnecessary operation of the FCA or AEB may occur. Our proposed method exploits the different patterns of changes in the estimated range, relative velocity, and signal amplitude as the radar approaches the overhead structure and ground-level targets. These patterns can be used to classify overhead structures by designing the appropriate classifier. Therefore, additional antenna elements arranged in the vertical direction required for elevation angle estimation are unnecessary. First, we obtain the radar signals by measuring the metal reflector installed at different heights. Then, by processing the received signal, we generate the input data for the proposed bidirectional long short-term memory-based height classifier. Our proposed method can classify the installation height of the reflector with an average accuracy of 98.18% within the range of 13 m, and an average accuracy of 94.97% within the range of 20 m in a complicated environment.

In radar-based air-writing, the continuous measurements of the hand movements may result in the addition of unnecessary strokes for certain characters or digits (e.g., 4 and 5), making it difficult to accurately recognize the air-written results when observed by human eyes. Therefore, we propose a deep neural network (DNN)-based classifier designed to identify unnecessary strokes and clutter that arise during the radar-based air-writing. First, while air-writing the digits from 0 to 9, the range, angle, and signal amplitude of the hand movement are obtained through a radar system. Then, we represent the hand’s trajectory in the form of x and y coordinates by using the information of range and angle. Next, we train the DNN-based classifier using the acquired x and y coordinates, signal amplitude, and frame index as input features. To ensure the classifier’s performance would not be impacted by changes in the position and size of the air-writing area, we apply the normalization to x and y coordinates. Finally, the performance of the classifier is verified using the results of air-writing digits from 0 to 9. The proposed method identifies unnecessary strokes and clutter regardless of the position and size of the air-writing area, demonstrating an average classification accuracy of 94.57%. Furthermore, when the classifier was validated with different individuals conducting the air-writing, the classifier exhibited an average classification accuracy of 93.9%.

In this paper, we propose a method for detecting the misalignment of automotive radar sensors. Ensuring the accurate operation of automotive radar sensors is essential for the safety of drivers and passengers. However, when radar sensors are misaligned, they may perceive the surrounding road environment inaccurately, providing incorrect information to the vehicle’s control systems. This can lead to erroneous decisions and potentially cause traffic accidents. Therefore, a method of examining the alignment state of the automotive radar sensors is required. Furthermore, due to the time-consuming and expensive nature of removing the bumper to access the radar sensor for direct inspection, an alternative method of detecting misalignment indirectly is necessary. Our method enables the detection of misalignment in automotive radar sensors through the utilization of ground reflections of radar signals. Depending on the mounting angle of the radar sensor, the range-velocity (RV) spectrum of the received signal reflected from the ground varies significantly. This frequency spectrum can be used to classify the mounting angle of the radar sensor to detect the misalignment. We design a convolutional neural network (CNN)-based classifier with the RV spectrum as input. Using the CNN-based classifier, we estimated the mounting angle of the radar sensor with an average accuracy of 94.72%, demonstrating that our proposed method can detect the misalignment of the radar sensor effectively.

Due to availability of maintaining the orthogonality between transmit signals and effectively leveraging time resources, Doppler division multiplexing (DDM) method has been widely investigated for operating multiple-input multiple-output (MIMO) radar systems. However, the number of peaks equal to the number of transmitting antenna elements appear along the velocity axis for every target, and it leads to Doppler ambiguity problem. As a solution, we propose a novel approach to resolve the Doppler ambiguity using the phase information of folded array, which overlaps certain antenna array elements when forming the entire MIMO antenna array. The proposed method consists of two main steps. First, we apply the uniform phase DDM to simultaneously transmit signals for each transmitting antenna elements. This method results in multiple peaks for the same target occurring at constant intervals along the velocity axis. By using the similarity of phase values between overlapping MIMO array elements, our proposed method can accurately determine the precise order of the transmitting antenna elements. Since the proposed method resolves Doppler ambiguity and enables to use non-coherent integration (NCI) on both the receiving and transmitting antenna channels, it significantly improves the signal-to-noise ratio (SNR) and increases the detection probability. Simulation and experimental results based on automotive radar sensor demonstrate the effectiveness of the proposed method and we expect that this method enables reliable perception tasks for autonomous driving.

Phase-modulated continuous-wave (PMCW) radar systems have been gaining attention as next-generation radar systems due to their advantages, such as the availability of detecting high-velocity targets, multiple-input–multiple-output operation capability, and simplified transmitter architecture. Because the autocorrelation property of phase modulation codes plays a crucial role in determining the system performance of PMCW radar, proper signal design is essential. Golay complementary sequences are well-suited for PMCW radar systems because of their outstanding autocorrelation characteristics. These properties yield a sharp mainlobe and suppressed sidelobes in the range profile, which, in turn, enhance the accuracy of range estimation. To make more efficient use of these Golay sequences, we propose a signal-processing method based on frequency-domain (FD) accumulation with Doppler shift compensation. In addition, we propose an efficient frame design method based on the frame structure used in the IEEE 802.11ad preamble. The combination of the proposed frame design and signal-processing method enables significant improvements in Doppler robustness and time efficiency. Simulation results confirmed that the proposed FD accumulation with the Doppler shift compensation method improves the peak power loss ratio (PPLR) by approximately 4 dB compared to the conventional accumulation method. In addition, the proposed frame design outperforms existing Golay-based designs, with improvements of up to 13 dB in the PPLR, 19 dB in the peak sidelobe level ratio (PSLR), and 11 dB in the integrated sidelobe level ratio (ISLR). Moreover, it achieved superior performance when compared to a widely used pseudorandom binary sequence (PRBS)-based design, with an enhancement of 0.68 dB in PSLR and 10.57 dB in the ISLR.

In this paper, we propose a method for recognizing kick motions using a multiple-input multiple-output (MIMO) frequency-modulated continuous wave (FMCW) radar system combined with deep learning techniques. Smart trunk opener (STO) systems that provides users with hands-free trunk operation have been gaining attention. To address the limitations of prevalent STO systems, which rely on capacitive or ultrasonic sensors, we propose employing a 60 GHz MIMO FMCW radar system. Our design of a 60 GHz MIMO FMCW radar system can detect kick motions and estimate their range, velocity, and angle, using a signal processing chain that includes a two-dimensional fast Fourier transform and multiple antenna elements. Additionally, using the acquired information on the range, velocity, and angle of the kick motion, we propose a deep learning-based model for recognizing specific kick motions to operate the STO system. This model is designed to take sequences of range, velocity, and angle as its input, unlike the conventional motion recognition methods that treats radar data as an image. We analyze the performances of various deep learning-based models using a dataset of various kick motions obtained through real-life measurements. Results showed that the one-dimensional convolutional neural network achieved more than 97% in accuracy, F1 score, recall and precision. Furthermore, when compared with the models soly based on velocity information, overall performances decreased to around 66%, proving the effectiveness of the MIMO FMCW radar-based kick motion recognition.