基于机器学习的车用主动防撞预警雷达信号识别系统研究

Research on Vehicle Learning Radar Signal Recognition System Based on Machine Learning

摘要:

为了避免同向和相向干扰信号对识别精准度影响，引入机器学习，研究车用主动防撞预警雷达信号识别系统。在机器学习支持下，设计预警雷达信号识别系统总体架构，采用BGT24MTR12E6327XUMA1型号原装雷达收发器，通过TendaA9信号放大器将混频信号送到信号处理系统之中，以此控制汽车行驶速度。TMS320F206 DSP通过CAN总线连接外部设备和TJA1041A总线收发器，使PC和DSP之间能够串行通信。使用抗干扰流水线结构转换方式，基于机器学习获取无干扰实时状态信号。通过计算雷达信号相似度，设计具体识别流程。依据各个子雷达在汽车上分布情况设计实验，由实验结果可知，相向干扰下机器学习技术信号识别精准度最高可达到96%；同向干扰下机器学习技术信号识别精准度最高可达到94%，为车辆安全行驶提供设备支持。

Abstract:

In order to avoid the influence of the same direction and the opposite direction interference signals on the recognition accuracy, machine learning is introduced to study the signal recognition system of the active anti-collision early warning radar for vehicles. Under the support of machine learning, the overall architecture of the early warning radar signal recognition system is designed. The original bgt24mtr12e6327xuma1 radar transceiver is used, and the mixing signal is sent to the signal processing system through the tendaa9 signal amplifier to control the vehicle speed. TMS320F206 DSP connects external equipment and tja1041a bus transceiver through CAN bus, which enables serial communication between PC and DSP. Based on machine learning, the interference free real-time state signal is obtained by using the anti-interference pipeline structure conversion method. By calculating the radar signal similarity, the specific recognition process is designed. The experiment is designed according to the distribution of each sub radar on the vehicle. According to the experimental results, the recognition accuracy of machine learning technology signal can reach 96% at most under the condition of the opposite interference; the recognition accuracy of machine learning technology signal can reach 94% at most under the condition of the same interference, providing equipment support for the safe driving of the vehicle.