Abstract
Keywords - Ultrasonic sensor, Pre-crash sensing, Pseudo-Wigner Distribution, Maximum likelihood method, Relative speed
Conventional ultrasonic sensors for automotive applications can detect a target in a short range, but the devices must be compact and waterproof. Those requirements limit their application to user-friendly systems such as parking assist systems and rear sonar systems. For pre-crash applications, generally proposed sensors like stereo cameras, radar or lidar are able to detect obstacles at long distances, but they are much more expensive than ultrasonic sensors. This paper proposes a range enhancement method for ultrasonic sensors to make them usable in pre-crash applications.
The time-of-flight method is one of the typical techniques used for measuring distance with ultrasonic sensors in general applications. The distance to the target can be calculated from the time difference between the transmitted pulse and the received pulse. The relative speed of the received pulse is found from the Doppler frequency calculated with a fast Fourier transform (FFT). However, a wave reflected from a far distance is quite difficult to detect, and the relative speed and distance are difficult to calculate because the signal is corrupted by surrounding random disturbance noise, e.g., echoes from the ground.
For pre-crash applications, the sensor needs to detect targets at farther distances because the targets move faster than in the case of conventional user-friendly systems. Hence, we developed a method of detecting and calculating the reflected signal masked by random noise in order to increase the detection range of conventional ultrasonic sensors. This method uses a pseudo-Wigner distribution (P-WD)-based likelihood function.
The proposed method was applied to observation data obtained experimentally with actual ultrasonic sensors. First, the P-WD of the reflected wave of a target at a near distance which was free from any noise was calculated as a reference. Then, using the whole length of the noisy observation data, the P-WD was calculated off-line to obtain the log-likelihood ratio function. The reflected wave was detected by maximizing the P-WD-based likelihood function with respect to the parameter related to the time delay and the frequency. Finally, the distance was calculated from the time delay and the relative speed from the frequency shift.
The results of this preliminary experiment confirmed that the maximum distance for detecting the reflected signal can be extended by as much as about 1.4 times. The distance and relative speed of a moving target were also estimated simultaneously with high accuracy.