Reliable and fast detection of maneuvering target in complex background is important for both civilian and military applications. It is rather difficult due to the complex motion resulting in energy spread in time and frequency domain. Also, high detection performance and computational efficiency are difficult to balance in case of more pulses. In this paper, we propose a fast and refined processing method of radar maneuvering target based on hierarchical detection, utilizing the advantages of moving target detection (MTD), and the proposed sparse fractional representation. The method adopts two-stage threshold processing. The first stage is the coarse detection processing screening out the rangebins with possible moving targets. The second stage is called the refined processing, which uses robust sparse fractional Fourier transform (RSFRFT) or robust sparse fractional ambiguity function (RSFRAF) dealing with high-order motions, i.e., accelerated or jerk motion. And the second stage is carried out only within the rangebins after the first stage. Therefore, the amount of calculation can be greatly reduced while ensuring high detection performance. Finally, real radar experiment of UAV target detection is carried out for verification of the proposed method, which shows better performance than the traditional MTD method, and the FRFT-FRAF hierarchical coherent integration detection with less computational burden.

Motion monitoring systems are often designed and researched to detect the movement of human lower limbs, and play an important role in the field of exoskeleton control. However, current wearable devices can still be improved to be more convenient or accurate in motion recognition. In this work, a comfortable smart wearable gait monitoring system was designed and tested. Inertial measurement units (IMUs) and flexible membrane compression sensors were implemented, integrated to a comfortable sport pant and insoles of both feet, respectively. Data acquisition module was designed, while software with user interface for data collection and storage was realized based on LABVIEW. Experiments were conducted to evaluate the recognition performance of the smart wearable gait monitoring system among nine common actions. Results show that the combined data set of IMUs and compression sensor provided by the system can highly improve classification performance. Based on the self-designed sensing network and the K-nearest neighbor machine learning algorithm, the recognition rate of nine motion patterns can reach as high as 99.96%, showing that the multi-channel wearable gait monitoring system is more effective for motion detection and prediction compared to that with single-type sensors.

No abstract

No abstract

No abstract