Rui Zhang

Bio: Rui Zhang is an academic researcher from Tsinghua University. The author has contributed to research in topics: Gesture recognition & Inverse synthetic aperture radar. The author has an hindex of 5, co-authored 7 publications receiving 128 citations.

Sparsity-Driven Micro-Doppler Feature Extraction for Dynamic Hand Gesture Recognition

Abstract: In this paper, a sparsity-driven method of micro-Doppler analysis is proposed for dynamic hand gesture recognition with radar sensors. First, sparse representations of the echoes reflected from dynamic hand gestures are achieved through the Gaussian-windowed Fourier dictionary. Second, the micro-Doppler features of dynamic hand gestures are extracted using the orthogonal matching pursuit algorithm. Finally, the nearest neighbor classifier is combined with the modified Hausdorff distance to recognize dynamic hand gestures based on the sparse micro-Doppler features. Experiments with real radar data show that the recognition accuracy produced by the proposed method exceeds 96% under moderate noise, and the proposed method outperforms the approaches based on principal component analysis and deep convolutional neural network with small training dataset.

Sparsity-based dynamic hand gesture recognition using micro-Doppler signatures

Abstract: In this paper, a sparsity-driven method of micro-Doppler analysis is proposed for dynamic hand gesture recognition with radar sensor. The sparse representation of the radar signal in the time-frequency domain is achieved through the Gabor dictionary, and then the micro-Doppler features are extracted by using the orthogonal matching pursuit (OMP) algorithm and fed into classifiers for dynamic hand gesture recognition. The proposed method is validated with real data measured with a K-band radar. Experiment results show that the proposed method outperforms the principal component analysis (PCA) algorithm, with the recognition accuracy higher than 90%.

Micro-doppler interference removal via histogram analysis in time-frequency domain

Abstract: Micro-Doppler interference generated by mechanical vibration or rotation of a target or its parts may degrade the quality of radar images. In this paper, a method based on histogram analysis of the time-frequency distribution is proposed for enhanced radar image reconstruction with effective micro-Doppler interference removal. First, based on the different frequencies of occurrence between the micro-Doppler and rigid-body components as revealed in the time-frequency analysis, the rigid-body data contaminated by micro-Doppler components are removed at each frequency bin. Then, the full data set is restored based on the data preserved in the first step. Finally, the radar image is reconstructed by using the preserved and restored data rendered from the previous two steps. Experimental results show that compared with the recently developed L-statistics-based method, the proposed method offers better micro-Doppler interference removal capability and, thereby, enhances the reconstructed radar images with higher peak-sidelobe ratio and integrated sidelobe ratio.

Multi-aspect micro-Doppler signatures for attitude-independent L/N quotient estimation and its application to helicopter classification

Abstract: Micro-Doppler signals returned from the main rotor of a helicopter can be used for feature extraction and helicopter classification. An intrinsic feature of a helicopter that may be extracted from the micro-Doppler signatures is the L/N quotient, where N denotes the number of rotor blades and L is the blade length. However, in monostatic radar, the L/N quotient cannot be accurately estimated due to the unknown attitude angles of non-cooperative helicopters. To solve this problem, an attitude-independent L/N quotient estimation method based on multi-aspect micro-Doppler signatures is proposed in this study. The helicopter is observed from different aspect angles, and the multi-aspect micro-Doppler signatures are jointly processed to solve the attitude angles of the helicopter and estimate the L/N quotient unambiguously. Experiments with both simulated and real data demonstrate that, the proposed method is robust with respect to the attitude of the helicopter and, therefore, significantly improves the accuracy of L/N quotient estimation compared with only using the signature observed from single-aspect angle. This implies that the proposed method has the potential to increase the success rate of helicopter classification.

Helicopter classification via period estimation and time-frequency masks

Abstract: The rotation of blades of a helicopter induces a Doppler modulation around the main Doppler shift, which is commonly called the micro-Doppler signature and can be used for target classification. In this paper, an automatic helicopter classification method is proposed by estimating the period of the micro-Doppler signature and identifying the number of blades via time-frequency masks. The advantages of this method are threefold: (1) it determines the number of blades automatically; (2) it significantly reduces the computational burden compared to the classical model dictionary-based classification methods; (3) it is robust with respect to the inclination of the helicopter. The effectiveness of the proposed approach is validated by using both synthetic and real data.

'The Micro-Doppler Effect in Radar' by V.C. Chen

Abstract: This is a book review of 'The Micro-Doppler Effect in Radar' by V. C. Chen. Artech House, 16 Sussex Street, London, SW1V 4RW, UK. 2011. 290pp + diskette. Illustrated. £90. ISBN 978-1-60807-057-2.

Radar-Based Human-Motion Recognition With Deep Learning: Promising applications for indoor monitoring

Abstract: Deep learning (DL) has shown tremendous promise in radar applications that involve target classification and imaging. In the field of indoor monitoring, researchers have shown an interest in DL for classifying daily human activities, detecting falls, and monitoring gait abnormalities. Driving this interest are emerging applications related to smart and secure homes, assisted living, and medical diagnosis. The success of DL in providing an accurate real-time accounting of observed humanmotion articulations fundamentally depends on the neural network structure, input data representation, and proper training. This article puts DL in the context of data-driven approaches for motion classification and compares its performance with other approaches employing handcrafted features. We discuss recent proposed enhancements of DL classification performance and report on important challenges and possible future research to realize its full potential.

Personnel Recognition and Gait Classification Based on Multistatic Micro-Doppler Signatures Using Deep Convolutional Neural Networks

Abstract: In this letter, we propose two methods for personnel recognition and gait classification using deep convolutional neural networks (DCNNs) based on multistatic radar micro-Doppler signatures. Previous DCNN-based schemes have mainly focused on monostatic scenarios, whereas directional diversity offered by multistatic radar is exploited in this letter to improve classification accuracy. We first propose the voted monostatic DCNN (VMo-DCNN) method, which trains DCNNs on each receiver node separately and fuses the results by binary voting. By merging the fusion step into the network architecture, we further propose the multistatic DCNN (Mul-DCNN) method, which performs slightly better than VMo-DCNN. These methods are validated on real data measured with a 2.4-GHz multistatic radar system. Experimental results show that the Mul-DCNN achieves over 99% accuracy in armed/unarmed gait classification using only 20% training data and similar performance in two-class personnel recognition using 50% training data, which are higher than the accuracy obtained by performing DCNN on a single radar node.

Robust Gesture Recognition Using Millimetric-Wave Radar System

Abstract: Gesture recognition is one of the most intuitive forms of humancomputer interface. Gesture sensing can replace interfaces such as touch and clicks needed for interacting with a device. In this article, we present a short-range compact 60-GHz mm-wave radar sensor that is sensitive to fine dynamic hand motions. A series of rangeDoppler images are extracted and processed using a long recurrent all-convolution neural network for real-time dynamic hand gesture recognition. Furthermore, we make use of novel data augmentation techniques for the proposed gesture recognition system to generalize for multiple users and operating environments. The results show accurate classification performance requiring very low processor footprint facilitating implementation in embedded platforms with real-time user feedback.

Sparsity-Driven Micro-Doppler Feature Extraction for Dynamic Hand Gesture Recognition

Abstract: In this paper, a sparsity-driven method of micro-Doppler analysis is proposed for dynamic hand gesture recognition with radar sensors. First, sparse representations of the echoes reflected from dynamic hand gestures are achieved through the Gaussian-windowed Fourier dictionary. Second, the micro-Doppler features of dynamic hand gestures are extracted using the orthogonal matching pursuit algorithm. Finally, the nearest neighbor classifier is combined with the modified Hausdorff distance to recognize dynamic hand gestures based on the sparse micro-Doppler features. Experiments with real radar data show that the recognition accuracy produced by the proposed method exceeds 96% under moderate noise, and the proposed method outperforms the approaches based on principal component analysis and deep convolutional neural network with small training dataset.