Haoran Ji

Mr. Haoran Ji

Tianjin International Engineering Institute

Research Focus & Metrics

Academic Career Length

6Years (2021-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
27 May 2021 in IEEE Sensors Journal
Verified
Journal Article
Closed access
YLYue Lang
Yue Lang
CHChunping Hou
Chunping Hou
HJHaoran Ji
Haoran Ji
YYYang Yang
Yang Yang
Radar sensors and micro-Doppler signatures have been widely used to recognize human motions. Apart from the motion classification task, human motion detection has attracted much attention as an emerging topic. A majority of existing motion detectors are designed for a specific motion, such as falling. In some scenarios, however, a broader range of human actions is of interest, hence a general motion detector is desired. In this paper, we propose a radar-based motion detection model named dual generative adversarial network (DGN). The proposed model tackles the detection task as a one-class classification problem and is applicable to detecting various motions. Unlike prior fall detection algorithms, which depend on manually collected alien data, the DGN employs a dual generation scheme to automatically produce valid alien samples in both the pixel level and the semantic level. The model is verified on two measured radar datasets containing individual motions and interactive motions, respectively. The experimental results show that our method outperforms other existing models on the human motion detection task.

26 October 2021 in IEEE Signal Processing Letters
Verified
Journal Article
Closed access
HJHaoran Ji
Haoran Ji
CHChunping Hou
Chunping Hou
YYYang Yang
Yang Yang
FFFrancesco Fioranelli
Francesco Fioranelli
YLYue Lang
Yue Lang
In this letter, a radar-based gait authentication method is proposed. We focus on the overfitting problem on the target category caused by limited training data in authentication models and propose a one-class classification model to alleviate this problem. The effectiveness of such model is verified by establishing a radar-based gait dataset, which is composed of gait micro-Doppler spectrograms derived from nine human subjects. The experimental results demonstrate that, under the condition of limited training data, the performances of an authentication model degrade because misclassification of the non-target samples easily occurs. The proposed method effectively avoids this risk, performing the other existing authentication and one-class classification methods on the metric Equal Error Rate.

19 September 2022 in IEEE Signal Processing Letters
Verified
Journal Article
Closed access
YYYang Yang
Yang Yang
YZYutong Zhang
Yutong Zhang
HJHaoran Ji
Haoran Ji
BLBeichen Li
Beichen Li
CSChunying Song
Chunying Song
In recent years, radar-based human activity recognition has received considerable interest, but it faces the difficulty of a shortage of training data. In this research, a two-stage domain adaption method is proposed for this issue. This method combines simulated radar spectrograms with an elaborated domain-translation network to perform domain adaptation, hence enabling human activity recognition with limited measurement data support. To validate the efficacy of our method, we conduct radar simulation and measurements, and the experimental results demonstrate that the proposed method outperforms other comparisons in terms of accuracy, showing its great capacity for radar-based human activity recognition.

14 August 2023 in IEEE Transactions on Instrumentation and Measurement
Verified
Journal Article
Closed access
YLYue Lang
Yue Lang
CWChenyang Wu
Chenyang Wu
YYYang Yang
Yang Yang
HJHaoran Ji
Haoran Ji
Advancements in sensing techniques have fueled the construction of a worldwide smart environment. An accompanying concern is the security issue. This paper presents a noninvasive user authentication technique using an ultra-wideband radar sensor. Human gait micro-Doppler signatures captured by the radar are used as the biometrics of individuals. Unlike the existing authentication techniques, our proposed method does not require a gallery set for retrieval during the testing stage. Instead, we formalize the authentication task as a one-class classification problem and utilize a generative adversarial network to characterize the legal users’ movement modes, especially the fine-grained distinctions of micro-Doppler signatures. Meanwhile, the discriminator automatically outputs the prediction result, indicating whether a user is legal or not. The fully convolutional network architecture and a fine-grained recognition module are added to enhance the discrimination ability of the model. Experiments are carried out using measurement data from 15 subjects, and the results demonstrate that the proposed method achieves an equal error rate of 0.234, outperforming the comparative algorithms by at least 9.8%. Moreover, the model is evaluated for its robustness against various attacks as well as different walking styles. An ablation study is conducted to verify the effectiveness of the network design.

1 March 2024 in IEEE Transactions on Instrumentation and Measurement
Verified
Journal Article
Closed access
YYYang Yang
Yang Yang
XSXiaochun Shang
Xiaochun Shang
BLBeichen Li
Beichen Li
HJHaoran Ji
Haoran Ji
YLYue Lang
Yue Lang
The utilization of multi-sensors data mining and processing exhibits significant promise in forthcoming applications, with cross-modal retrieval serving as a crucial element within this realm of research. Radar offers continuous surveillance capabilities over a 24-hour period, whereas cameras provide visual information about the surrounding environment. This characteristic has led to the widespread utilization of the sensors across various applications. This study centers on the task of video-radar retrieval, which, to the best of our understanding, represents the inaugural investigation of this particular subject matter. The proposed model differs from previous cross-modal retrieval tasks that utilize video data. It operates in a detection-free scenario, eliminating the need to actively identify the specific region where the human subject is located. In order to address the challenges posed by inconsistent semantic information and a large modality gap in radar spectrograms and video frames, we propose a two-step approach. First, we introduce an identity-excluded feature removal scheme to eliminate redundant information in videos. Then, we present a two-stage feature alignment strategy to align the multi-modal features, enabling retrieval without being hindered by the inherent discrepancies between the irrelevant modalities. In order to assess the efficacy of our model, we have implemented a video-radar system and constructed a cross-modal dataset. Subsequently, we have conducted a comparative analysis between our model and seven state-of-the-art cross-modal algorithms. The exemplary performance exhibited by our model demonstrates the effectiveness of our approach, while the conducted ablation studies provide empirical evidence supporting the efficiency of our design choices.

29 September 2023 in IEEE Transactions on Instrumentation and Measurement
Verified
Journal Article
Closed access
YXYan Xu
Yan Xu
WLWenyu Li
Wenyu Li
YYYang Yang
Yang Yang
HJHaoran Ji
Haoran Ji
YLYue Lang
Yue Lang
Radar echo signal separation of multiple persons is a key issue in signal processing, whose objective is to separate the superimposed echo signals into their respective signals. However, the task is challenged by the heavy superimposition regions of echo signals. When the echoes of distinct individuals exhibit substantial overlap, the independent signals of each individual may become significantly distorted, providing a possible threat to subsequent tasks such as human activity classification. In this paper, we introduce a new technique for signal separation known as Non-linear superimposed Mask-guided and multi-person Domain Contrastive Regularization (NM-DCR). Our aim is to address the challenge of separating radar Range-Doppler (RD) maps that exhibit significant superimposition of backscattering echoes from multiple individuals. In light of the non-linear relationship between the superimposition coefficients on the different range and Doppler components, we propose a superimposed mask-guided scheme with the proposed superimposition coefficients learning process to explore the intrinsic relationship, which is embedded into the separation model. Additionally, we develop a contrastive regularization term to further avoid signal distortion by constructing multi-person domain negative samples as the lower boundary, which shrinks the solution space for decoupling the superimposed signal. To validate our model, we create a multi-person radar dataset to conduct radar simulations. Experimental results show that our model outperforms other comparable models, demonstrating its significant potential for improving the performance of radar echo signal separation under the challenge of heavy superimposition regions.