Unsupervised Adversarial Domain Adaptation for Micro-Doppler Based Human Activity Classification

Du, Heng; Jin, Tian; Song, Yongping; Dai, Yongpeng

doi:10.1109/lgrs.2019.2917301

Cited by 36 publications

(10 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…By utilizing the motion capture database as the source dataset for knowledge transferring, Lang et al [14] proposed a UDA method to learn the domain-invariant features to classify the unlabeled measured radar data. Du et al [13] utilized an unsupervised adversarial domain adaption method to reduce the domain discrepancy between the simulated radar spectrogram dataset and the measured spectrogram dataset. Chen et al [22] proposed two adaptation networks that utilized DA to eliminate the impact of aspect angle on HAR with micro-Doppler (MD) spectrograms.…”

Section: B Unsupervised Deep Damentioning

confidence: 99%

“…Such an approach can achieve good performance when sufficient labeled data are available for each class. On the contrary, the unsupervised TL [13], [14], based on domain adaptation (DA) with unlabeled target data, is employed to learn domain-invariant feature representation. However, due to the lack of label information, the performance of the unsupervised methods is generally not as good as the supervised ones.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Semisupervised Human Activity Recognition With Radar Micro-Doppler Signatures

Fioranelli

et al. 2022

IEEE Trans. Geosci. Remote Sensing

View full text Add to dashboard Cite

Human activity recognition (HAR) plays a vital role in many applications, such as surveillance, in-home monitoring, and health care. Portable radar sensor has been increasingly used in HAR systems in combination with deep learning (DL). However, it is both difficult and time-consuming to obtain a large-scale radar dataset with reliable labels. Insufficient labeled data often limit the generalization of DL models. As a result, the performance of DL models will drop when being applied to a new scenario. In this sense, only labeling a small portion of data in the large-scale radar dataset is more feasible. In this article, we propose a semisupervised transfer learning (TL) algorithm, "joint domain and semantic transfer learning (JDS-TL)," for radar-based HAR, which is composed of two modules: unsupervised domain adaptation (DA) and supervised semantic transfer. By employing a sparsely labeled dataset to train the HAR model, the proposed method alleviates the need of labeling a significantly large number of radar signals. We adopt a public radar micro-Doppler spectrogram dataset including six human activities to evaluate JDS-TL. Experiments show that the proposed JDS-TL is able to recognize the six activities with an average accuracy of 87.6% when there are only 10% instances labeled in the training dataset. Ablation analysis also demonstrates the efficiency of the DA and the semantic transfer modules.

show abstract

Section: B Unsupervised Deep Damentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Semisupervised Human Activity Recognition With Radar Micro-Doppler Signatures

Fioranelli

et al. 2022

IEEE Trans. Geosci. Remote Sensing

View full text Add to dashboard Cite

show abstract

“…Due to this limitation in data availability, some authors resort to simulated data to train their deep-learning algorithms [7,8]. The core of the problem is then shifted to the choice of a suitable model to simulate human motion; this can be either analytically tailored to specific movements [18] or rely on MOtion CAPture (MOCAP) data [19].…”

Section: Related Workmentioning

confidence: 99%

“…Methods like domain adaptation, transfer learning and Generative Adversarial Networks (GANs) assist the improvement of classification or regression tasks from incomplete datasets [4]. In this paper, we focus on human activity classification with radar due to the research interest that it has currently awakened [5][6][7][8], for which we count on the help of millimeter-wave radar sensors operating at 60 GHz. Problems with datasets in this context can originate from several reasons: 1.…”

Section: Introduction 1motivationmentioning

confidence: 99%

Improving Radar Human Activity Classification Using Synthetic Data with Image Transformation

Hernangómez

Visentin

Servadei

et al. 2022

Sensors

View full text Add to dashboard Cite

Machine Learning (ML) methods have become state of the art in radar signal processing, particularly for classification tasks (e.g., of different human activities). Radar classification can be tedious to implement, though, due to the limited size and diversity of the source dataset, i.e., the data measured once for initial training of the Machine Learning algorithms. In this work, we introduce the algorithm Radar Activity Classification with Perceptual Image Transformation (RACPIT), which increases the accuracy of human activity classification while lowering the dependency on limited source data. In doing so, we focus on the augmentation of the dataset by synthetic data. We use a human radar reflection model based on the captured motion of the test subjects performing activities in the source dataset, which we recorded with a video camera. As the synthetic data generated by this model still deviates too much from the original radar data, we implement an image transformation network to bring real data close to their synthetic counterpart. We leverage these artificially generated data to train a Convolutional Neural Network for activity classification. We found that by using our approach, the classification accuracy could be increased by up to 20%, without the need of collecting more real data.

show abstract

“…Most existing relevant research involving human activity recognition is focused on classification and recognition of different activities such as walking, running, and jogging using machine learning (ML) techniques or deep learning based on the m-D spectrograms obtained by Short-Time Fourier Transform (STFT) [3][4][5][6][7][8]. In many applications, however, fine-grained analysis of human motion state is needed rather than recognition of different activities.…”

Section: Introductionmentioning

confidence: 99%

Human Multicomponent Micro-Doppler Signals Separation Based on a Novel Local Time-Frequency Sparse Reconstruction Method

Ni¹,

Huang²

2021

PIER C

View full text Add to dashboard Cite

The use of radar micro-Doppler (m-D) signatures for human activities classification, surveillance and healthcare has become a hot topic in recent years. While m-D signals are always multicomponent, it is necessary to separate them into mono-components signals associated with individual body parts for easier features analysis and extraction. In this paper, a novel method called local time-frequency sparse reconstruction (LTFSR) is proposed to iteratively extract and separate m-D components one by one in a descending intensity order from a time-frequency (T-F) representation. For the current strongest m-D component, we first estimate its instantaneous frequency (IF) by dividing the signal into short overlapping time intervals and selecting the best matching chirp atom to approximate the local frequency in each time interval based on matching pursuit. Then, a T-F filtering is used to extract and remove the strongest component from the multicomponent signal. Repeat the above steps until all m-D components are separated. Simulations are given to validate the effectiveness and robustness of the proposed method.

show abstract

Unsupervised Adversarial Domain Adaptation for Micro-Doppler Based Human Activity Classification

Cited by 36 publications

References 23 publications

Semisupervised Human Activity Recognition With Radar Micro-Doppler Signatures

Semisupervised Human Activity Recognition With Radar Micro-Doppler Signatures

Improving Radar Human Activity Classification Using Synthetic Data with Image Transformation

Human Multicomponent Micro-Doppler Signals Separation Based on a Novel Local Time-Frequency Sparse Reconstruction Method

Contact Info

Product

Resources

About