Radar Human Activity Recognition with an Attention-Based Deep Learning Network

Abstract

Radar-based human activity recognition (HAR) provides a non-contact method for many scenarios, such as human–computer interaction, smart security, and advanced surveillance with privacy protection. Feeding radar-preprocessed micro-Doppler signals into a deep learning (DL) network is a promising approach for HAR. Conventional DL algorithms can achieve high performance in terms of accuracy, but the complex network structure causes difficulty for their real-time embedded application. In this study, an efficient network with an attention mechanism is proposed. This network decouples the Doppler and temporal features of radar preprocessed signals according to the feature representation of human activity in the time–frequency domain. The Doppler feature representation is obtained in sequence using the one-dimensional convolutional neural network (1D CNN) following the sliding window. Then, HAR is realized by inputting the Doppler features into the attention-mechanism-based long short-term memory (LSTM) as a time sequence. Moreover, the activity features are effectively enhanced using the averaged cancellation method, which improves the clutter suppression effect under the micro-motion conditions. Compared with the traditional moving target indicator (MTI), the recognition accuracy is improved by about 3.7%. Experiments based on two human activity datasets confirm the superiority of our method compared to traditional methods in terms of expressiveness and computational efficiency. Specifically, our method achieves an accuracy close to 96.9% on both datasets and has a more lightweight network structure compared to algorithms with similar recognition accuracy. The method proposed in this article has great potential for real-time embedded applications of HAR.

A Non-destructive Radar Device for Detecting Additive Materials in Concrete

“…The amplitude levels in Figure 5 also confirm that the additive materials in concrete can be clearly identified. In previous studies on non-destructive radar inspection devices [1,7,8,[11][12][13], the measured data were visualized and could be in several aspects, such as additive materials, defects, and scanned mapping. However, these devices had a complicated structure with a special radar sensor and an elaborated signal processing system.…”

“…The HB 100 Doppler radar sensor is widely used for motion detection thanks to its low cost and high reliability [11][12]. Recently, the HB 100 sensor was used for additional functions such as underground metal detection [13] or wooden pole inspection in forests [14].…”

Pedestrian Pose Recognition Based on Frequency-Modulated Continuous-Wave Radar with Meta-Learning

“…Often, LSTM can be combined with other networks. In [16], a combination of CNNs and LSTM neural networks was used. After obtaining features through a one-dimensional CNN, the features were integrated into an LSTM neural network with attention mechanisms as a time series to achieve human activity recognition.…”

A lightweight hybrid vision transformer network for radar-based human activity recognition

“…2D FFT processing compresses the signal energy at the corresponding position on the range-angle plane. A phase average cancellation method 29 is then utilized for the static clutter suppression, which will preserve the micro-Doppler signal components. Two-dimension constant false alarm rate (2D-CFAR) is applied to detect the target against the noise background.…”

“…Considering the embedded application background of radar-based HAR, some work has attempted to solve the efficiency and performance issues 28 , 29 , but new networks need to be developed to improve the recognition performance on the lightweight structures more effectively. To achieve high-accuracy HAR, this paper developed a lightweight hybrid Vision Transformer (LH-ViT) network.…”