Deep learning methods have been widely explored in motor imagery (MI)-based brain computer interface (BCI) systems to decode electroencephalography (EEG) signals. However, most studies fail to fully explore temporal dependencies among MI-related patterns generated in different stages during MI tasks, resulting in limited MI-EEG decoding performance. Apart from feature extraction, learning temporal dependencies is equally important to develop a subject-specific MI-based BCI because every subject has their own way of performing MI tasks. In this paper, a novel temporal dependency learning convolutional neural network (CNN) with attention mechanism is proposed to address MI-EEG decoding. The network first learns spatial and spectral information from multi-view EEG data via the spatial convolution block. Then, a series of non-overlapped time windows is employed to segment the output data, and the discriminative feature is further extracted from each time window to capture MI-related patterns generated in different stages. Furthermore, to explore temporal dependencies among discriminative features in different time windows, we design a temporal attention module that assigns different weights to features in various time windows and fuses them into more discriminative features. The experimental results on the BCI Competition IV-2a (BCIC-IV-2a) and OpenBMI datasets show that our proposed network outperforms the state-of-the-art algorithms and achieves the average accuracy of 79.48%, improved by 2.30% on the BCIC-IV-2a dataset. We demonstrate that learning temporal dependencies effectively improves MI-EEG decoding performance. The code is available at https://github.com/Ma-Xinzhi/LightConvNet.

The monitoring of physiological parameters is a crucial topic in promoting human health and an indispensable approach for assessing physiological status and diagnosing diseases. Particularly, it holds significant value for patients who require long-term monitoring or with underlying cardiovascular disease. To this end, Visual Contactless Physiological Monitoring (VCPM) is capable of using videos recorded by a consumer camera to monitor blood volume pulse (BVP) signal, heart rate (HR), respiratory rate (RR), oxygen saturation (SpO₂) and blood pressure (BP). Recently, deep learning-based pipelines have attracted numerous scholars and achieved unprecedented development. Although VCPM is still an emerging digital medical technology and presents many challenges and opportunities, it has the potential to revolutionize clinical medicine, digital health, telemedicine as well as other areas. The VCPM technology presents a viable solution that can be integrated into these systems for measuring vital parameters during video consultation, owing to its merits of contactless measurement, cost-effectiveness, user-friendly passive monitoring and the sole requirement of an off-the-shelf camera. In fact, the studies of VCPM technologies have been rocketing recently, particularly AI-based approaches, but few are employed in clinical settings. Here we provide a comprehensive overview of the applications, challenges, and prospects of VCPM from the perspective of clinical settings and AI technologies for the first time. The thorough exploration and analysis of clinical scenarios will provide profound guidance for the research and development of VCPM technologies in clinical settings.

Because unmanned forklifts need to recognise and locate pallets in warehouses, a detection algorithm based on deep learning framework was proposed. First, the authors collected a large number of pictures including people and pallet in the real warehouse and marked the corresponding label to build a logistics warehouse pallet database. Second, the object detection algorithm based on a single shot multibox detector is improved and trained by the database. In the prediction phase, the network combines the multiscale feature maps with different resolution, which enhances the adaptability of the network to the detection task. Third, the algorithm is an end‐to‐end detection network, i.e. uses a single network for detection tasks, which can be easily combined with other systems which need detection tasks. The experimental results show that the accuracy of the improved pallet detection algorithm can reach 92.7% and the test rate is 42 frames per second, which can meet the requirements of the efficiency and accuracy of the pallet detection while using the TITAN X GPU.

Oxygen saturation (SpO ₂ ) is a vital parameter in clinical practice. Clinical research suggests that SpO ₂ may significantly decrease in patients with lung infections before evident symptoms, such as COVID-19, appear. Therefore, monitoring the cardiopulmonary status of individuals with lung disease is indispensable. Although contactless SpO ₂ measurements perform well in laboratory settings, their clinical application still faces several challenges. To bridge this gap, a camera-based contactless SpO ₂ measurement foundation model called CCSpO ₂ Net is proposed, which is appropriate for both clinical and laboratory settings. This approach is based on a spatial feature extractor (SFE) and a global temporal feature extractor and estimator (GTFEE). The SFE employs two residual networks to enhance interchannel correlations and capture spatial information. The GTFEE module consists of a multilayer perceptron mixer (MLP-Mixer) network and an SpO ₂ estimator. The MLP-Mixer operation facilitates global temporal information exchange across positions by rearranging and swapping the features along the channel dimension. Additionally, the Segment Anything Model (SAM) was first utilized to accurately extract skin regions at the pixel level to measure SpO ₂ . Finally, we are the first to explore the topic of contactless SpO ₂ estimation based on facial videos in a clinical environment; remarkable results were achieved, with a mean absolute error (MAE) of 1.79 and a root mean square error (RMSE) of 2.85 on the intensive care unit (ICU) dataset. Moreover, the extensive experimental results further validate that CCSpO ₂ Net outperforms the state-of-the-art approaches significantly, achieving MAEs of 0.54 (improved by 16.7%) and 2.48 (improved by 101.6%) on the PURE and smartphone camera oximetry (SCO) datasets, respectively.

Millimeter wave (MMW) communication, noted for its merit of wide bandwidth and high-speed transmission, is also a competitive implementation of the Internet of Everything (IoE). In an always-connected world, mutual data transmission and localization are the primary issues, such as the application of MMW application in autonomous vehicles and intelligent robots. Recently, artificial intelligence technologies have been adopted for the issues in the MMW communication domain. In this paper, MLP-mmWP, a deep learning method, is proposed to localize the user with respect to MMW communication information. The proposed method employs seven sequences of beamformed fingerprints (BFFs) to estimate localization, which includes line-of-sight (LOS) and non-line-of-sight (NLOS) transmissions. As far as we know, MLP-mmWP is the first method to apply the MLP-Mixer neural network to the task of MMW positioning. Moreover, experimental results in a public dataset demonstrate that MLP-mmWP outperforms the existing state-of-the-art methods. Specifically, in a simulation area of 400 × 400 ${\mathrm{m}}^2$, the positioning mean absolute error is 1.78 m, and the 95th percentile prediction error is 3.96 m, representing improvements of 11.8% and 8.2%, respectively.