Automatic sleep staging is of great significance for the diagnosis and treatment of sleep disorders. However, the published works for sleep staging have not efficiently characterized temporal information from EEG, and thus face the problems of future information leakage and high time cost for model training. To address the above issues, this paper proposes a novel network architecture named CausalAttenNet. It avoids future information leakage and extracts time-relevant sleeping features with wide temporal receptive field through causal dilation convolution. The traditional recursive network is replaced by multi-head attention to largely reduces time cost and form a faster network for sleep staging. The accuracy rates of CausalAttenNet on the Sleep-EDF-20, Sleep-EDF-78 and Sleep Heart Health Study (SHHS) datasets are 84.7%, 81.6% and 83.3%, respectively, which are better than those of the state-of-the-art methods. The performances of the proposed CausalAttenNet on different datasets demonstrate its effectiveness and superiority in the field on sleep staging. Individual difference is allowed and hand-crafted feature design is unnecessary for CausalAttenNet. The high accuracy and efficiency of the proposed method indicate its clinical potential for detecting sleep disorders.

Roboticized coal mining equipment is an important way to achieve intelligent coal mining. The perception system of robots is crucial for coal mine robots to understand their operating environment and conditions, serving as the functional foundation for autonomous operation in coal mines. However, due to the harsh and complexly changing conditions in underground coal mine environments, it is difficult for robotized coal mining equipment to obtain information about the underground environment accurately and reliably. In response to the characteristics of underground coal mine environments, an adaptive enhancement method for coal-mine underground images based on no-reference image quality evaluation is proposed in that paper. Its core is to establish a quality evaluation model for visible light images for underground coal mines, and to develop an adaptive optimization algorithm for image enhancement parameters based on the improved Sparrow Search Algorithm. Relevant experiments are conducted to demonstrate the performance of the proposed method. Experimental results show that, for three common enhancement methods, including CLAHE, homomorphic filtering and AutomateMSRCR algorithms, the results obtained from the optimized algorithms are improved by an average of 41.28%, 2.67%, 19.06%, 43.62%, and 129.46% respectively in terms of the SMD function, entropy function, variance function, Brenner gradient function, and Laplacian gradient function indicators, and the proposed adaptive image enhancement method shows strong application potential for the scene perception tasks including the anchor-rod tail detection and etc.

Coal-rock cutting state (CRCS) recognition plays an important role in improving shearer automation and ensuring the safe and efficient mining of coal mines. However, due to the randomness and unevenness of the distribution of underground coal seams, the vibration monitoring data of different cutting states are usually imbalanced, and exhibit complex structural relationships. Therefore, this paper novelty proposes a hypergraph convolution rebalancing cascade broad learning (HCRCBL) model for CRCS recognition. Specifically, HCRCBL recasts the cascaded feature mapping nodes of the original cascade broad learning system (CBLS) through a random hypergraph convolution mechanism to fully leverage the high-order relationships among cutting vibration data. Meanwhile, according to the data prior distribution, a rebalancing weight assignment strategy is designed for HCRCBL to pay more attention to the minority class and low-density distribution samples, which enhances the CRCS recognition performance of imbalanced data. Besides, HCRCBL adopts the ridge regression method to obtain the global optimal solution, assuring its high computation efficiency. The effectiveness of the HCRCBL-based CRCS recognition framework is demonstrated via the monitoring data from a coal-rock cutting experiment platform of shearer. The comparison results verify that the proposed method achieves high recognition accuracy while costing low training time consumption and effectively handling imbalanced data.