辽宁石油化工大学
TP391
辽宁省教育厅基本科研项目面上项目
睡眠分期在睡眠障碍诊断中具有重要意义。目前的自动睡眠分期方法大多集中在研究时域信息,且睡眠阶段之间的过渡规则往往无法被识别和捕获,导致睡眠分类准确率低。为解决这一问题,提出基于单通道脑电(EEG)信号的融合多尺度特征和注意力门控多层感知器的睡眠分期方法(Multi-scale features and Attention gated multi-layer perceptron SleepNet? MA-SleepNet)。该模型由多尺度特征提取模块、压缩激励网络模块和注意力门控多层感知器网络模块组成。多尺度特征提取模块采用双通道卷积从脑电信号中提取不同尺度波形特征；压缩激励网络模块采用压缩激励模块学习多尺度特征的重要程度,提升有效特征；注意力门控多层感知器模块将多层感知器与门控机制结合起来,同时加入简单的自注意力机制,实现不同维度之间的数据通信,整合信息中的有效特征。在Sleep-edf-20和sleep-edf-78数据库上MA-SleepNet模型分别达到了86.1%和83.2%的睡眠分期准确率。与现有典型研究结果相比,该方法提高了分类性能。
Sleep staging has attracted much attention as an important method for studying sleep disorders in recent years. The majority of the current automatic sleep staging methods focus on studying time-domain information and ignore the interrelation between features, resulting in low sleep classification accuracy. To solve these problems, a multi-scale features and attention gated multi-layer perceptron mechanisms named MA-SleepNet is proposed for automatic sleep stage classification, using single-channel electroencephalogram (EEG) signals. The network consists of a multi-scale feature extraction (MFE), squeeze and excitation network (SE), and an attention gated multi-layer perceptron mechanism(aMLP). The MFE module uses convolutional kernels of different sizes to fully extract different scale features from EEG signals. The SE module further optimizes the weight of features and improves the feature expression ability of the network. The aMLP module combines multi-layer perceptron with gating mechanism, adds tiny self-attention mechanism to realize data communication between different dimensions and integrates powerful feature representation.The MA-SleepNet model is evaluated on two public datasets, Sleep-EDF-20 and Sleep-EDF-78. It achieves the accuracy of 86.1% and 83.2% on the Fpz-Cz channel, respectively. Compared with the existing sleep staging methods, our method improves the classification performance.
