Abstract: Sleep is important for the survival of human beings．Collecting and analyzing EEG signals in sleep will help clinicians and researchers to diagnose and study the health of human subjects．The application of deep learning technology to the study of sleep EEG, of sleep spindle in particular, is discussed．Deep learning algorithm was found to show higher accuracy and stronger data adaptability in comparion with traditional signal processing algorithm．To further improve detection and applicability of the network, feature fusion and spiking neural networks are proposed．Higher detection performance demonstrates potential of deep learning technology in sleep EEG analysis.