Abstract: ‍ ‍An electroencephalogram （EEG） records the brain biological potential by collecting electrical signals from the human scalp through precision amplification instruments. It is widely used in medical diagnosis and scientific research fields for its advantages of safety， low cost， and high temporal resolution. However， the amplitude of an EEG signal is weak， and the actual EEG signal is normally mixed with noises. In general， the EEG signal is vulnerable to the external environment and physiological activities during the process of acquisition， which means it can be contaminated easily. Among these noises， the one caused by physiological activities of the subjects overlaps with the pure EEG signals in the time or frequency domain， making it difficult to separate them from the EEG signal with simple preprocessing methods. Consequently， algorithms for the recognition and removal of EEG artifacts， able to effectively remove the noises， have been a research focus in the brain-computer field. Conventional algorithms for the removal of artifacts include regression， wavelet transform， empirical mode decomposition， and blind source separation. They separate artifacts and the pure EEG signal based on the time-frequency characteristics of the signal itself or the statistical characteristics between signals， and play an important role in the development and application of EEGs. However， although many studies have been conducted on EEG artifact removal， a method that can be applied to all cases has not been developed owing to the complexity of artifact components. This causes an unnecessary burden of choice for the matching between target and algorithm in practical applications. This paper contributes to the solution of the problem. First， the paper summarizes the causes and categories of artifacts， and explores the morphological characteristics of different physiological artifacts. This enables researchers from different fields to have a more detailed understanding of artifacts. Second， various advanced methods for removing artifacts from EEG signals globally are summarized. Further， the advantages， disadvantages， and differences in applicability of those methods in terms of artifact removal performance are discussed. This can provide a theoretical basis for researchers in different fields to choose suitable algorithms for removing EEG artifacts in the future. Finally， some existing problems in current research are analyzed， and the development direction of the research on EEG artifact removal is discussed.