Aiming at the current problems of single feature extraction of motor imagery EEG signals and low accuracy of classification and recognition, a feature extraction method of motor imagery EEG signals based on the fusion of PSD and CSP (PSD-CSP) is proposed. Firstly, the FastICA algorithm is employed for artifact signal removal from the raw EEG data. Subsequently, features are extracted from the Power Spectral Density (PSD) and Common Spatial Pattern (CSP), followed by their serial fusion. Finally, a Support Vector Machine (SVM) classifier is used for classification. In binary motor imagery classification experiments on the BCI Competition IV Dataset I, the average classification accuracy reaches 91.43%, and comparative analysis with other methods demonstrates the feasibility of the proposed fusion algorithm.