Abstract: Electroencephalogram (EEG) is one of the useful methods in analysis and simulation of different organs of human body. Human hand motion detection is one of the interesting issues in robotic, computer vision and many other applications. One of the main problems in human hand motion detection is inaccuracy in classification of extracted features from EEG signals. Although many motion detection methods have been proposed and developed, many of them suffer from extracting less accurate data from EEG signals. This paper proposes an effective feature extraction model to enhance the quality of hand motion detection using a combination of obtained feature extraction parameters from autoregressive model, Hjorth parameters, fractional dimensions in time, frequency, and spatial domains. Furthermore, one-second Hamming window with half-second overlap is used for signal windowing. The Competition-III data set and mean absolute error of prediction method are used to evaluate the performance of the proposed method. The obtained results indicate that the proposed method shows more accuracy in feature classification when compared with the other hand motion detection methods.