Abstract: The vibroarthrographic (VAG) signal is the vibration and contact friction generated by the knee joint during flexion and extension. It reflects the characteristics and status of chondromalacia, meniscal tears and ruptured ligament and other diseases, and it is gradually getting the attention of clinical medicine. In this paper, the characteristics of normal and abnormal VAG signals are quantitatively analyzed according to the multifractal detrended fluctuation method. The features of the VAG signals are extracted, including the fractal scale exponent, the multifractal spectrum and extremum values, the generalized fractal dimension and the time-frequency information entropy. SVM is used to classify the normal and abnormal VAG signals, getting a high classification accuracy, which is very important for noninvasive detection and assistant diagnosis of knee joint  injury.