Crack on piezoelectric ceramic is the main reason leading to the degradation of ultrasonic motors. Monitor electrode voltage generated by the positive piezoelectric effect is usually used to monitor the vibration of stator. However, the complexity of monitor electrode voltage signal changes slightly until failure state because of the bonding layer between piezoelectric ceramic and metal elastomer, and it brings difficulties for the degradation state identification. In order to improve the accuracy rate, a method based on segmented fractal dimension and sparse representation was presented in this article. Firstly, segmented fractal dimension was proposed to extract the complexity information of local signal, and the ones of standard samples were taken as the atoms to construct a dictionary. Then, sparse representation of the test sample was calculated according to the constructed dictionary, and the specific steps for the solution were also detailed. Lastly, the test sample’s deviation vectors corresponding to different degradation states were obtained, and the modulus of the vectors were employed to identify the degradation state. The experimental results show that this method is feasible and effective for the degradation state identification of piezoelectric ceramic. It is meaningful for the condition-based maintenance of ultrasonic motors.