The working environment of the quay crane is harsh and special. As an important transmission equipment of hoisting mechanism, health condition of gearbox is very important for reliable operation. In order to extract degradation features from complex vibration monitoring signals, an improved KW symbol entropy feature extraction technique is proposed. Considering the unity of symbol standard, the method takes the root mean square of the normal condition signal as the symbol standard and combines the symbol coefficient to construct a unified symbol scale. At the same time, symbols number variable is introduced to expand symbols set and improve the information expressing ability. On this basis, combining with information entropy theory, complexity of symbol sequence in symbol sequence and symbol distribution is calculated respectively, and two features named improved symbol sequence entropy (IKSE) and improved symbol distribution entropy (IKDE) are obtained. The Logistic chaotic sequence and the lifetime signal of the hoisting gearbox are used for analysis respectively. The result show that the proposed features are able to characterize the complexity of nonlinear time series, so as to describe the whole process of the performance degradation of the hoisting gearbox sensitively. The parameters have little influence and the technique have a good stability.