In order to explore the influence of human disturbance activities on soil organic matter content and improve the estimation accuracy of soil organic matter in arid area. The soil in Fukang City, northern Xinjiang, is studied. The hyperspectral curves of 90 sampling points are successively transformed by continuous wavelet transform (CWT) and compared with R′and lg(1/R), two common spectral transformation methods. The results show that the spatial variability of soil organic matter is enhanced with the increase of human disturbance degree. Moreover, in the common spectral transformation methods, R2 of partial least squares models established by the soil organic matter institute and R′ in zone I, zone II and zone III are both higher than those by R and lg(1/R). The precision of the model established after CWT is very high. Compared with R2 of the model established by R, R2 of the measured value and the predicted value are 0.717, 0.689, 0.630, increased by 0.382, 0.4, 0.389, respectively. In addition, the relative percent deviation correspondingly reach 2.150, 2.090, 2.013, indicating that CWT can well predict the soil organic matter contents. The usage of CWT does not greatly reduce the model accuracy with the increase of human disturbance degree and is more suitable for the prediction of organic matter content in arid regions.