In this study, speckle-visibility spectroscopy is used to measure the time-resolved particle dynamics in the passive layer of a continuous granular flow in a half-full rotating drum. A signal diagram of the mean fluctuation in the speed of the particles, , which is equal to the collision velocity with noise, is obtained. We discuss the traditional threshold method of signal processing for the extraction and how statistics depend on several subjective factors that are chosen in the analysis process. We propose a new threshold method using wavelet analysis for extracting valid signals from noise. Experiments show that the useful signal can be clearly distinguished from the noise using wavelet analysis, and compared with the traditional method, the statistical distribution map of the is closer to the actual movement phenomenon. (C) 2018 Elsevier B.V. All rights reserved.