Smartphones, as a new tool for structural health monitoring, require further validation of the signal accuracy collected by their built-in accelerometers and the effectiveness of signal preprocessing methods. This paper presented a shaking table test on a scaled model of a five-story single-bay hybrid structure composed of wooden modules and a reinforced concrete frame. To address the issue of baseline drift in the integrated acceleration signals collected by smartphone accelerometers, the feasibility of the methods was demonstrated by comparing the detrending effects and signal characteristics of time-history curves processed by EMD and EEMD. The test results showed that the signals processed by the EEMD method were richer than those processed by the EMD method. On the basis of removing the trend term of the original signal, the EEMD method effectively reconstructed and preserved the signal characteristics of the velocity time-history curve, making it more suitable as a preprocessing method for the integrated acceleration signals measured by smartphone accelerometers.