Landslides are powerful, destructive events that can generate slidequakes. Monitoring passive slidequake signals can facilitate hazard assessments, dynamic calculations, and early warning. However, there have been few direct observations of landslide dynamics. This hampers the classification and identification of landslide movements based on spectral features of slidequake signals. Here, physical analogue tests were conducted to simulate the landslide movements of three boundary conditions: open flat (OF), gully constrained curve (GC) and barrier blocking (BB). Different slope gradients, movement paths, blocking conditions and slide material structures were employed in 96 experiments to analyse the spectral features of slidequake signals and compensate for the lack of active observation data. Via spectrum identification, we analysed the vibration characteristics of the collected three-component high-frequency slidequake signals of analogue modelling landslides. Then, we modelled the landslide dynamics through feature extraction of the fractal box dimension of the slidequake signals. The OF and BB slidequake signals were identified with a prediction accuracy of 87.5%. Hence, the accurate classification and identification and capability for early warning of remote landslide dynamics based on passive slidequake signal monitoring is presented.