One of research fields in studying dynamics of gravel-bed rivers is the interaction between sediment particles in motion and incision rates in rock-bottom river reaches. This natural phenomenon of rock abrasion was studied in a laboratory in a Dubree-type abrasion (tumbling) mill with the diameter of 711 mm, using different mixtures of fluvial sediments as abrasive media. A set of rock plates of different lithologies was fixed to the inside mill wall to evaluate rock abrasion by moving sediment particles. The dynamicsof the abrasion process of the rock plates was studied by a spherical instrumented tracer with the diameter of 99 mm. This paper describes our solution to the problem of recognizing and differentiating between impacts of the instrumented tracer with different bodies: sediment particles, rock plates, soft lining of the mill and steel side plates of the mill. For this purpose, the signal analysis of measured 3D accelerations of the instrumented tracer gave sufficient information to recognize the tribological surrounding and sufficiently describe the intensity of the abrasion process (number and amplitudes of contact forces). An effective and computationally inexpensive algorithm for automatic impact recognition and evaluation was developed, basedon time domain analysis. Furthermore, the frequency domain analysis gave a method of discriminating different signals. Both mentioned methods allow us to classify all recorded signals into groups based on similarity of measurement conditions.