Mining dumpers are heavy-duty off-highway trucks designed for transporting materials, such as ore, overburden, or other mined materials, within a mining site or between a mine and a processing plant. These trucks are essential in the mining industry, facilitating the efficient transport of large quantities of materials across distances, often through various activities, challenging terrain, and harsh environmental conditions. Indeed, dumpers’ activities are critical factors in determining their cycle times. Many factors contribute to a dumper's overall cycle time. Optimizing the dumper activities is essential for improving the efficiency and productivity of mining operations. This paper presents a suitable approach for identifying dumper activities using statistical features and power spectral density (PSD) from the vibration signals of the dumper. A comparison of artificial neural network (ANN) and convolutional neural network (CNN) models was used to find dumper activity and their performance was tested using evaluation metrics. The results indicate that while both ANN and CNN models demonstrated high accuracy in recognizing dumper activities, CNN outperformed ANN in accuracy. Additionally, the study highlighted the importance of selecting appropriate features and model architectures for achieving optimal performance in dumper activity recognition tasks. The study investigated the efficacy of the models by conducting a case study using vibration data from two dump trucks.