A critical aspect of intelligent transportation systems, including autonomous vehicles, is the ability to detect anomalies on road surfaces accurately. This information is vital for ensuring the safety of both passengers and autonomous vehicles. In this study, we introduce a new method for road surface detection that utilizes data from inertial sensors and vehicle speed, processed through a Long Short-Term Memory (LSTM) model. We evaluated our model using publicly accessible data collected from inertial sensors mounted on a vehicle’s dashboard and corresponding labels. Our results show that our approach effectively identifies different road surface types with a high classification accuracy of approximately 95.02%.