Despite advancements in instrumentation and control system of lower limb prostheses, individuals face significant challenges when walking on irregular surfaces, increasing the risk of falls. Current prosthetic designs often struggle to adapt dynamically to varying surfaces, leading to compromised mobility. The lack of automatic irregular surface recognition and adaptive control mechanisms in prosthetic systems exacerbates these challenges, diminishing users’ confidence and safety while walking. Addressing this issue is crucial to enhance the functionality and safety of lower limb prostheses. Combination of inertial measurement units and artificial intelligence approaches such as deep learning methods (DL) have emerged to detect walking surfaces and provide an alert to individuals while walking in irregular surfaces. This paper conducts a pilot study toward development of a robust intelligent surface recognition approach for lower limb prosthesis that can predict walking surfaces and notify user accordingly. In this study, IMU data were collected from top of the prosthetic leg worn by lower limb amputees. Statistical tools were applied to analyze significant changes in IMU data (acceleration and gyroscope) across different surfaces. The results indicate that acceleration data exhibit greater variability compared to gyroscope data. Then three DL methods, such as Long short-term memory (LSTM), Convolution neural network (CNN) model, and a hybrid CNN-LSTM model, were employed utilizing most significantly varying IMU data to classify regular and irregular surfaces, with a focus on achieving high accuracy while processing fewer input data. Consequently, the CNN-LSTM hybrid model demonstrated superior performance in detecting irregular surfaces using data from top of the prosthetic leg of amputees, achieving a mean classification accuracy of 97.17 ± 3.42%, outperforming other DL models. The findings of this study may contribute to IMU based AI software development that can be integrated into prosthetic legs, thereby enhancing walking safety for lower limb amputees’ healthcare.