The increase of performance has always been a key topic of the research activities on the internal combustion engines. Nowadays this is even truer as the performance is strictly correlated to the pollutant emissions. In this sense, an interesting approach could be the improvement of the effectiveness of engine control system and optimize the combustion process. To pursue this goal it would be very important to know the in-cylinder pressure during engine operation. The measurement of this quantity is performed generally with a pressure sensor flush mounted on the cylinder head. The measurement is very accurate, but the severe ambient conditions strongly limit the lifetime of these sensors, which, therefore, are not well suited to act as a feedback to the control system of on-road engines. Even though several approaches to measure indirectly the in-cylinder pressure have been developed, their diffusion is still hampered by reliability and sturdiness problems. To overcome these issues, an innovative methodology to measure the in-cylinder pressure has been conceived and tested on a two strokes single cylinder engine. The proposed approach was based on the analysis of the mechanical stress of the engine studs. Particular attention was paid to the signal processing and elaboration (based on a neural network architecture) to obtain a reliable in-cylinder pressure description.