Classification of microphones of mobile devices via blind channel estimation

The present work concerns microphone classification, a research topic still in its early stages strictly related with audio authenticity and forgery detection. In this paper we propose a new method of microphone classification based on a recent single-microphone blind channel estimation algorithm, focusing on audio recordings from mobile devices: this method aims to correctly identify the source device with high accuracy, even if the sample audio file was compressed at extremely low bit-rates. A Support Vector Machine (SVM) was used in order to perform closed-set identification experiments, and to assess the performance of the algorithm.