Forensic Recognition of Narrowband AMR Signals

A vast number of audio files are submitted to forensic analysis every year, originating from both lawfully intercepted signals over communication networks and user recording devices. The forensic recognition of such signals in uncompressed format revolves around the detection of single and double adaptive multi-rate (AMR) compression traces, as well as detection of modification traces of AMR signals. Given that the high-bitrate AMR compression traces easily eschew the human sensory scrutiny, there is a large interest in exact interpretation of such traces. In this paper, we propose a method to detect and recognize different narrowband AMR single or double compression modes, using a deep neural network.