Forensic audio authentication techniques include an extensive toolbox for verifying recording originality and exposing forgery traces. The AMR (adaptive multirate) codec is a widespread standard audio format used to store speech in smartphones or digital recorders, making AMR audio is an important source of audio to be authenticated. An original AMR audio file is supposed to be single compressed, while a tampered version should be double compressed. This makes AMR double-compression detection an interesting topic for multimedia forensics. In this paper a new method is proposed to detect AMR double compression using compressed domain features based on linear prediction coefficients and a support vector machine (SVM). By using a robust scaling procedure, a detection accuracy of 98% was achieved with TIMIT database, reaching the same performance as the state-of-the-art methods. The feature extraction and computation are designed for the specific problem, representing AMR audio files by a more appropriate set of vectors. The main conclusion is that robust scaling is a key tool to increase the accuracy of the method. Compared to the previous experiments using compressed-domain features and min-max scaling, the method offers an increase of about 5% in average accuracy.
Click to purchase paper as a non-member or login as an AES member. If your company or school subscribes to the E-Library then switch to the institutional version. If you are not an AES member and would like to subscribe to the E-Library then Join the AES!
This paper costs $33 for non-members and is free for AES members and E-Library subscribers.