Audio Recording Device Identification Based on Deep Learning

摘要

In this paper we present a research on identification of audio recordingdevices from background noise, thus providing a method for forensics. The audiosignal is the sum of speech signal and noise signal. Usually, people pay moreattention to speech signal, because it carries the information to deliver. So agreat amount of researches have been dedicated to getting higherSignal-Noise-Ratio (SNR). There are many speech enhancement algorithms toimprove the quality of the speech, which can be seen as reducing the noise.However, noises can be regarded as the intrinsic fingerprint traces of an audiorecording device. These digital traces can be characterized and identified bynew machine learning techniques. Therefore, in our research, we use the noiseas the intrinsic features. As for the identification, multiple classifiers ofdeep learning methods are used and compared. The identification result showsthat the method of getting feature vector from the noise of each device andidentifying them with deep learning techniques is viable, and well-preformed.