An digital audio can be easily recorded by handheld devices such as digital voice recorders and Smartphone. A nd these audio are using as evidence in courts in many cases . One of the most important problem is that many audio often contains content forgery. Here we analyze the authenticity of AMR audio . A AMR is a audio codec for speech compression ,these format is widely used in today's handheld devices such as digital audio recorder or in Smartphone etc . Designing hand-crafted features is a challenging and time-consuming problem. In this paper, instead of manually extracting the features, we investigate how to use deep learning techniques in this audio forensics problem.. For an audio clip with many frames, the features of all the frames are aggregated and classified by classifier . Here we use three classifier and compare them .Instead of hand-crafted features, we used the SAE to learn the optimal features automatically from the audio waveforms. Audio frames is the input to feature extractor and the last hidden layer’s output constitutes the features of a single frame. At last the features of all the frames are aggregated and classified by either UBM-GMM or SVM or Bayesian classifier . when comparing and analyzing these three classifier the SVM and Bayesian classifier shows high degree of accuracy for detecting the authenticity of an audio .