Inverse Decoding of PCM A-law and µ-law

In this paper, we present a new approach for detecting traces of PCM A-law and PCM µ-law encoding within 16-bit PCM content. For both codecs, no previous detection approaches have been developed, despite the fact that they are often applied for land-line phone calls. An evaluation of our method has been carried out using the publicly available Librispeech ASR Corpus as a basis. The detection of both codecs, using analysis windows of 0.1s, is highly reliable even if automatic dithering noise is present, with an Equal Error Rate (EER) of 0% for PCM A-law and 1.05% for PCM µ-law.