Analysis of Spectral Parameters of Audio Signals for the Identification of Spam Over IP Telephony

A method is presented which analyses the audio speech data of voice calls and calculates an "acoustic fingerprint". The audio data of the voice calls are compared with each other based on spectral parameters and voice calls are identified which have a high degree of similarity. The method which is resistant to various modifications of the audio signal can be used to detect SPIT which is typically characterized by similar or identical voice calls being distributed to a large number of receivers. Privacy protection is assured since only a fingerprint but not the complete audio data of a call is stored, which does not allow a reconstruction of the content or identification of the speaker.

• Type: Conference
• Year: 2008
• Where: CEAS
• Authors: Christoph Pörschmann, Heiko Knospe
• Keywords: Audio Data, Audio Speech Data, CEAS 2008, Internet Technology, Voice Calls