Voice activity detection and discontinuous transmission for a GSM AMR speech codec

Abstract

In this thesis, a voice activity detector (VAD) and a discontinuous transmission (DTX) algorithms have been developed for the codec candidate. The VAD is used to detect whether the signal contains speech and background noise or only background noise. Taking advantage of the output of the VAD, the DTX is applied to switch off transmission during speech pauses. During the pauses, an approximation of the background noise is transmitted with silence descriptor (SID) frames. In the receiver, a synthetic noise similar to the transmit side background noise is generated according to the SID frames. This synthetic noise is called comfort noise. The DTX saves power in the mobile phone enabling smaller batteries and a longer operation time for the phone and reducing the interference between radio channels. The thesis presents the design process and evaluates the performance of the developed VAD and DTX.