Performance Analysis of AMR Voice Activity Detection Algorithm for Mandarin Based on Objective Test Method

To evaluate the performance of adaptive multi rate (AMR) voice activity detection (VAD) algorithm standardized by 3GPP applied to Mandarin, the formal objective test method is adopted as the performance metric, this kind of method is based on statistics of clipping probability, voice activity factor, noise detected as speech probability and speech input level. With different background noises and different signal to noise ratios, the AMR VAD algorithm is simulated and compared with both reference VADs including enhanced full rate (EFR) and ITU G.729B. The results show that the AMR VAD performs better than G.729B VAD and worse than EFR VAD in most of cases.