一种语音信号重建算法

为了减轻因信包丢失而造成的语音失真,提出了一种基于双边线性预测的信包丢失隐藏算法。这种方法利用丢失信包的前一信包或邻接信包(在后一信包可获得的情况下)预测丢失信包,通过线性加权双边线性预测的样点获得最终的重建信号,使用重叠相加和幅度调整操作平滑重建信号和真实信号之间的边界。经过非正式试听和ITU-T P.862协议所推荐的PESQ算法测试,该算法的重建语音信号质量与其他四种流行重建算法相比,有了较为明显的改善。

A Speech Signal Reconstruction Algorithm

In this paper,a two-sided linear prediction-based packet loss concealment algorithm is presented to alleviate the speech distortion caused by lost packets in VoIP.This algorithm predicts lost packets from speech segments via either preceding or both preceding and subsequent packets(If subsequent packet is available).Predicted samples are smoothed by linear weighting to obtain the final reconstructed signal.It also performs overlap-and-adding and amplitude-scaling operations to alleviate discontinuity at packet boundaries.The assessment of resultant speech quality via informal listening test and objective measurement with PESQ algorithm from ITU-T P.862 shows that the proposed algorithm has a significant improvement over other four popular reconstruction algorithms.