论文部分内容阅读
提出一种用于可变码率音频编码的正弦+噪声(SN)模型。提出了对正弦模型进行本质上的增强。从大、中、小三个前后衔接的尺度上对音频信号进行时域重叠相加(overlap-add)的正弦分析时引入了心理声学模型加权的匹配跟踪算法(matching pursuits algorithm),将大尺度正弦分析-合成后的余量送入相对小的尺度进行分析,以达到相应的分辨率。这种算法有效的解决了正弦模型固有的预回声效应,提高了重建音频的质量。这一模型适用于可变码率、高保真的音频压缩和发音速度、音调的调整。
A sinusoidal + noise (SN) model for variable rate audio coding is proposed. Proposed to the sinusoidal model of the nature of the enhancement. In the sinusoidal analysis of the overlap-add of the audio signals from the large, medium and small coalescence scales, a psychoacoustic model-weighted matching pursuit matching algorithm is introduced. The large-scale sine Analysis - After the synthesis of the margin into the relatively small scale analysis, in order to achieve the appropriate resolution. This algorithm effectively solves the inherent pre-echo effect of the sinusoidal model and improves the quality of reconstructed audio. This model is suitable for variable bit rate, high-fidelity audio compression and pronunciation speed, pitch adjustment.