A Bark-band residual noise model integrated with the human hearing mechanism is proposed to efficiently complement sinusoidal model in parametric audio coding. The time-varying spectrum of the residual noise is retrie...A Bark-band residual noise model integrated with the human hearing mechanism is proposed to efficiently complement sinusoidal model in parametric audio coding. The time-varying spectrum of the residual noise is retrieved by Bark-scale piecewise constant magnitude estimates along with random phases. In the proposed noise model, Bark bands information is obtained by short-time FFT method and window overlap-add technique is exploited to remove boundary discontinuities. SVQ is also incorporated into parameter quantization process for the low bit-rate coding demand. Simulation results and informal listening tests show that when the sinusoidal model is combined with the Bark-band noise model, better synthesis audio quality can be achieved compared with the original sinusoidal modeling audio codec.展开更多
为提高噪声模型的估计精度,改善系统率失真性能,文中提出了一种基于残差子带分组聚类的自适应噪声模型估计方法.首先根据频率高低对残差子带进行分组,然后由组内子带残差样本生成特征矢量,进而利用改进的模糊c-均值聚类算法对当前解码...为提高噪声模型的估计精度,改善系统率失真性能,文中提出了一种基于残差子带分组聚类的自适应噪声模型估计方法.首先根据频率高低对残差子带进行分组,然后由组内子带残差样本生成特征矢量,进而利用改进的模糊c-均值聚类算法对当前解码子带进行聚类,最后计算出每类残差系数的噪声参数.实验结果表明,相比于相邻子带聚类-方差估计算法,文中所提算法能够更加准确地匹配残差分布特征,率失真性能平均提升0.60 d B,且解码时间平均节省40.59%.展开更多
文摘A Bark-band residual noise model integrated with the human hearing mechanism is proposed to efficiently complement sinusoidal model in parametric audio coding. The time-varying spectrum of the residual noise is retrieved by Bark-scale piecewise constant magnitude estimates along with random phases. In the proposed noise model, Bark bands information is obtained by short-time FFT method and window overlap-add technique is exploited to remove boundary discontinuities. SVQ is also incorporated into parameter quantization process for the low bit-rate coding demand. Simulation results and informal listening tests show that when the sinusoidal model is combined with the Bark-band noise model, better synthesis audio quality can be achieved compared with the original sinusoidal modeling audio codec.
文摘为提高噪声模型的估计精度,改善系统率失真性能,文中提出了一种基于残差子带分组聚类的自适应噪声模型估计方法.首先根据频率高低对残差子带进行分组,然后由组内子带残差样本生成特征矢量,进而利用改进的模糊c-均值聚类算法对当前解码子带进行聚类,最后计算出每类残差系数的噪声参数.实验结果表明,相比于相邻子带聚类-方差估计算法,文中所提算法能够更加准确地匹配残差分布特征,率失真性能平均提升0.60 d B,且解码时间平均节省40.59%.
