论文部分内容阅读
首先对 ＭＢＥ模型分析算法的复杂度进行了深入的分析，认为原算法中的双闭环基音搜索结构有很大的改进余地．为此提出了一种基于Ｂ－样条二进离散子波变换（Ｂ—ＳＤｙＷＴ）基音检测法的改进ＭＢＥ模型分析算法。其关键在于利用Ｂ－ＳＤｙＷＴ基音检测原理将这种双闭环搜索算法改造成先开环后闭环的结构。理论分析可以说明这种改造可以大幅度地降低基音搜索的复杂度，而不降低其性能。通过仿真实验，我们对改进前、后的ＭＢＥ模型基音搜索算法的复杂度、准确度以及抗噪声能力等方面作了一个全面的对比分析。得出这样的结论：改进算法与原算法相比，各项性能均不低于原算法甚至有所改善，特别是证实了理论分析中的复杂度降低作用。
First of all, the complexity of MBE model analysis algorithm is analyzed in depth, and the double closed-loop pitch search structure in the original algorithm has a lot of room for improvement. To this end, an improved MBE model analysis algorithm based on B-Spline Binary Discrete Wavelet Transform (B-SDyWT) is proposed. The key lies in the use of B-SDyWT pitch detection principle of this double closed-loop search algorithm into a closed loop after the first open-loop structure. Theoretical analysis shows that such a modification can significantly reduce the complexity of the pitch search without degrading its performance. Through the simulation experiments, we make a comprehensive comparative analysis of the complexity, accuracy and anti-noise ability of the pre-and post-MBE model pitch search algorithm. The conclusion is drawn that compared with the original algorithm, the performance of the improved algorithm is not lower than that of the original algorithm or even improved, especially the complexity reduction in the theoretical analysis is confirmed.