计算机系统结构学习指导与题解
计算机系统结构教程配套的学习指导和题解,适用于期末复习。
矩阵入门必读,非常详细,内容包括矩阵的各种运算技巧,可以当做工具性的参考书,随查随用
论文中带有相关的笔记和注释,方便查看和修改、标注.主要内容为利用低通滤波实现声门闭合时刻的检测和利用CNN实现声门闭合时刻的分类.
Determination of the instants ofglottal closure (GC) from speech wave using wavelet transform is equivalent to finding a particular local modulus maxima pattern across several scales in the time-scale plan. It is shown that the local modulus maxima of wavelet transform corresponding to the event of GC are generally zigzag around the instant of GC and that the amount of the shift of the local modulus maxima from the instant of GC when evolving across scales can be well sketched with an evolution cone of GC maxima. An efficient algorithm that make use of the evolution cone of GC maxima to detect the instant of GC is presented together with three erformance measurements regarding the accuracy of location, false alarm rate, and fail alarm rate of the GC instant detection algorithm. Intensive experiments are carried out to test the validity of this novel method.
根据语音信号在声门闭合时刻(GCI)的锐变特性 ,通过检测语音信号小波变换发生在 GCI 附近的局部极大值的位置估计声门闭合时刻 ,两个相邻 GCI 的距离为该段语音的基音周期 ,它的倒数即为基频 F0 .实验结果表明 ,采用本方法检测到的基音周期准确 ,能按基音周期及时反映它的迅速变化 , 且抗噪声性能强 ,还能可靠的判别清/浊音
利用小波变换模极大值点同信号突变点之间的关系, 以及声门闭合时刻(GCI)引起语音信号锐变的特点, 通过检 测两个相邻 GCI 的距离来估计语音的基音周期, 由此求出基音频率对时间的化曲线(即声调曲线)。 通过仿真, 该方法提取声调准确, 而且抗噪音的能力比较强