摘要
通过对ＬＰＣ（线性预测编码）方法提取语音信号共振峰进行的研究表明，采用相－频特性与对数幅－频特性同样能提取语音信号共振峰。与对数幅－频特性的二次导数相比，相－频特性的三次导数有更高的频率分辨率，能更有效地解决共振峰合并的问题，提取更精确的共振峰参数。
The study on speech signal formant abstracted by LPC shows that it can also be abstracted by adopting phase-frequency characteristics and logarithmic amplitude-frequency characteristics. Compared with the second derivative of logarithmic amplitude-frequency characteristics, the third derivative of phase-frequency characteristics has better frequency resolution, so it can solve the combination of formant more effectively and abstract formant parameters more precisely.
出处
《电声技术》
北大核心
2000年第3期3-8,共6页
Audio Engineering
基金
浙江省自然科学基金!698064
