论文部分内容阅读
本文采用自回归（ＡＲ）模型，对汉语元音（正常儿童）进行时间序列分析。ＡＲ模型根据Ｍａｒｐｌｅ递推算法得到线性预测系数，然后求出汉语元音的功率谱密度估计。分析结果显示：ＡＲ模型分析得到功率谱曲线较光滑，谱峰清晰可辨，能较好地反映汉语元音在频域上的特性，克服了以往用ＦＦＴ法所得功率谱曲线是锯齿状，需计算语频包络中振幅最大的谐波频率的偏峰值作为共振频率值的不足之处。
In this paper, an autoregressive (AR) model is used to analyze Chinese vowels (normal children) in time series. The AR model obtains the linear prediction coefficient according to Marple recursion algorithm, and then obtains the power spectral density estimation of Chinese vowel. The analysis results show that the power spectrum curve is smoother and the spectrum peak is clear and distinct, which can better reflect the characteristics of Chinese vowel in the frequency domain. The power spectrum curve obtained by the FFT method is jagged, Calculating the peak envelope of the amplitude of the harmonic peak frequency as the resonance frequency value of the inadequacies.