论文部分内容阅读
提出了一种基于希尔伯特-黄变换瞬时能频值的含噪耳语音声韵分割算法。根据耳语音声韵母幅频特性,运用HHT,分离出耳语音中高频部分的瞬时幅值与频率,同步去除低频噪声,提取出能够反映声韵母过渡信息的特征参数——瞬时能频值,利用该参数对耳语音进行声韵分割。实验结果表明:与相对熵算法相比,该算法对含噪耳语音进行的声韵分割正确率较高,能够较准确地进行耳语音声韵分割。
This paper proposes a noisy speech phonological segmentation algorithm based on the Hilbert-Huang transform instantaneous energy value. According to the amplitude frequency characteristics of ear phonetic vowel, HHT is used to separate the instantaneous amplitude and frequency of middle and high frequency part of whispered speech. The low frequency noise is removed synchronously. The characteristic energy instantaneous frequency value, which can reflect the transition information of vowels, is extracted. Parameters of the ear phonology rhyme segmentation. The experimental results show that compared with the relative entropy algorithm, this algorithm has a higher correct rate of rhyme segmentation for noisy-ear speech, which can be used to segment ear phonology more accurately.