论文部分内容阅读
基频轮廓就是基频随时间变化的曲线,刻画了汉语中最主要的韵律特征—声调和语调的变化趋势.而藤崎模型则是由日本东京大学藤崎博也教授建立的一种韵律模型,可以用来非常好地逼近汉语语音的基频轮廓(1).本文提出了一种基于该模型的从汉语单音节的基频轮廓中提取参数的方法.我们首先采用基于小波变换的基音检测技术(3)获取非常准确的给定单字的各基频值,并连接形成基频轮廓,然后根据最小均方误差准则,用藤崎模型来拟合各基频点,最后分析得出模型的最优化参数,作为此基频轮廓的参数.实验结果充分说明了这一方法的合理性和可靠性
The fundamental frequency contour is the curve of the fundamental frequency changing with time, and depicts the most important prosodic features in Chinese - the trend of tone and intonation. The Fujisaki model, on the other hand, is a prosody model built by Professor Fujisaki Hiroyuki of the University of Tokyo in Japan, which can be used to very well approximate the fundamental frequency profile of Chinese speech (1). This paper presents a method based on the model to extract parameters from the fundamental frequency contour of Chinese monosyllable. We first use pitch detection technology based on wavelet transform (3) to obtain very accurate reference value of a given word, and connect to form the basic frequency contour, and then according to the minimum mean square error criterion, with Fujisaki model to fit the fundamental frequency Point, the final analysis of the model to obtain the optimal parameters, as the parameters of the pitch profile. The experimental results fully demonstrate the rationality and reliability of this method