Apriori算法在语音合成应用中的一种改进

An Improvement to Apriori Algorithm in the Application of Speech Synphesis

针对目前语音合成技术中提高语音合成自然度这一难点,研究了数据挖掘在语音合成中的应用.首先通过选取基频曲线中的最高音基频值(N1)和最低音基频值(N2)优化韵律参数,然后将其运用到关联规则对韵律参数进行规则提取的方法中,提出了优化韵律参数后的规则提取过程,并对原有的Apriori算法进行改进而获得更适合语音合成的HLApriori算法,通过该算法可以将原有Apriori算法得到的规则进一步细分,从而得到更多研究者感兴趣的规则.

This paper studied the application of data mining technology in speech synphesis in order to overcome the current technical difficulty in improving the spontaneousness of speech synphesis. Firstly, we optimized the prosodic parameter by choosing the highest pitch values(N1) and the lowest pitch values(N2). Secondly, we applied this optimized prosodic parameter to the rules concerned to pick out the rules for the prosodic parameter. We put forward the process of distilling rules after the optimized prosodic parameter. We proposed a new algorithm,HLApriori algorithm, to better fit the speech synphesis, and obtain more interesting rules for investigators.