汉语单词语声区的检测

汉语单词语声区的检测 DETECTING THE ENDPOINTS OF ISOLATED CHINESE UTTERANCES 李汉 1 2 first-author 原勇 1 2 宋骞 1 2 王延平 1 2 武汉大学物理系 武汉大学物理系 武汉大学物理系 研究生 武汉大学物理系 研究生 准确地判定在噪声背景中语声区的位置,不但在语声处理中可以有效地限制待处理的数据量,而且对语声识别率,也有极大影响。已有一些文献专门讨沦这一问题,并且企图找到简单可靠的方法检测语声区的端点。本文分析了汉语发声起终点的特点,指出可以只用能量作为汉语语声区的检测量,并通过实验确定出了起终点鉴别机构中的待定参量值。 The determination of the location of the endpoints is very important for speech recognition. It influences the recognition rate as well as processing effieciency. According to the analysis of Chinese utterances and a large amount of experiments, it is found that to use energy to derive the several criterions is enough for isolated Chinese word recognition The algorithm based on this idea is given in this paper 1984-01-01 2021-04-01 1