基于人工神经网络的多媒体语音特征挖掘

基于人工神经网络的多媒体语音特征挖掘 The Feature Mining of Multimedium Speech Recognition on Artificial Neural Network 曹加恒 CAO Jia- heng 1 2 first-author 曹斌 CAO Bin 1 2 张凯 ZHANG Kai 1 2 叶可 YE Ke 1 2 刘茂福 LIU Mao- fu 1 2 武汉大学数学与计算机科学学院/教育部多媒体软件开放研究实验室,湖北,武汉,430072 武汉大学数学与计算机科学学院/教育部多媒体软件开放研究实验室,湖北,武汉,430072 College of Mathm atics and Computer Science/ The Open Reserch L ab for Multimedia Software U nder the Departm ent of Education, Wuhan University,Wuhan 0 72 College of Mathm atics and Computer Science/ The Open Reserch L ab for Multimedia Software U nder the Departm ent of Education, Wuhan University,Wuhan 0 72 武汉大学数学与计算机科学学院/ 武汉大学数学与计算机科学学院/ 提出了一种基于人工神经网络的聋儿语音训练识别的多媒体特征挖掘技术 .构造了邻域三层神经元合作竞争的动态行为神经元模型 ;实验选取了动态递减函数、动态最小覆盖矩阵和动态确定阈值形成的 SOL A挖掘算法 ;以及聚类分布的网络图技术 .解决了聋儿语音训练系统中基本语音识别的难题 . The mining techniques which use artificial neural network to solve the speech recognition for Deaf- mutes is given.The artificial neural model,its neighbour field,their interaction of co- operation and competition in neighbour fild,and the dynamic behavior of neurons,have been described.The SOLA- mining algorithm adopted descending function and dynamic threshood,the smallest covering rectangle, clustering distribution and mapping technique. The experimental result has shown that it can converge quickly. 人工神经元 动态行为 聚类分布 SOLA挖掘算法 artificial neural unit dynamic behavior clustering distribution SOLA- mining Algorithm 10.14188/j.1671-8836.2000.05.014 TN912 国家863计划资助项目!（ 863 -3 0 6-ZT0 4-0 1-1） 2000-05-01 2021-04-01 5