发明名称 CODE BOOK GENERATOR, SPEAKER RECOGNIZING DEVICE, PROGRAM AND CODE BOOK GENERATING METHOD
摘要 <p><P>PROBLEM TO BE SOLVED: To provide a code book generator in which recognition rate is improved. <P>SOLUTION: A representative vector sequence is obtained from a featured vector sequence by a first code book generating means 11 and a code book is generated once. Then, an unnecessary vector discriminating means 12 conducts discrimination to determine whether each representative vector of the obtained representative vector sequence is an unnecessary representative vector. Based on the obtained discrimination result, a vector eliminating means 13 eliminates a vector from the featured vector sequence, and a second code book generating means 14 generates a code book again based on the remaining featured vector sequence. Thus, occurrence of the representative vector to be assigned to the featured vector such as friction sound and noise or the like is prevented, and the recognition rate is improved. <P>COPYRIGHT: (C)2006,JPO&NCIPI</p>
申请公布号 JP2005309138(A) 申请公布日期 2005.11.04
申请号 JP20040126787 申请日期 2004.04.22
申请人 TOSHIBA TEC CORP 发明人 YOSHIZAKI KEISUKE;KAKINO TOMONARI
分类号 G10L15/06;G10L17/00;(IPC1-7):G10L15/06 主分类号 G10L15/06
代理机构 代理人
主权项
地址