发明名称 SIGNAL SECTION PREDICTION APPARATUS, METHOD, PROGRAM AND RECORDING MEDIUM THEREOF
摘要 PROBLEM TO BE SOLVED: To provide a means for accurately estimating sections where target signals, such as voice signals and music signals exist, from a mixed signal in which strength changing noise and sporadic noise are included. SOLUTION: According to the invention, a power ratio between periodical component power which is a feature amount with superior noise resistance, and non-periodic component power, is used. That is, a sound signal is input, and the sound signal is segmented for each predetermined section. The power of the segmented sound signal is calculated (11). Based on the fundamental frequency of the segmented sound signal estimated (14), the calculated power and the estimated fundamental frequency, the calculated power is divided into periodic component power and the periodic component power (20a). The power ratio between the periodical component power and the non-periodical component power is calculated (17), and whether target signal is contained in the acoustic signal cut out is determined, depending on the power ratio is larger than the prescribed threshold T. COPYRIGHT: (C)2008,JPO&INPIT
申请公布号 JP2008064821(A) 申请公布日期 2008.03.21
申请号 JP20060239796 申请日期 2006.09.05
申请人 NIPPON TELEGR & TELEPH CORP <NTT> 发明人 ISHIZUKA KENTARO;NAKATANI TOMOHIRO;MIYAZAKI NOBORU
分类号 G10L11/00;G10L11/04 主分类号 G10L11/00
代理机构 代理人
主权项
地址