发明名称 VOICE RECOGNIZING SYSTEM
摘要 PURPOSE:To decrease erroneous recognition by stressing parts which are not similar and executing matching. CONSTITUTION:Input patterns are formed from an input signal in a characteristic extraction part 2. A control section 9 changes a switch 3 to a standard pattern registration section 4 side to form a standard pattern when a registration mode is assigned by a keyboard 8. The switch 3 is changed to a collation part 6 side at every formation and the standard pattern of a standard pattern memory section 5 and the input pattern are collated. A mark is put when the similarity is larger than the threshold. The input pattern is fed to the collation section 6 where the matching of the input pattern and the standard pattern of the memory section 5 is executed in the case of assigning the recognition mode by the keyboard 8. The segment data of the similar marks are thinned out between the respective candidates in order of the larger similarity. Thus, the matching of the input pattern and the standard patterns of a 1st and 2nd candidates is executed again and the marks of the larger similarity are outputted to a result display section 7, by which the wrong recognition is decreased.
申请公布号 JPH02247698(A) 申请公布日期 1990.10.03
申请号 JP19890070033 申请日期 1989.03.20
申请人 RICOH CO LTD 发明人 KAWAMOTO TOSHIKI
分类号 G10L15/28;G10L11/00;G10L15/06;G10L15/10 主分类号 G10L15/28
代理机构 代理人
主权项
地址