| 摘要 |
PROBLEM TO BE SOLVED: To provide a method for generating a database of speech samples that can construct a system which can use information which cannot be made language as a key and intuitively retrieve a body that a user faces, a phenomenon that the user recognizes, unknown information, and information on a body and a phenomenon whose key words are forgot with sound information. SOLUTION: The largest value in speech data is detected and time intervals within a specified range (e.g. 0.5 second) before and after the time when the largest value is obtained are regarded as a sampling object. The largest value of speech data inside the sampling object range is regarded as an upper-limit value and the smallest value is regarded as a lower-limit value to grasp variation in speech data value as time-series bit data. Pieces of data of a plurality of time in the time-series bit data are divided into three by colors by using a proper tricolor coloration system such as an RGB coloration system to generate gradation data by the pieces of divided data, thereby a database of those pieces of gradation data is constituted. COPYRIGHT: (C)2004,JPO
|
