| 摘要 |
PROBLEM TO BE SOLVED: To provide an utterance signal extraction method for extracting a sound signal (that is, an utterance signal) of a whole utterance period, and to provide a speech recognition device and a program for performing the utterance signal extraction method. SOLUTION: In an utterance period extracting processing which is performed by the speech recognition device, a signal level P of the sound signal specifies a period which is a speech determination threshold Pt or more, continuously for a time threshold T1 or more , as a speech period, and stores it in a second memory (S260 to S320). A sound signal of a time threshold T3 before a reference start point which is a start point of the speech period (S250), and a sound signal of a time threshold T4 after a reference end point which is an end point of the speech period (S340), are also stored in the second memory as a part of speech retention period. When a user's instruction is inputted during a period (from a time threshold T5 to a time threshold T6) including the speech retention period, speech recognition processing is performed on the utterance signal, the utterance signal being set to the sound signal in the speech retention period, which is stored in the second memory. COPYRIGHT: (C)2009,JPO&INPIT
|
