| 摘要 |
<P>PROBLEM TO BE SOLVED: To restrain, in noise estimation, reflection of such voice sections as vowel sections and suppressed voice sections in a noise model. <P>SOLUTION: A noise estimating device 10 comprises a correlation calculating unit 6 that calculates spectral correlation values among a plurality of frames in sound information acquired with one or more microphones; a power calculating unit 7 that calculates the power value of the sound level of at least one object frame in the sound information; an update decision unit 8 that, using the power value of the object frame and correlation values among frames including the object frame, decides a degree of update indicating to what extent the object frame should be reflected in a noise model recorded in a recording unit 12 or whether or not the noise model needs to be updated; and an updating unit 9 that causes the sound information to be reflected in the noise model in accordance with the decision. <P>COPYRIGHT: (C)2012,JPO&INPIT |
