| 摘要 |
PROBLEM TO BE SOLVED: To provide an audio classifying device capable of readily and promptly discriminating noise/non-noise sections. SOLUTION: A sound determining section 4 determines that there is sound, when a value σ<SB>e</SB><SP>2</SP>, calculated in an energy dispersion arithmetic section 3, is larger than the threshold. When it is determined as being a sound, the audio information stored in a memory 5 is read and input to an arithmetic section 11 of a total sub-band energy center of gravity. The sub-band energy center of gravity Gi is calculated in the arithmetic section 11, and an energy center of gravity mean value Eg and a dispersion value σ<SB>g</SB><SP>2</SP>are calculated, in an arithmetic section 12 of an energy center of gravity average and dispersion. In a noise discriminating section 13, noise/non-noise sections are determined, by using a discrimination function, based on the characteristic that noise, such as, cheering concentrates at a fixed region of the mean value and the dispersion value. COPYRIGHT: (C)2010,JPO&INPIT |
