| 摘要 |
PROBLEM TO BE SOLVED: To accurately perform signal section classification even in the case of an observation position being unknown.SOLUTION: Digital sound signals in frequency domains per prescribed time section obtained by a plurality of channels are inputted, and normalized signals are obtained by normalizing, per channel, magnitudes of digital sound signals in sound sections by magnitudes of digital sound signals in non-sound sections, and harmonic structure signals in which, in respect to one or more fundamental frequencies, normalized signals having integral multiples of the fundamental frequencies and frequencies neighboring the integral multiples of the fundamental frequencies are placed above normalized signals having frequencies other than the integral multiples of the fundamental frequencies and frequencies other than frequencies neighboring the integral multiples of the fundamental frequencies are obtained, and a feature quantity sequence is obtained from harmonic structure signals obtained for the plurality of channels, and the feature quantity sequence is clustered to determine a signal section classification to which the feature quantity sequence belongs. |
