| 摘要 |
PROBLEM TO BE SOLVED: To make it possible to select, with less labor, a discrimination condition or discrimination function that can accurately discriminate sounds including different sounds and sounds not including different sounds, without relaying on the skill of an individual.SOLUTION: A plurality of sound data including different sounds and a plurality of sound data not including different sounds are subjected to the process of dividing the predetermined frequency range of the frequency spectrum of the sound data into an arbitrary number of blocks and calculating an amplitude average value for each block, and the process of cutting out an arbitrary amplitude range of the frequency spectrum of sound data, and dividing the amplitude value of each frequency component in the amplitude range into 10 levels, for example. For sound data subjected to these processes, the degree of increase in the amplitude of each frequency range is analyzed, thereby determining the discriminability of these sound data. Only in the case of sound data with high discriminability, a discrimination function is generated using the sound data. While the number of blocks or the like is changed, these process are performed several times. Among the several discrimination functions generated as a result of the processes, a discrimination function with the highest accuracy rate of discrimination is selected. |
