摘要 |
PROBLEM TO BE SOLVED: To provide embodiments that measure content similarity between two audio segments.SOLUTION: As to content similarity to be calculated between two audio segments, a characteristic generator 521 is configured to extract a simplex characteristic vector from the audio segment, and supply the extracted simplex characteristic vector to a model generator 522. The model generator 522 is configured to generate a statistical model for calculating the content similarity on the basis of Dirichlet distribution from the simplex characteristic vector, and supply the generated statistical model to a similarity calculation unit 523. The similarity calculation unit 523 is configured to calculate the content similarity on the basis of the generated statistical model. |