| 摘要 |
A method comprises determining a divergence measure between a statistical distribution of audio features of a first audio track and a statistical distribution of audio features of at least one further audio track, which may be obtained from a store of audio tracks. The divergence determination may involve quantising parameters associated with the statistical distributions and using them to read a divergence measure from a look-up table. Next, a divergence measure threshold value is determined from at least the divergence measure between the statistical distribution of audio features of a first audio track and the statistical distribution of audio features of the at least one further audio track. The divergence measure is then compared with the divergence measure threshold value. The at least one further audio track is retained dependent on the comparison step and its similarity to the first audio track is determined. The parameters associated with the statistical distributions may comprise at least one of a mean vector and a standard deviation of the mel-frequency cepstral coefficients (MFCCs) for an audio analysis frame. The divergence measure may be based on an approximation of a symmetrised Kullback-Leibler divergence measure. Corresponding system and computer program claims are also disclosed. |
