Latent topic model for audio retrieval

作者：

Highlights：

• A topic model named Gaussian-LDA is introduced for audio retrieval, which integrates topic modeling and clustering in a framework.

• The variational inference method is adopted to train the parameters of Gaussian-LDA.

• It is beneficial to model audio data using latent topic model in the task of audio retrieval.

• The Gaussian-LDA demonstrates better performance than LDA and other compared methods.

摘要

Highlights•A topic model named Gaussian-LDA is introduced for audio retrieval, which integrates topic modeling and clustering in a framework.•The variational inference method is adopted to train the parameters of Gaussian-LDA.•It is beneficial to model audio data using latent topic model in the task of audio retrieval.•The Gaussian-LDA demonstrates better performance than LDA and other compared methods.

论文关键词：Topic model,LDA,Gaussian distribution,Audio retrieval

论文评审过程：Available online 19 June 2013.

论文官网地址：https://doi.org/10.1016/j.patcog.2013.06.010