Generative Models for Retrieval of Video, Audio and Text Data

"We propose a general approach to audio segment retrieval, which allows a user to query audio data by an example audio segment of a short duration and to find similar segments. The basic idea of our approach is to first train a hidden Markov model (HMM) using the given example, it is called the theme HMM. The total audio data available is used to train a background HMM. We combine these individual HMMs to form a synthesized ""background-theme-background"" HMM. This synthesized HMM can then be applied to any audio stream as a parser to detect the most likely theme segment. A major advantage of this approach is that it does not assume any predefined segment boundaries as in previous work and thus can be expected to retrieve theme segments with more accurate boundaries. We overcome the problem of only being able to use a short duration query to train a theme HMM by using the maximum a posteriori estimator with the background model as a prior model. Evaluation of the proposed retrieval scheme using short duration example audio clips of narration as queries gives quite promising results."

Name thesis:degree_name

Ph.D.

Level thesis:degree_level

Dissertation

Discipline thesis:degree_discipline

Electrical and Computer Engineering

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2015

Author dc:creator

• Velivelli, Atulya

Contributors dc:contributor

• Huang, Thomas S.

• Engineering, Electronics and Electrical

Language dc:language

eng

Identifier

(MiAaPQ)AAI3425399

OAI identifier oai:identifier

oai:www.ideals.illinois.edu:2142/81157