On enabling techniques for personal audio content management

Abstract

State-of-the-art automatic analysis tools for personal audio content management are discussed in this paper. Our main target is to create a system, which has several co-operating management tools for audio database and which improve the results of each other. Bayesian networks based audio classification algorithm provides classification into four main audio classes (silence, speech, music, and noise) and serves as a first step for other subsequent analysis tools. For speech analysis we propose an improved Bayesian information criterion based speaker segmentation and clustering algorithm applying also a combined gender and emotion detection algorithm utilizing prosodic features. For the other main classes it is often hard to device any general and well functional pre-categorization that would fit the unforeseeable types of user recorded data. For compensating the absence of analysis tools for these classes we propose the use of efficient audio similarity measure and query-by-example algorithm with database clustering capabilities. The experimental results show that the combined use of the algorithms is feasible in practice.

Semi-supervised context adaptation: case study of audience excitement recognition

“…Our audio classification is based on the Mel frequency cepstral coefficient (MFCC) feature-a spectral feature used in various areas of speech analysis and in audio classification, due to its good capability to model fast and low variations of speech signal [22]. As the previous experiences of the authors show [17,23], MFCC allows to distinguish between various audio classes with fairly similar accuracy, while classification accuracy in case of using temporal and spectral audio features (e.g. spectrum centroid, low energy ratio) is often more classdependent and the overall result is not guaranteed to be much better than with MFCC.…”

SpEx: A Tool for Visualising and Navigating Speech Audio

“…More complex audio features such as identifying who was speaking, speaker emotion, and voiced/unvoiced segments were plausible, and potentially useful, additions to TAFE. However, the required machine learning training and a priori knowledge of different conditions [56] may have made configuring TAFE correctly more difficult for Peter. In turn, SpEx may become inaccurate for Jack and Amy.…”

SpEx: A Tool for Visualising and Navigating Speech Audio

“…More complex audio features such as identifying who was speaking, speaker emotion, and voiced/unvoiced segments were plausible, and potentially useful, additions to TAFE. However, the required machine learning training and a priori knowledge of different conditions [56] may have made configuring TAFE correctly more difficult for Peter. In turn, SpEx may become inaccurate for Jack and Amy.…”