Voice: New Approaches to Audio Analysis

Presented by
Filipe J. Zabala (UFRGS, Brazil)

We introduce 'voice' package, developed to be a free and easy-to-use set of tools to perform audio analysis in R. Three main approaches are presented: (i) extract_features, (ii) tag and (iii) poetry. Such approaches aim to (i) easily build data frames containing the main features from multiple audio files, (ii) attach summarized columns from audio files to canonical data and (iii) map the best spoken words in the best order. The open datasets (Mozilla) Common Voice and VoxForge are presented to illustrate the main functionalities of (i) and (ii). Public domain audios from Internet Archive are used to exemplify paradigm (iii).