Detecting Solo Phrases in Music Using Spectral and Pitch-related Descriptors

Abstract

In this paper we present an algorithm for segmenting musical audio data. Our aim is to identify solo instrument phrases in polyphonic music. We extract relevant features from the audio to be input into our algorithm. A large corpus of audio descriptors was tested for its ability to discriminate between solo and non-solo sections, which resulted in a subset of five best features. We derived a two-stage algorithm that first creates a set of boundary candidates from local changes of these features and then classifies fixed-length segments according to the desired target classes. The output of the two stages is combined to derive the final segmentation and segment labels. Our system was trained and tested with excerpts from classical pieces and evaluated using full-length recordings, all taken from commercially available audio. We evaluated our algorithm by using precision and recall measurements for the boundary estimation and introduced new evaluation metrics from image processing for the final segmentation. Along with a resulting accuracy of 77%, we demonstrate that the selected features are discriminative for this specific task and achieve reasonable results for the segmentation problem.

Essentia

“…It has been used for music classification [13], and in particular for mood classification, semantic autotagging [12], music similarity and recommendation [2,1], visualization and interaction with music [1,7], sound indexing [10], musical instruments detection [6], cover detection [11], and acoustic analysis of stimuli for neuroimaging studies [8]. The systems based on Essentia/Gaia have been enrolled in the the Music Information Retrieval Evaluation eXchange (MIREX) campaigns for the tasks of music classification, music similarity, autotagging, and beat detection, and they have usually ranked among the best ones.…”

Automated Segmentation of Folk Songs Using Artificial Neural Networks

“…One class is the "vocal" where a singing voice is performing and the other is the "instrumental" where only instruments are performing. Several methods that tend to solve similar classification problems have been proposed in the past by Lu et al (Lu, Zhang, Li, 2003), Scheirer and Slaney (Scheirer and Slaney, 1997), Fuhrmann et al (Fuhrmann, Herrera and Serra, 2009) and Vembu and Baumann (Vembu and Baumann, 2005). Panagiotakis and Tziritas (Panagiotakis and Tziritas 2004) propose a speech/music discriminator based on the Root Mean Square (RMS) and the zero crossing rates (ZCR).…”