1 This thesis presents a method of detecting
\emph{singing
}-voice segments in non
2 standard music genres. Growling singing voices have different acoustic
3 characteristics compared to regular singing voices. This research uses existing
4 \emph{singing
}-voice detection techniques on extreme styles like
\gls{dm
} and
5 \gls{dom
} and achieves a good performance.
\emph{Singer
}-voice recognition and
6 detection has also been attempted with similar positive results. This forms a
7 strong base for attempting lyrics synchronization and lyrics recognition in