Singing Language Identification using a Deep Phonotactic Approach
This work addresses the problem of identifying languages in singing for music processing applications, representing an incremental improvement over existing methods.
The paper tackles Singing Language Identification (SLID) in polyphonic music by proposing a deep phonotactic system, achieving unprecedented performances on a large public dataset and presenting initial results for out-of-set languages.
Extensive works have tackled Language Identification (LID) in the speech domain, however their application to the singing voice trails and performances on Singing Language Identification (SLID) can be improved leveraging recent progresses made in other singing related tasks. This work presents a modernized phonotactic system for SLID on polyphonic music: phoneme recognition is performed with a Connectionist Temporal Classification (CTC)-based acoustic model trained with multilingual data, before language classification with a recurrent model based on the phonemes estimation. The full pipeline is trained and evaluated with a large and publicly available dataset, with unprecedented performances. First results of SLID with out-of-set languages are also presented.