A neural network was trained and tested to provide automated classification of singing voices, both recognizing voice quality (amateur, semiprofessional, and professional) and voice type (bass, baritone, tenor, alto, mezzo-soprano, and soprano). Parameters related to singing were defined to form feature vectors. Single vowel samples for each singer were judged by six experts to establish a quality index. In a test based on a database of 2690 samples, 90% of the decisions were correct. These results show that it is possible to use neural networks to create an expert system to evaluate singing.
Click to purchase paper as a non-member or login as an AES member. If your company or school subscribes to the E-Library then switch to the institutional version. If you are not an AES member and would like to subscribe to the E-Library then Join the AES!
This paper costs $33 for non-members and is free for AES members and E-Library subscribers.