New Feature For Automatic Speech/Music Discrimination

Jayme Garcia Arnal Barbedo
Amauri Lopes

Department of Communications ­ FEEC ­ UNICAMP
C.P. 6101 ­ CEP 13.081-970 ­ Campinas ­ SP ­ Brazil


This paper presents a new mechanism for automatic speech/music discrimination (SMD). Such feature is based on the concept of multiple fundamental frequencies. The performance of the feature in terms of correct classifications is evaluated for a wide variety of audio signals, and factors such as computational complexity and robustness are also investigated. The results are compared to those ones reached by previous techniques.

Download paper: [pdf] 46216 bytes.

Download presentation slides: [pdf] 191661 bytes.