Abstract
In this paper a new geometrical approach for separating speech signals is presented. This approach can be directly applied to separate more than two speech signals. It is based on clustering the observation points, and then fitting a line (hyper-plane) onto each cluster. The algorithm quality is shown to be improved by using DCT coefficients of speech signals, as opposed to using speech samples. © Springer-Verlag 2004.
Cite
CITATION STYLE
Babaie-Zadeh, M., Mansour, A., Jutten, C., & Marvasti, F. (2004). A geometric approach for separating several speech signals. Lecture Notes in Computer Science (Including Subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 3195, 798–806. https://doi.org/10.1007/978-3-540-30110-3_101
Register to see more suggestions
Mendeley helps you to discover research relevant for your work.
