Third International Workshop on Models and Analysis of Vocal Emissions for Biomedical Applications (MAVEBA 2003)
The separation of independent sources from mixed observed data is a fundamental and challenging signal processing problem. A method for directly extracting clean speech features from noisy speech is implemented. This process is based on independent component analysis (ICA) and a new feature analysis technique to reduce the computational complexity of the frequency-domain ICA. For noisy speech signals recorded in real environments, this method yielded consider-able performance improvement. Thus the process for extracting clean speech features can be performed without recovering the actual source signal.
Full Paper (reprinted with permission from Firenze University Press)
Bibliographic reference. Somkuwar, Ajay / Singh, R. P. (2003): "Blind signal separation of vocal signals taken in noisy environment", In MAVEBA-2003, 115-117.