Feature extraction and signal reconstruction of air and bone conduction voices by independent component analysis
Proc. of International MultiConference of Engineers and Computer Scientists (IMECS2008), pp.55-59
T. Azetsu, E. Uchino, R. Kubota, N. Suetake
Oral presentation (general)
We discuss in this paper the feature extraction and signal reconstruction of air and bone conduction voices in the time domain by applying the independent component analysis (ICA). The basis functions of the air and bone conduction voices extracted by applying ICA are shown ﬁrst, and then the quality of the signal reconstruction by using several of those basis functions is investigated from the signal compression viewpoint.