|
|
|
| Comparison of traditional independent component analysis with variational Bayesian independent component analysis |
| School of Mechanical Engineering, Zhengzhou Universty, Zhengzhou, He＇nan 450001 , China） |
|
|
|
|
Abstract The capabilities of blind source separation （BSS） with the traditional independent component analysis（ICA） and with variational Bayesian independent component analysis（VbICA） were discussed and verified by the experiment, The experimental results show that both methods can give a satisfactory separation performance in a noise-free BSS. However the VbICA method is superior to the traditional ICA method in the noise BSS, especially in the lower signal-to-noise BSS. In addition, the VbICA method can estimate the optimal number of source signals. However the number of source signal is always assumed to be known in the traditional ICA, otherwise the source signals can not be separated.
|
|
|
|
|
|
|
|
|
2010, Vol. 31 
