Bounded multivariate generalized Gaussian mixture model using ICA and IVA

A bounded multivariate generalized Gaussian mixture model with a full covariance matrix is proposed for modeling data in a bounded support region. For model selection, we propose minimum message length criterion. Furthermore, this paper proposes a bounded multivariate generalized Gaussian mixture model with independent component analysis. By employing the mixture model with independent component analysis, the assumed independence of the sources can be relaxed. For data with multiple sources such as functional magnetic resonance imaging and electroencephalogram databases, we propose the bounded multivariate generalized Gaussian mixture model with independent vector analysis as a generalized technique for the independent component analysis-based one. For a more insightful model analysis, we validate the proposed mixture model in data clustering through a variety of medical applications. We also propose the application of the independent component analysis-based model in speech (Romanian read-speech corpus), electrocardiogram, and electroencephalogram databases. For validation of the independent vector analysis-based model performance, different medical and speech databases are used. The results presented in the paper demonstrate the effectiveness of the proposed approaches for modeling different types of data.