Speech analysis for medical predictions based on Cell Broadband Engine

Speech signals analysis can provide useful clinical information that may be used in order to predict certain diseases. Voice analysis can be done quickly and with minimal costs, in comparison with other medical investigations, such as Nuclear Magnetic Resonance. Analysis of speech signals may be used for sorting patients who will be subject to these expensive investigations. In this paper, to perform a preliminary prediction of patients with Parkinson´s disease, we propose the usage of FLAME clustering algorithm on the speech signals acquired from the patients. The algorithm has been optimized for CBEA-based processors in order to use intensive computing resources.