A QUANTITATIVE STUDY ON HUMAN VOICE SEPARATION ABILITY

Human brain is a well developed intelligent system which can process information with high efficiency. However, the brain mechanism of information processing has not been revealed clearly. As an example, people can separate and distinguish a minding voice easily from simultaneously spoken voices sea, while it is difficult to be realized by a computer system satisfactorily. This kind of voice separation intelligence depends on complicated brain neural network, and correlated to brain health status. If this kind of ability can be quantitatively measured, it is possible to evaluate human brain intelligence status, and even to find a convenient method to check brain cognitive function. Thus, this research had an integrated study on human voice separation ability in engineering and physiological views, and proposed a quantitative measurement and spectral calculation method on this ability. Based on large numbers of measurement experiments, the validity of this method was proved basically, and the relation between voice separation ability and human age was also found. Furthermore, through brain activation analysis and clinic validation, the voice separation ability measurement is possible to be applied in early-check of brain cognitive deficit and brain function enhancement. In this paper, parts of the results were reported.