Near-infrared spectroscopy (NIRS) has been extensively developed for in-vivo measurements of tissue vascular oxygenation, breast tumor detection, and functional brain imaging, by groups of physicists, biomedical engineers, and mathematicians. To quantify concentrations of oxyhemoglobin, deoxyhemoglobin, and total hemoglobin (hemodynamics), extinction coefficients of hemoglobin (ε) have to be employed. However, it is still controversial what ε values should be used and relatively what calibration should be done in NIRS quantification to achieve the highest precision, although that the differences in ε values among published data resulted in ~20% variation in quantification of hemoglobin concentration is reported based a single human blood test. We collected 12 blood samples from 12 healthy people, and with each blood sample performed blood tissue model experiments. 4 teams of published extinction value widely used in NIRS fields were employed respectively in our quantification. Calibrations based least square analysis and regression between real and estimated hemodynamics for 12 subjects were performed with each team of ε values respectively. We found that: Moaveni’s ε values contributed to highest accuracy; Regression method produced quite effective calibration, and when it combined with Moaveni’s ε values, the calibration reduced the std/mean of estimation by two orders of magnitude. Thus Moaveni’s ε values are most recommended to use in NIRS quantification, especially with our calibration matrix based on regression analysis with a group of subjects’ blood sample.