There is no corresponding fingerprint characteristic spectrum detecting complex ensemble biological samples in liquid, in the paper, such urine of kidney disease patients as samples of the research, using terahertz time-domain spectroscopy emphatically explores response characteristics of the urine albumin in the terahertz spectrum characteristics, and combined with stoichiometric method, we find a certain kind of relationship between terahertz spectrum data and the content of urine albumin, which offsets the defects of other spectroscopy in measuring liquid protein, and in accordance with hospital clinical data. This study established a semi-qualitative method of using terahertz spectroscopy in detecting non-purification of biological liquid sample, which provides a simple, nondestructive, cheap and fast reference method in identifying the early nephropathy for medical test.