Wavelength selection of spectra plays an important role in the quantitative analysis of terahertz spectroscopy. In this paper, a new method for quantitative measurement of paper sheet thickness is presented based on wavelength selection of terahertz phase spectra. Besides, binary particle swarm optimization (BPSO) algorithm is applied in wavelength selection of terahertz phase spectra and partial least square (PLS) algorithm is employed to establish the quantitative analysis model of paper sheet thickness. In addition, the root mean square error of prediction (RMSEP) is analyzed to show the performance of the method. Finally, the prediction error of the method is compared with both the methods using the raw spectra and the method using a single wavelength of the spectra. The results show that the method of this paper gives the most satisfactory quantitative accuracy.