The aim of this study was to propose a spectral characterization model using BP neural network based on subspace partition and PCA so as to achieve spectral characterization of LCD. Color space of monitor was divided into subspaces. Training and testing of the model was processed in subspaces. Experimental results showed that model parameter optimization in the subspace was beneficial to the improvement of colorimetric and spectral accuracy of the model after subspace division. PCA can improve the operating efficiency of algorithm and reduce spectral dimension under the circumstances of maintaining accuracy. The experiment implies that this is a high-precise characteristic model for display.