With the increasing application of 3D sound binaural rendering algorithms,establishing an effective and comprehensive evaluation system for binaural rendering algorithms has become an important issue.Based on previous research results,this paper optimized the evaluation method of location perception,refined acoustic objective features extraction,and established a subjective-objective perception model through the partial least squares regression method to explore the location perception mechanism of binaural rendering.The results show that binaural rendering algorithms have a significant optimization effect on localization perception,especially for film and television signals.The binaural features at low and medium frequency below 1 kHz and the energy distribution of interaural signal spectrum are important features affecting subjective perception of localization.