The sound quality of the electric powertrain significantly impacts the comfort of new energy vehicles. The article uses an artificial head to collect noise samples from six electric drive assemblies, and invites non-sound-quality professionals representing ordinary consumers to conduct subjective evaluations using the pairwise comparison method. A multiple linear regression approach was adopted to establish a predictive model for the sound quality of the electric powertrain, with the average sharpness and tonality from artificial head as independent variables and subjective evaluation results as the dependent variable. Upon testing, the model shows good predictive ability in relative ranking of sound quality. This study provides a feasible model and method for the rapid assessment of the sound quality of electric powertrains aimed at the experience of ordinary consumers.