Source apportionment is a crucial step toward reduction of heavy metal pollution in soil. Existing methods are generally based on receptor models. However, overestimation or underestimation occurs when they are applied to heavy metal source apportionment in soil. Therefore, a modified model (PCA-MLRD) was developed, which is based on principal component analysis (PCA) and multiple linear regression with distance (MLRD). This model was applied to a case study conducted in a peri-urban area in southeast China where soils were contaminated by arsenic (As), cadmium (Cd), mercury (Hg) and lead (Pb). Compared with existing models, PCA-MLRD is able to identify specific sources and quantify the extent of influence for each emission. The zinc (Zn)-Pb mine was identified as the most important anthropogenic emission, which affected approximately half area for Pb and As accumulation, and approximately one third for Cd. Overall, the influence extent of the anthropogenic emissions decreased in the order of mine (3 km)> dyeing mill (2 km) industrial hub (2 km)> fluorescent factory (1.5 km) > road (0.5 km). Although algorithm still needs to improved, the PCA-MLRD model has the potential to become a useful tool for heavy metal source apportionment in soil.