Quality Characteristics (QC) are divided into product planning QCs ,conceptual design QCs, detailed design QCs and manufacturing QCs. A kind of progressive reflectingrelationship of four stages QCs has been established by using Voice of the Customer(VOC) specifications as the input indicators. Through the statistics methods of Artificial Neural Networks (ANNs), the relative importance of technical specifications of VOC and the reflecting weights of each level are obtained. An extraction model of the key quality characteristics of complicated mechanical and electrical products based on ANNs technology is formulated. The model can avoid putting nonkey controlling QCs as key controlling ones in the process of manufacturing and vice versa. The correctness and effectiveness of the model are verified by application examples.