As one of the key methods for ensuring and improving the teaching quality of university, teaching quality assessment has been widely studied. Effectively processing and analyzing the big original dataset collected during teaching procedure can provide the basis of decision support for teaching quality assessment and policy making. Considering the diversity and huge volume of original teaching data, this paper proposes a teaching quality assessment model based on deep learning technology, in which a deep neural network is implemented by MatConvNet and applied for data fusion. The model achieves good performance in university teaching quality assessment which illustrates its application value.