At present, there is a lack of real-time monitoring system for MOS value of VoLTE voice calls, effective end-to-end analysis methods and problem demarcation/location methods. With the wide application of emerging technologies such as big data and artificial intelligence, it is urgent to promote the digital intelligent transformation of network operation and maintenance. This paper proposes a voice intelligent evaluation system based on AI+ big data. Through AI+ big data analysis, model development is carried out based on deep learning. Through a unified DPI system, combined with two modules of offline training and online application, accuracy, timeliness and effectiveness of fault/complaint/performance discovery and positioning are realized. The successful implementation of voice intelligent evaluation technology fills the gaps of three supporting means in the field of communication: real-time automatic speech MOS value evaluation, poor quality event history information backtracking, and network cutover full verification.