The construction of acoustic emission criterion system is crucial for monitoring and providing early warning of geological hazards. In the current soil acoustic emission monitoring methods, the signal generated by soil deformation and failure is weak and experiences high attenuation, resulting in a low level of the monitored signal. One approach to enhance the quality of monitoring data is by utilizing the active waveguide model. However, the current research on the active waveguide model system is not extensive. To address these issues, a set of active waveguide system was designed to improve the data quality of acoustic emission monitoring and early warning. The deformation and failure process of loess monitored by acoustic emission was divided into stages, and the precursor information of acoustic emission for geological disasters in loess areas was comprehensively deconstructed. The data quality advantage of the active waveguide model was verified through comparative experiments of with the passive waveguide model. This study investigates the AE signal characteristics of the active waveguide model. It explores various aspects such as the AE waveform parameter characteristics, the discrimination method for failure mode based on RA-AF value, the AE r-value characteristics, the AE b-value characteristics, and the frequency-amplitude characteristics. The study reveals the evolution law of AE signals in the active waveguide model, including early warning signs and failure morphological characteristics. Furthermore, it constructs a warning criterion for the active waveguide system. The development of this criterion system is of great importance in guiding the monitoring and early warning of geological disasters in loess areas.