The audio signals of polyester filament and ring-spun yarn were collected respectively during tensile failure by self-built Acoustic Emission detection system (AE for short), which is based on the conventional yarn tensile strength tester. And then the Hilbert Huang transform (HHT) program was compiled by Matlab software to analysis and process the collected signals, in order to get the time-frequency characteristics of each damage source in the yarn tensile fracture process. The results show that, AE signal via HHT can be identified two modes involved in polyester ring-spun yarn tensile fracture process, which are fiber fracture and fiber slippage. And the characteristic frequencies of the modes are 30, 20 kHz and 6, 3 kHz respectively. Thus, this paper provides an effective new approach for the analysis and further research of yarn tensile fracture mechanism.