Gas ultrasonic flowmeters are widely applied in industry because of high accuracy. The time of flight (TOF) measurement is significant which determines the flowmeter performance. The traditional TOF method are easily influenced by unstable flow and noise that leads to bad results especially in large velocities. In this paper, a new signal processing method based on the double-threshold method is proposed. The peak's characteristics of the received signal are used to find the feature point on the received signal, and then using this feature point and the zero-crossing detection to complete the calculation of TOF. By comparing the new method with other traditional method based on flow experiments, higher precision and reliability are obtained. And the calibration test results show that the flowmeter can achieve 1% measurement error in the wide flowrate range from 1m/s to 35m/s.