Abstract: Real-time monitoring of the temperature distribution is of great significance to the control and diagnosis of large coal-fired boilers. Time of flight (TOF) as an important parameter of acoustic pyrometry is difficult to measure in the complex combustion furnace environment. In this paper, the coal-fired boiler of 50 MW unit capacity is taken as the research object. The background noise and the attenuation characteristics of the single-frequency signals with different frequencies in the furnace are analyzed to determine the frequency band of the acoustic signal in the temperature measurement. And, the cross-correlation time delay estimation method combined with the Hilbert transform for envelope extraction is used to obtain the time of flight (TOF). The experimental results show that the proposed method has better accuracy and stability over the non-envelope extraction method.