The disclosure provides a method for predicting Signal to Interference plus Noise Ratio (SINR) and a device for implementing the method. The method includes: setting a time domain threshold and/or a frequency domain threshold, dividing a time-frequency distance space into different regions according to the set time domain threshold and/or frequency domain threshold, and setting an SINR determination manner for each time-frequency region; calculating a time domain difference and/or a frequency domain difference between a channel to be estimated and a Reference Signal (RS), determining a time-frequency region to which the time domain difference and/or the frequency domain difference belong(s), and determining the SINR of the channel to be estimated according to the SINR determination manner corresponding to the time-frequency region to which the time domain difference and/or the frequency domain difference belong(s). In the disclosure, the relevant fading characteristics of an RS in time domain and frequency domain are fully considered, the SINR of the channel to be estimated is predicted combining the SINR feature of the RS that was transmitted previously by the system, and the prediction result is more accurate.