Gaussian frequency shift keying (GFSK) modulation is widely used in low data-rate wireless audio/video transmission and personal communication standards like Bluetooth. In this letter, least squares (LS) and maximum likelihood (ML) methods are exploited for decision-aided carrier recovery in a GFSK receiver. Through the analysis of estimation error variance, we show that the LS/ML carrier frequency offset estimation method outperforms the conventional method using the DFT approach. Besides, the new method is also effective to deal with non-DC-free data sources. Simulation results show that the new decision-aided GFSK receiver achieves better results than the feedforward receivers with DC-free and non-DC-free sources.