Time-domain thermoreflectance (TDTR) method is a widely used measurement technology for thermal property determination in micro and nanoscales. However, due to the relatively weak thermoreflectance coefficient, it is quite meaningful to improve the sensitivity and signal-to-noise ratio of the TDTR system. In general, the method of extracting thermal properties is to fit the signal (amplitude, phase or –Vin/Vout signal) collected from the lock-in amplifier with higher sensitivity. In this paper, we propose a complex signal fitting (CSF) method, which directly fits the complex signal without considering the sensitivity level of the signal to the unknown parameters before solving them. As the entire output signal of the lock-in amplifier, the complex signal contains all the thermal information carried by the amplitude signal and the phase signal, so the CSF method always provides a higher sensitivity to any thermal property than the other fitting methods. Through sensitivity analysis and numerical simulation, we intuitively derive the advantage of the CSF method for various materials. To further demonstrate the validity of the CSF method, we also measured and extracted the thermal conductivity and the interfacial thermal conductance of a typical Al-SiO2 sample with our TDTR system and CSF method. The CSF method not only simplifies the fitting process but also improves the signal-to-noise ratio of the TDTR system and the accuracy of the measured thermal properties for nanoscale materials and structures.