Fiber optic gyroscope (FOG) is a typical optical fiber interferometer. The nonreciprocal errors in its output are mainly induced by parasitic interferences and its zero bias error demonstrates a periodical characteristic. The sinusoidal noise induced by the parasitic interferences is found to be the main impact factor of bias error by using the Allan variance analysis method. By adding the sinusoidal noise to the original FOG output, we find that the “rate ramp and rate random walk” sections in the Allan variance curve are due to the sinusoidal noise. The sinusoidal noise identification method is implemented by using wavelet analysis and Fourier analysis. Based on this, the calculation methods for bias instability coefficient, angle random walk coefficient and quantization noise coefficient are proposed with wavelet analysis and Allan variance analysis and validated in experiment.