Abstract: Empirical Mode Decomposition(EMD) is one of the method which is usually used to process the signal. Utilizing the idea of the Ensemble Empirical Mode Decomposition(EEMD), an adaptive phase evaluation algorithm is proposed based on the frame of Intrinsic Mode(IMF) Function combined with the sparse optimization. This algorithm uses a kind of noise assisted decomposition method to eliminate parts of noise. Then the normalization method is employed to evaluate the phase. Theoretical analyze and numerical simulation demonstrate it can persist the disturbance of noise compared with the present phase evaluation algorithm so that Sparse Time Frequency Representation(STFR) method can have a comparative better representation.