In order to realize the noise source recognition of carpet tufting machines, a method of noise source recognition based on modified ensemble empirical mode decomposition (MEEMD) and Akaike information criterion (AIC) is proposed. First, MEEMD is used to decompose the measured noise signal into a finite number of intrinsic mode function (IMF) components. Then the singular value decomposition (SVD) of the covariance matrix of the component matrix is performed to get the eigenvalues of the matrix. Next, the number of effective components is estimated by using the AIC criterion. At the same time, the effective components are selected by combining the energy characteristic index and the Pearson correlation coefficient method. Finally, the time-frequency analysis of the effective components is carried out one by one to realize the noise source identification of the carpet tufting machine. The results show that the method is suitable for the noise source recognition of the carpet tufting machine used for the experiment，the noise signal of the carpet tufting machine is mixed by multiple noise sources, in which the vibration of hook shaft in the coupling shaft system is the main noise source of the carpet tufting machine. The results provide theoretical support for the realization of the active noise reduction of the carpet tufting machine.