Abstract: Of the existing sound source localization algorithms for digital hearing aids, the accuracy is low while the computational complexity is high. To solve these problems, a new binaural sound source localization algorithm was proposed. First, the collected binaural sound signals were decomposed into several subband signals by the Gammatone filter, and the data was compressed according to the energy. Then, the binaural cues contained in the head-related transfer function (HRTF), namely interaural time difference, interaural level difference, and interaural coherence, were extracted to characterize the position of the sound source. Finally, the source location was estimated by the Gaussian mixture model (GMM) classifier. Experiment results show that the proposed algorithm has high accuracy, strong robustness, and low complexity.