Abstract: Traditional sound source localization methods are often susceptible to adverse acoustic conditions such as low signal-to-noise ratio, and it is difficult to achieve localization accuracy and real-time performance at the same time. Therefore, a sound source localization method based on windowed least absolute shrinkage and selection operator (LASSO) is proposed. The windowed LASSO was used to sparse decompose the audio signal to extract the high-energy transient and tonal components. The two components were used to calculate SRP-PHAT to realize the spatial localization of the target sound source. The experimental results show that the method can effectively suppress environmental noise and keep the positioning error at about ±10°. At the same time, the computational complexity is reduced, and the positioning time per frame is reduced to less than 1 s.