Based on multi-fractal theory, a method of on-line condition monitoring of tool wear was presented. The generalized fractal dimensions of acoustic emission (AE) signals in cutting process were calculated using box-counting method. The generalized dimension spectrums of AE signals to different tool wear condition were gained, and the relation between tool wear condition and generalized dimensions was analyzed. The feature distances and correlation coefficients of generalized dimensions of AE signals were calculated. The classification for tool wear condition was made through comparing the values of correlation coefficients of generalized dimensions. The experimental results show that the method can be used effectively for on-line condition monitoring of tool wear.