This paper presents a model-based computationally efficient method for detecting milling chatter in its incipient stages and for chatter frequency estimation by monitoring the cutting force signals. Based on a complex exponentials model for the dynamic chip thickness, the chip regeneration effect is amplified and isolated from the cutting force signal for early chatter detection. The proposed method is independent of the cutting conditions. With the aid of a one tap adaptive filter, the method is shown to be capable of distinguishing between chatter and the dynamic transients in the cutting forces arising from sudden changes in workpiece geometry and tool entry/exit. To facilitate chatter suppression once the onset of chatter is detected, a time domain algorithm is proposed so that the dominant chatter frequency can be accurately determined without using computationally expensive frequency domain transforms such as the Fourier transform. The proposed method is experimentally validated.