A new approach was presented in this paper to partition slub yarn into slub parts and base parts. A voltage signal proportional to the line mass density of slub yarn was firstly acquired with a capacitance sensor and a DAQ (Data AcQuisition) card. Some digital signal processing methods, such as digital filtering and re-sampling, were adopted to reduce the negative affection of sampling noises. An unevenness elimination method was presented to reduce the negative affection of yarn count unevenness, and a greedy algorithm was also presented to calculate the key threshold separating the base and slub parts. Experimental results show that the slub yarn can be partitioned effectively with the presented approach.