A new density- and grid-based clustering algorithm is proposed to identifying free shape clusters. The proposed algorithm is a non-parametric method, which does not require user specifying parameters for clustering. The algorithm divides each dimension of the data space into certain intervals to form a grid structure. The valley seeking procedure is employed to find the cluster centers where the data density is higher than neighbor grids and to initialize clusters. Then, the discrimination between any two clusters is evaluated by Fisher's linear discriminant, and cluster pairs which don't have a density valley between them are merged. Compared with many conventional algorithms, this algorithm is computational efficient because it clusters data by grids rather than by points. The accuracy and efficient of the proposed algorithm was verified on extracellular recorded neural spikes.
展开▼
