The characters of any languages having scripts are formed by basic units called primitives. It is necessary to practice writing the primitives and their appropriate combinations while writing different characters. In order to automate character generation, primitives‟ recognition becomes important. In this paper, we propose a zone-features based nearest neighbor classification of Kannada printed and handwritten vowel and consonant primitives. The normalized character image is divided into 49 zones, each of size 4x4 pixels. The classifier based on nearest neighbor using Euclidean distances is deployed. Experiments are performed on images of printed and handwritten primitives of Kannada vowels and consonants. We have considered 9120 images of printed and 3800 images of handwritten 38 primitives. A K-fold cross validation method is used for computation of results. We have observed average recognition accuracies are in the range [90%, 93%] and [93% to 94%] for printed and handwritten primitives respectively. The work is useful in multimedia teaching, animation; Robot based assistance in handwriting, etc.