The impact of the number of strokes for each character in online handwriting author identification.

Abstract

We investigated the accuracy of online handwriting recognition, focusing on stroke speed and pressure rather than letter shapes. Testing 20 letters randomly chosen from a handwriting sample (2019, 401 individuals, 5 samples/person, 400px x 400px x 3), with a stroke count of a character ranging from 2 to 15 strokes, we calculated seven online indices per character. Normalizing differences in mean values by standard deviation, we assess identification rates. Findings indicate a correlation between a higher stroke count of a character and increased writer identification rates.