In this paper, we present a system for the recognition of cursive handwriting that utilizes the Hough transform and a neural network. The Hough transform is a line detection technique, which has the ability of tolerating deformation, disconnections and noise. Instead of searching for linear strokes in the image, we compute global directional information at each pixel of the image. This information is stored into several feature maps. Thus, we avoid assigning to each pixel a single orientation in order to preserve useful information. Zones then process each feature map in order to estimate the local orientation of the strokes. Finally, we recognize the image by means of a neural network classifier. We have tested the system for the recognition of segmented cursive characters, cursive words and the first letter of cursive words. The results obtained are encouraging and compare well with respect to other results.