Handwriting Recognition Uses Thinning Scheme to Reduce Data Points

Reducing Data for Handwriting Recognition

Handwriting recognition systems often generate numerous data points, especially when a pointer moves slowly across a tablet. Many of these points are redundant for capturing the general shape of a stroke. A thinning scheme addresses this by removing unnecessary data points to optimize processing.

Geometric Thinning Process

The thinning scheme operates by drawing a square around the first data point. Subsequent points falling within this square are discarded. The square is then repositioned around the next point that falls outside its boundaries, and the process repeats. The size of this square determines the aggressiveness of the thinning, allowing for control over how many points are removed.

Impact on Data Processing

This thinning process compares the position of a new smoothed data point with the last point in a thinned track. If the points are sufficiently far apart, the new point is accepted; otherwise, it is discarded. This method eliminates small perturbations and drastically reduces the number of data points, thereby lowering data processing requirements.

Efficiency and Effectiveness

Experiments show that this thinning scheme can preserve the essential shape of most strokes even with aggressive thinning. Researchers found that satisfactory results could be achieved even when approximately 70% of the raw data points were discarded, demonstrating the efficiency of the technique.