This paper presents an expanded system for the online recognition of handwritten mathematical formulas. Our target handwritten mathematical formulas are strokes drawn on a data tablet. This system recognizes such strokes as components of mathematical formulas on the basis of their positions and combinations. Including matrix structures, general mathematical expressions are acceptable for this system. Each recognition result is acquired as a L/sup A/T/sub E/X source code. This system also has a preview function to enable a more highly intuitive recognition result. In recognition experiments, this system proved to be fairly feasible in handling handwritten mathematical formulas in real-time.