Hand Writing Recognition System for Musical Notes

Hand held devices have recently became available that can scan printed text and create computer-generated speech. These devices are generally designed for people with some type of disability. This paper describes the software portion of a similar device to identify isolated small sequences of hand-written or printed sheet music. The software we designed analyses scanned images of sheet music and identifies musical notes by kind and location on the staff. Our program was designed to identify sets which use the note durations (whole notes, half notes and quarter notes, etc.) and the note frequencies (locations relative to the staff). This paper introduces the method implemented to isolate and recollect the information from the data sets. We describe how the information was used and manipulated through statistical models in order to identify the musical notes. The efficiency of the different models was compared and those with the best accuracy were chosen to be implemented in the final program. The final program was tested with multiple data sets. This paper discussed the results of our program along with possible future enhancements.