BobLewiston
In Runtime
- Messages
- 182
Can anybody recommend an OCR (Optical Character Recognition) program that can do what I need it to do?
Here's the situation: Over a period of time, I figured out the chords to about 280 songs. Yeah, I know you can get the chord progressions to songs online, but they tend to be very inaccurate. I worked on each song until I was positive it was exactly right. (It made sense to do it since I was getting paid.) And I wrote down each song, by hand, one song per sheet of paper.
Now I want a permanent, neat record of all of these chord progressions, so I can print them out as requested.
So I scanned the song sheets I had written, with the idea of then translating them to text via an OCR program, and then formatting them in Word.
The problem is the OCR programs I tried. First I tried FreeOCR. It can't handle handwriting at all - only typewriting. So then I tried SimpleOCR. This involved a lengthy exercise, the intent of which was for SimpleOCR to “learn†my handwriting. Towards this end, I wrote down by hand a newspaper article of about 500 words, and then scanned it and submitted it to SimpleOCR for analysis. This involved proofreading by me of SimpleOCR's analysis.
But SimpleOCR's performance was barely better than FreeOCR's. Here is the problem, as I understand it:
Older OCR systems yielded limited accuracy because they simply compared scanned images of characters to stored bitmap images of known characters rendered in various specific fonts. More modern OCR systems achieve better accuracy because they can recognize characters based on the context of the word in which they appear (as with the Predictive Optical Word Recognition algorithm).
But in this case, this more modern approach is actually working against me. Apparently SimpleOCR thinks that anything that is submitted to it consists of sentences, and it's trying to decipher each word based on having seen a sample, in the subject's handwriting, of that word before.
But what I'm trying to translate into text isn't a bunch of sentences composed of words. It's chord progressions. Stuff like:
G6 Asus2 Asus2/#C E Bm7+11
Am9 Bm7 CM7 D Dsus4 D
SimpleOCR is completely failing to read this kind of thing correctly. It's just spitting out gibberish.
Does anybody know what OCR program CAN handle this? Or failing that, can anyone recommend a more appropriate forum for this question?
Here's the situation: Over a period of time, I figured out the chords to about 280 songs. Yeah, I know you can get the chord progressions to songs online, but they tend to be very inaccurate. I worked on each song until I was positive it was exactly right. (It made sense to do it since I was getting paid.) And I wrote down each song, by hand, one song per sheet of paper.
Now I want a permanent, neat record of all of these chord progressions, so I can print them out as requested.
So I scanned the song sheets I had written, with the idea of then translating them to text via an OCR program, and then formatting them in Word.
The problem is the OCR programs I tried. First I tried FreeOCR. It can't handle handwriting at all - only typewriting. So then I tried SimpleOCR. This involved a lengthy exercise, the intent of which was for SimpleOCR to “learn†my handwriting. Towards this end, I wrote down by hand a newspaper article of about 500 words, and then scanned it and submitted it to SimpleOCR for analysis. This involved proofreading by me of SimpleOCR's analysis.
But SimpleOCR's performance was barely better than FreeOCR's. Here is the problem, as I understand it:
Older OCR systems yielded limited accuracy because they simply compared scanned images of characters to stored bitmap images of known characters rendered in various specific fonts. More modern OCR systems achieve better accuracy because they can recognize characters based on the context of the word in which they appear (as with the Predictive Optical Word Recognition algorithm).
But in this case, this more modern approach is actually working against me. Apparently SimpleOCR thinks that anything that is submitted to it consists of sentences, and it's trying to decipher each word based on having seen a sample, in the subject's handwriting, of that word before.
But what I'm trying to translate into text isn't a bunch of sentences composed of words. It's chord progressions. Stuff like:
G6 Asus2 Asus2/#C E Bm7+11
Am9 Bm7 CM7 D Dsus4 D
SimpleOCR is completely failing to read this kind of thing correctly. It's just spitting out gibberish.
Does anybody know what OCR program CAN handle this? Or failing that, can anyone recommend a more appropriate forum for this question?
